Reading Infinite Games, they tell an anecdote about a Microsoft exec on a flight telling an Apple exec that the Zune was a way better portable music player than the iPod. The Apple exec was just “yup, you’re probably right” and then soon after the iPhone dropped
I had the first iPod, and later the first Zune. The Zune had a beautiful UI, but by then Apple came out with the iPod Nano 2nd gen and that took the cake.
I had the OG white Zune. It was the same price as an iPod ~$250. The reason I chose Zune was the Zune Pass. A precursor to Spotify. Plus growing up in the Seattle area, I knew other people who had them and we would trade songs.
100% - the iPod wasn't just good enough, it was more than enough for almost everyone. The Zune was firmly in the diminishing returns category by the time it came out.
Actually, Microsoft came with "modern" smart phone much sooner.
Actually too soon, technology wasn't there (price, computing power, size, weight, battery life).
Apple never came first ... but often just at the right moment and had marketing skills to make it a new trend.
I'm the opposite of an Apple fanboy (typing this on a Windows box), but this is some grade A nonsense.
Sure, MS launched PDAs and phone-ish devices long ago, running Windows CE and whatnot but they were awful. It's absolute bollocks that the iPhone was a splashing success because of Apple marketing. It was a splashing success because it worked spectacularly well. Random non-tech people would randomly pull out their newest purchase to show their friends. "And now it's a notepad!" "Look and now suddenly it's a calculator!" Sure, your awful HP Tablet had all that, and a call function, well before. But it sucked. It felt like using a computer while squinting, and not like a magic calculator that can turn into a notepad and then into a phone and then into an iPod.
The iPhone was a success because it worked so well. And it worked so well because the technology was there - in part because they invented it and in part because they had the taste to not bring out a shit product but wait a bit instead.
Zune was probably a better player. It was too late, and arrived when MS didn't much care.
There are emails unearthed in various lawsuits where you can read Bill Gates screaming at his subordinates: "why the hell can't our partners like Sony and Creative create a similar device? Give them all, give them early access to everything, work with them". In the end MS felt compelled to make their own.
Sony can't be bothered to compete with Apple in that era. They were trying to recover from their DRM dreams, and their devices were already sounding great with in-house software and hardware.
Creative's Muvo^2 already was the poor man's iPod with surprisingly good audio quality as well.
I feel like this completely misses the mark. Audio quality was never the compelling feature of the iPod and people weren't clamoring for it because it sounded good.
When the iPod came out you largely had two options for carrying your music collection on the go. You either carried a binder of CDs, or you had some niche player like Mini-Disc or an MP3 player. Both alternatives were expensive and had limitations similar to a CD in terms of number of tracks you could carry.
I had an MP3 player on either side of 2000 that was slightly smaller than a deck of playing cards that could use Smart Media flash memory cards. The largest card at the time was either 16 or 32mb and was enough to hold 1 album at near CD quality.
Creative's Muvo was a weird form factor that was larger than an iPod. It had a horrid interface both on device and for loading music. It's only grace was that it was slightly cheaper than an iPod and didn't need a Mac with FireWire. Although iircc this was pre USB 2.0 so not having FireWire would mean loading music took forever and a day.
The iPod allowed you to carry most, if not all, of your music collection in a package slightly larger than a deck of playing cards. And it had a fantastic interface for navigating music on the device.
This was at a time before most people had laptops and if you had a PC it was at home and used sparingly. The iPod was such a compelling mobile computing device that it drove adoption of the iMac. Apple would eventually release iTunes for Windows and USB support but that was many years later.
> I had an MP3 player on either side of 2000 that was slightly smaller than a deck of playing cards that could use Smart Media flash memory cards. The largest card at the time was either 16 or 32mb and was enough to hold 1 album at near CD quality.
I had something similar. The storage was the iPod's killer feature, along with iTune's $0.99 songs. Suddenly you no longer had to buy whole albums, and you didn't have to swap out what was on your MP3 player every day when you wanted a different playlist. A 5GB hard drive in your pocket was a huge innovation then.
"1000 songs in your pocket" was the entire driving force behind the thing. It was constantly bellowed in the marketing and it's what Steve Jobs demanded of the engineering team from day one. The size and the storage were paramount and they knew no one else could match it.
I had a Rio Volt which somewhat bridged the gap. 700MB mp3 CD/RWs (about 10 hours of music per disc) and a CD player when traveling and picking up new music. Not as small as an iPod, but no book of CDs was necessary, either.
In some ways, anyway. Never owned a Zune myself, but a university classmate did and I was shocked by how poorly it handled non-Latin languages… she had a ton of Japanese and Korean songs loaded onto it, and their metadata all displayed as "missing character" blocks. She used it a lot like one might use an iPod Shuffle despite it having a nice screen because the only way to tell what was playing was by hearing it play.
By contrast my 4th gen B&W iPod which was about 5-6 years older handled unicode just fine.
“Squirting” was ridiculed at the time and the word itself scaring away women. Jobs killed it then in an interview with a classic quote: Microsoft = Cold tech and Apple = Humanity. MS scares her away, Apple gets the girl.
> QUESTION: Microsoft has announced its new iPod competitor, Zune. It says that this device is all about building communities. Are you worried?
> Steve Jobs: In a word, no. I’ve seen the demonstrations on the Internet about how you can find another person using a Zune and give them a song they can play three times. It takes forever. By the time you’ve gone through all that, the girl’s got up and left! You’re much better off to take one of your earbuds out and put it in her ear. Then you’re connected with about two feet of headphone cable.
I just want to take a minute and be grateful for getting the chance to live through the 90s again. I always felt a bit sorry for myself being a child through the og 90s. Now I feel like in a decade or so into the future I will look back and be happy I got to live in the naïve days of windows 95 again, as an adult this time. I really appreciate it.
I don’t get that sentiment. I have an m4 16gb and it’s such a sluggish machine. and that’s not even doing development, just browsers, word, excel, PowerPoint. (And so many eternal bugs… I feel like I’m on windows)
I have an M2 Pro MBP 16GB, compared to the latest MBP issued by my employer a couple months ago, I see zero difference in real world performance. Can’t speak for Office on Mac, I have not used it in over a decade, but for typical full stack web dev I literally don’t need anything more.
It's been really weird reading laptop reviews over the last few months. I've seen a bunch of reviews where they have their usual bar graphs comparing a bunch of laptops, and the laptop that is at the bottom is an Apple. It's the really inexpensive Neo, of course, but even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops.
I assume the M6 will take the crown back and then a few months later Intel/AMD will release a new chip and take that crown back again. That's the state of the world we used to expect, but it's a state that has been missing ever since the release of the M1 in 2020 until Intel finally caught up again this year.
> even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops.
When plugged in... This caveat is so enormous it should almost be legislated. If your computer use is at all portable, a computer that scales down to 20 - 40% of GPU power when unplugged is an enormously significant factor. So far as I'm aware (could be wrong about arm devices?) there's no non-apple laptop that operates at 100% speed on the road.
Newer Intel Panther Lake chips perform the same or very similarly on battery as they do when plugged in - as do Qualcomm chips. However, I don't know where they're seeing "consistently beat on performance and battery life by Intel Windows laptops". Multi-core performance can definitely beat M5 in some configurations but single core performance is still fairly far behind and battery life is comparable again depending on the exact specifications and design of the laptop. I've seen analysis showing M5 is still the perf/watt king though regardless of configurations.
Sorry, poor wording. It's the Neo that's on the bottom in many reviews, but the M5 is regularly beaten by Windows laptops. My usage of "consistently" was in the sense of regularly beaten, not in the sense of always beaten.
Because the naming convention was used for chips based on Skylake microarchitecture (but with other differences such as process node). Not changing the name implies they've either changed the naming convention (why?) or they haven't improved the microarchitecture since 2012.
> but even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops
The real question is whether all these Intel Windows laptops spin their fans at full speed when doing absolutely nothing. That’s something I can never go back to. I do some light gaming on my M2 Pro MBP and it gets hot when trying to push 120 fps. But my Lenovo Legion (that’s now collecting the dust) is so loud I could hear it through headphones.
> [...] but even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops.
Hold up, in what metric/benchmark? Feel personally like we life in an age where, no matter the SOC vendor, something high performant and efficient is offered, so seeing a claim that any vendor, be it Intel, AMD, Qualcomm or Apple, is consistently outperforming another, I'd like to get more context on that.
Intel were x86, Apple Silicon is ARM-based. ARM-based chips are more power-efficient.
Also, it's built on a much smaller process. 3nm, if I'm not mistaken, older Intel was something bigger than 10nm. Heck, if you take a really old Intel Mac, you have something like 65nm process, which is much less efficient than 3nm.
> It's the really inexpensive Neo, of course, but even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops.
Of course you can beat the entry level MacBook Neo by comparing it to larger, more powerful, more expensive laptops.
The M5 is an entire family with a range of performance. Intel/AMD have done a lot to improve performance but they’re not beating the high end M5 chips on performance or battery life yet.
It's also not hard to find a Windows laptop that beats the M5 on battery life. Single-thread performance is the only remaining measure where the M5 is king.
x86 is way more complex when compared to ARM processors, as a result their TDP is way higher when you request performance from them.
Intel had to reduce the frequency of their processors when running AVX2 instructions and the AVX2 frequency of the processors were non-disclosable to anyone.
Also, benchmarking Intel processors and publishing these numbers were forbidden in some cases. I don't know whether this ban is still in effect.
x86 processors can't keep up with the ARM processors TDP and thermal profile wise. So they slow down a ton when running on battery. See Jeff Geerling's last video on Apple Neo vs. some Intel laptop. It's as "efficient", but slow as a newborn tortoise learning to walk when unplugged and trying to get the most endurance out of the battery.
My M1 Mac gets almost 2 days of low-intensity use after ~6 years of use, and it got warm once or twice because something ran away in the background for tens of minutes.
Being mistaken on the CISC vs RISC debate (besides, modern x86 is closer to RISC via micro-ops then old school CISC) is understandable. There is a lot of misinformation out there and myths, plus, it just feels right to consider CISC overly burdened, etc. TLDR: Intel x86_64 is closer to RISC then you likely think and Apple Silicon arm is closer to CISC then you likely think. These lines are blurry and have been for decades, very much for good reason.
> But talking this authoritatively on something without doing the reading, that's grating...
Thanks for your prejudice on me without knowing anything about me. In short, I'm a HPC sysadmin and programmer who works in a HPC center, separated from the actual hardware by a couple of floors.
We can discuss how transistors' heat generation doesn't discern about ISAs or being in a DAC or a cutting edge microprocessor, and we can even discuss how implementation of some functional blocks generate heat regardless of the ISA being involved. If you want we can discuss how saturating memory controllers affect pipeline saturation in processors even...
But talking this authoritatively on something with that amount of prejudice, that's grating.
Because being too confident about someone you don't know is a bad habit. Where I work doesn't matter though, but what I do is.
Pointing me to C&C is a nice touch though. Not only I read the site and very article you sent me before, I used to consume Anandtech before that.
As a mere mortal, I can make mistakes and gladly accept them, but I can't accept rude replies. Pardon my French, but being called a low-key liar or smoke blower gets me a little upset.
> Pointing me to C&C is a nice touch though. Not only I read the site and very article you sent me before, I used to consume Anandtech before that.
All I'll say is, that's worse then. Presuming you had not read up before promoting a long disproven myth, that was an assumption by me, I'll admit that and maybe I should not have done that, my mistake. But it was a gracious mistake, it was done in your favour, it was giving you credit.
- Apple: We have matched Intel in CPU performance thanks to this new PowerPC! (Shows ad that uses carefully handpicked benchmarks to suggest that the PowerPC is actually faster when it really isn’t on average)
- Intel: Oh, we just found a 30% clock rate increase in the pocket of our other fab pants.
- AMD: Hold my beer, I have the DEC Alpha guys making an x86 CPU… How about 64-bit while at it.
I said late 90s though. The insanity which poured through early 2000s.
5.25GHz Pentium 4s, intentionally lower binned Athlons, the era your CPU got obsoleted the moment you booted it for the first time.
I don't remember early 90s much. I was too young back then. I don't remember much stuff from that era. But late 90s, early 2000s.
Oh, boy.
P.S.: AMD64 was a great sucker punch though. One of the professors in our university rejected to believe and got mad when he learnt that Intel licensed AMD64 from AMD, heh.
That's sadly true though. Also booting something was really simple.
Now we boot an embedded microcontroller (or CPU) which boots the main CPU which boots another OS semi-persistently to boot the main OS (if it's allowed).
Sometimes there are other processors needs to be up to allow processor to continue booting as well (these are mostly servers, but eh).
To be fair, this kind of dominance is not unprecedented. Intel was even further ahead in the early 2000's. Every new competitor process was further behind the leading edge and not closer. TSMC started launching half nodes like 28nm just to have something in the market that would sell.
But then you started to see the cracks. New competitors would launch new products with very slightly better metrics than Intel's older stuff, just to be, heh, meep-meeped at the next press conference. But the overlap was real, if small. And it grew over time until everyone looked up around the 5nm node and realized Intel had lost.
That's where we are right now with Apple. "Funny in a way", sure. But history says this is more likely to be the beginning of the end. Everything goes in cycles.
I think it's great but it's important to remember that we haven't seen Xiaomi's chips in actual devices under real world tests. We don't know if the speeds are sustainable, under what wattage, etc. Competition is still great and I look forward to learning more, of course.
It's interesting to see how defensive some of these pro-China posters get on HN and elsewhere. I'm genuinely curious why there seems to be an inferiority complex here with respect to America/American companies.
> Xiamoi is some random chinese company not known for high end chips and was able to catch up impressivly in a short period of time
Xiamoi is a huge company and is a commonly known brand. They’ve been making chips for a long time. They didn’t start a few months ago and catch up with Apple on their first try.
Are they well known in China? I'm seeing that Chinese tech has decoupled from the West. They have all this cool stuff that we do not hear about because they aren't selling it to us because they don't need to. It's usually been stuff with better price to performance or ultra low prices though, rather than ultimate performance stuff. I wouldn't be surprised if they made a top end chip that everyone in China knew about, and we didn't.
That might be the funniest thing I read all day. Xiaomi is a massive company that makes all sorts of things, including a lot of pretty high-end mobiles, and has an annual revenue in the order of 75 billion US.
It’s no Apple, but it’s not exactly “some random Chinese company” either.
I find it funny not because I support Apple. I find it funny because it feels like the leapfrogging happened in early superscalar CPU evolution. The era when the Moore's Law was working.
I enjoy it because of progress, not because of Apple.
Even w/the pricing spike, inflation adjusted we are back to roughly the prices of a new Mac SE/30 for something that can beat a turing test w/o sweating.
I yield the floor to no one when it comes to pessimism, but that's incredible.
I don’t think the effects on the hardware market of open weight models triumphing has been reflected upon enough. It’s not clear that it will be less impactful if every enterprise decides to build for predominantly on-premise inference. In fact, before the question is ultimately decided, we’re probably heading towards a few years where hobbyists, enterprises and ‘hyper scalers’ are all competing for certain parts in common. Ram booked through 2027 sounds about right.
I built a new PC about two years ago, and I probably got it at the last possible opportunity for a while. CPU and motherboard have come down by maybe £100 in between the two of them, but a 7900 xtx (or any other 24GB GPU) for under £1,000 now seems like a bargain, and £180 for 64GB of DDR5 makes me feel like an old man talking about the halcyon days.
Welcome to the club. If you're _really_ competitive in cs2, I'd swap out to a 9800x3d setup, but it's still a maybe. Very little reason to upgrade right now other than to run LLMs.
It will never make sense to me to run Llms locally unless I had 50k. I don't even think that would compete with price perf of a remote llm and getting business done. And that's completely ignoring that sol/fable level is not local
I'm expecting it to somewhat collapse. I don't know if it'll go back to pre bubble prices (here's to hoping), but I do expect a pretty sharp decline around 2030... probably not before then.
Basically everyone that makes memory is building new fabs, meanwhile I'm not sure how much longer AI datacenter demand for ram will last. I think the decrease in AI ram demand and the new fabs will likely coincide leading to a collapse in pricing.
That is, of course, assuming the memory manufacturers don't pull their favorite trick and collude.
If there's a decrease in AI ram demand, it will not be because the models get better. Models getting better will increase RAM demand, because it grows the part of the economy that models are useful for. Classic Jevon's Paradox.
Cycles tend to be 5-10 years long not 25. Without knowing anything else I would expect a fab you seriously start planning today will be at full capacity in about 5 years. Nobody serious likes delays - in particular the banks don't like loaning money that won't at least start paying off. They know it takes some time to design a building - but factories typically are standard buildings so once you know about the size you can get it done fast - I expect 1 year to have the building done is the worst case (and it can be done in 3 months possibly if your project management is good - after interest this is cheaper than the 1 year). It takes time to build and install the specialized machines that go inside - this is the largest problem, but you typically order them first and then plan the building around the needed space and when they will arrive. Then you need 6 months to setup the inside of the building. From there it is just ramp up time.
The above is a standard project management problem. We do this for lots of industry all the time. There is every reason to think you can get a new factory running in 5 years.
Note that I said 1 factory above. Some of the special machines we don't have the ability to make them fast enough to do 2 (I don't know the real number!) new factories in 5 years. Existing factories are using most of the special machine capacity to replace machines that wore out on the way - this can be corrected as well, but it adds another year and the expenses are much larger. Realistically though 1 new factory is likely enough.
If you don't understand that the AI driven memory boom has totally broken the cycle you are going to lose a lot of money. There is infinite demand for Intelligence and that translates directly to chips.
What does that have to do with the Turing Test? The TT has clear rules: There are judges that have a dialogue with anonymized AI/humans. The humans cannot cheat and impersonate a machine, they have to act normally. The AI obviously should try to sound human.
No AI would pass this test with experienced judges.
I think that if you have a long enough chat, yeah, I think you can figure out who's meat. The original rules for the TT specified a short interaction, but I can probably accelerate it by pasting in large code snippets to force early compactions.
The test doesn't say that the judge has to be experienced. But I also don't care if some random gullible person can't tell the difference. Nothing passes the Turing Test for me yet.
Edit: Also doesn't say anything about who the human test subject is
Of course, and I'm sure OP wouldn't disagree with you, it was clearly a joke for emphasis. Some people round here need to clean and calibrate their humour detectors more often.
It was always a bad test, despite the greatness of Turing. The human organism is built to 'project' humanity onto anything available; apart from this none of the peculiar phenomena of the so-called 'modern human' is even intelligible, even the possibility of science. I bring all that is in me onto you as soon as you seem to be saying something, and reciprocally. We do this at the drop of a hat, and all specifically human life depends on it. But this power shows its 'gullibility' with 'gods' as also with Eliza. I am not snide about it because it is overreach by something the significance of which is overwhelming , but one is indeed amazed by the failure to reflect on the part of the ones eg giving LLMs rights - to take extreme case of a very widespread cultus - as if /they/ were the rational party, not ancients placating the storm god.
Yeah...but in context, "gullible" seem a bit pejorative. Humans are also hopelessly incapable of sensing radioactivity, methanol in their alcoholic drinks, carbon monoxide, and a great many other things that our ancestors just didn't encounter much.
Though we're pretty good at sizing up a person's emotional balance/maturity and competence at familiar tasks. So maybe have an old blacksmith watch the AI/robot interact with horse owners for a while, then shoe their horses, and see how well it does.
Nowadays, I prefer AI CS agents to humans. I just had a chat with an AI yesterday, it understood me perfectly even when I made mistakes, I was impressed.
In contrast, humans tend to paste me the same barely-relevant macro over and over, no matter how much time I spend explaining my issue.
Yeah, at least LLMs read everything you write (for now). Human first level support agents are incredibly frustrating if you have to explain anything with more than one logical step.
Customer service is very different. Crappiest audio quality possible & scripted answers all the way down.
Almost like humans are forced to behave like machines.
Yes, which is exactly and entirely the point of the whole paper. Somehow missed still -- despite how important AI has become, shockingly few people actually read the short, layperson-accessible paper that started the whole field.
The Turing test is more complex than what gets suggested.
And the "popularized" version is faulty also since it uses an ideal, abstract human judge (like the "spheroidal economic agent").
But if you want to add declinations to the said popularized image of the Turing test, you may add Maxim Lott's IQ tests at trackingai.org . Between the end of 2024 and the beginning of 2025 LLMs reached an equivalent IQ of 100, for example.
It depends who takes the test. I am not yet, to my knowledge, fooled by AI.
I've tried [1] and I almost 100% detect which is the AI. I really want to convince myself I have failed, does anyone know of a better site/resource for this?
I know it might be moving goalposts but I would consider AI to have passed in a well and truly undisputed manner when [2] is resolved.
But in a more practical sense, if AI can impersonate humans so well today then why are state of the art frontier models so obviously AI when they create PRs, commit messages, documentation, etc. Are the companies deliberately making them unnatural?
we might need to bring back the Voight-Kampff test. anthropic at the very least is introducing a water making system to Claude which might make them more identifiable to humans as well as much easier to detect for machines.
Not sure where you're quoting from but if it's the metaculus question comments, many of them are from 2023. The consensus is it will resolve in 2029. I believe it will not resolve before 2035.
Sure, when you look a little wider, since 2000 we have seen the following major improvements:
- Extreme poverty has dropped from 30% to under 10% globally.
- Child mortality rates have dropped in half
- Internet access has exploded from 10% to 70%
- Solar energy costs have dropped 90%
- Cancer death rates have declined by 30%
All of these massive improvements in less than 30 years.
While there certainly are issues to solve, and if you simply follow journalism you may think the world is worse off, but for many, their lives have been significantly improved.
It's a Substack that reports good things happening around the world, divided into sections like "Conservation and Restoration," "Climate and Energy," "Medicine," etc. And they also give part of their profits directly to projects in those categories.
(I'm not affiliated with them, I'm just a subscriber.)
These improvements are showing that we're producing plenty so that the cost is driven down and distributed to wider and wider portions of humanity.
But my personal observations of AI is that it's producing more and more of the same stuff and not moving the front forward much. The human innovation and invention seems to be lost.
Thanks for sharing this sentiment and including data. So few people seem aware of the wonderful progress humanity keeps making. Makes me worry that the progress will stall or even reverse because people don't even know it's happening.
In part, sure. From drug discovery to crop yields to education, compute and AI have material benefits. It's kind of shocking that anyone could doubt this.
This being HN, I hasten to add they also have massive downsides, we're all doomed, nobody programs the right way anymore, those poor people just think compute & AI are improving their lives, etc, etc.
AI in the sense of LLMs is too new to really make an impact yet. But it is compute driven. Modern science and engineering would be impossible as we do it now without high-end compute.
Just like the computer revolution, it will make a small number of people richer and put everyone else at their mercy. In real terms the average person is far poorer than in the 20th century. Used to be able to buy a home and support a family on a single income. Now it can take two just to survive.
In real terms, the median human globally today is vastly richer than at any point during the twentieth century.
And the median American is also richer in real terms, both in terms of wealth and in terms of income.
Of course, that all assumes that you use a reasonable measure of inflation that’s stable, well-designed, and applied methodically and consistently over many decades.
Alternatively, you can cherry pick data points and go based on vibes, which lets claim whatever you want!
Why in the world are you choosing to live that way?
I'm writing the best music of my life, realizing games and art projects I never had time for, and writing higher quality software in addition to dramatically more of it. Who has time for pablum?
How exactly are you using these tools that you have that experience?
It's a grovel-for-investor-dollars site that we've long pretended is for serious technical discussion. Comments should always be filtered through that lens.
I run a lot of local models (I am always experimenting) on my 32G M2-Pro MacMini - I would love to upgrade.
The financial aspects don’t work however: I can learn and experiment with what I have for local models, and I pay as I go on FireWorks.ai for open model inferencing and no matter how much I use this service my monthly bill is between $10 and $40 and much faster than any reasonable home rig.
Hybrid ‘small local’ and buying inference is the way I choose.
Apple Studio with maxed out M5 Ultra, 256GB RAM and 16TB storage is 18,299$. The 512GB RAM version apparently is coming in October, considering that the difference between 96GB and 256GB is priced at 4000$, the 512GB upgrade must be eye watering.
So, on the mini the RAM upgrade runs at 25$ per GB on all tiers, the same as the Studio therefore the upgrade to 512 will probably cost 6400$.
The fully maxed out Apple Studio then will be 24699$. It's 17199$ if you don't upgrade the storage(1TB).
So is downpayment on a house. I would buy the house and just pay for tokens as needed. The house will get more valuable and that wealth would buy a lot of tokens in the future - which will probably get cheaper.
EDIT: or buy AAPL. If I had bought Apple stock instead of buying a Mac LC II in 1992, then I would have about $2 million in Apple stock.
Or more simply – $25K (+ tax) put in a savings account will earn about enough interest to pay for a $100/month AI subscription indefinitely. And at the end of it you still have the $25K.
Not 'more simply', there are basically zero savings accounts that are going to net you a 5%+ interest rate to give you that $100 a month. And that $25k becomes less valuable over time. $25k is now only worth $19k because inflation.
It’s basically impossible to compete on economic terms with deeply subsidized hardware that is widely available to rent or as a service with zero commitment.
For general inference there’s no ROI that makes this work vs subscriptions.
25k for computer now, plus 9-10% sales tax, plus operating cost, plus time and cost for R&D tinkering with models, harnesses, and infra (assuming highly capable engineering talent that can get paid for your human inference) vs a HEAVILY subsidized subscription at 200 per month with free R&D has a pretty long ROI (15 years?)
At API costs, it’s like 6 months if you’re heavy on inference.
For training, specialized models will have their own ROI that makes this worthwhile. Then debate renting capacity and the platform to choose
I don't need privacy, so it would be financially imprudent for me to spend 20 grand on such a machine. But I have a financial management client who does need such privacy, and if I get more fully engaged with them then I would be able to justify getting a loaded Mac.
I tripled my money on my RAM purchase of three years ago. So, yes, for short-term appreciation that's hard to beat. But I don't think it's something that will continue.
Apple hasn't been selling just ram for a long time, they sell vram. Try getting 512 gb of HBM on current Nvidia cards - it's gonna cost way more than $ 24k. And here you get the same amount of memory for weights right in a quiet unit under your desk
Put together a similar build with a couple of rtx 6000 Ada cards and Apple's price tag suddenly looks pretty damn reasonable
HBM itself is very expensive but it’s not really fair to compare to LPDDR or GDDR
They’re very different things.
The more logical argument to me is that Apple uses its upgrade price points as more than just direct BOM and rather as a proxy for things that are amortized across all their sales like support/warranty/etc so higher SKUs subsidize the costs of the lower ones.
i remember when SGI boxes were $50k and then literally worthless just a couple bears later. i remember my university had a pile of them for free outside the deans office.
I think the reason they offer these options in the first place is the discontinuation of the MacPro and their remaining need to offer high end solutions.
I don't think 16TB storage is a right choice. Going 2TB and it's 11,299. Probably, if you buy the storage and install it yourself you can go higher and quite cheaper.
Ten years ago a bought an expensive MBP because I do a lot of stats in R, Python etc that benefited from it. But the next Mac I’ll buy will be a much lower-end model, because it’s just easier these days to do that work in notebooks in the cloud.
I live right next to a micro center and remember when they started offering that deal... still so pissed at myself for not buying one. I ended up just buying a raspberry Pi for what I was doing, but seeing as where the prices are now, I messed that up a bit. Also my worst sin was not buying 64GB of DDR5 when I was doing my computer upgrades back in August last year.
I returned an M4 Mac mini, 64GB, unopened... because I thought it was excessive for my needs then. I swear it'll be one of the things flashing before my eyes when this all ends.
Rumors say that Apple will only release M6 base variant and skips M6 Pro, M6 Max and M6 Ultra variants to concentrate all efforts to create a good AI capable M7:
"According to reports from Bloomberg, Apple will be skipping its M6 Pro, M6 Max, and M6 Ultra chips to accelerate development of the M7 chip. That means the only chip to be released from the M6 family will be the base M6.
The reason for this break with tradition: AI. Apple had been planning major neural-processing upgrades for the M7 family and ultimately decided those improvements were important enough to justify accelerating the next generation rather than completing the M6 lineup." https://9to5mac.com/2026/08/08/apple-m7-chip-heres-why-it-ma...
I'd skip M5 and M6 chips for LLM work and wait for a year for M7.
Apple still has the best hardware so I moved to it for the last few years, but the closed software ecosystem is terrible for taking advantage of it.
I wasn't able to debug network errors (restartin my Mac worked), Metal was missing low level disassembly / debugging tools (there is some hard to use UI), but the worst thing was the inflexible windowing system.
Even getting all the window handles on all screens/desktops with their titles and programs is impossible.
I just decided that I move to Omarchy 4 (basically Hyperland + QuickShell) + NVIDIA GPU, and I already was able to customize it more than my Mac in years.
I will miss Apple's hardware for sure, but not MacOS and the missing hardware documentation
I had a chance to try Omarchy past few days and it's just very different vs macOS. I honestly never had a problem with the windowing system ever since I built my own customization scripts (i.e. Hammerspoon). I can see the appeal for someone who wants ultimate customization though but macOS still wins overwhelmingly when it comes to polish, ecosystem, user experience, and apps (nothing comes close).
It’s not just Omarchy, there’s really not much out there in the desktop Linux sphere for those who are mostly happy with how macOS works out of the box. Everything is either in a similar vein to the Omarchy setup (hyper-minimal tiling WM), Windows-like (KDE, Cinnamon, most other DEs), or a chimera with a grab bag of design bits from every desktop and mobile platform (GNOME, Pantheon, COSMIC).
It’s a bit depressing because it means that if I ever feel forced to switch my daily driver, it won’t come without a dump truck load of friction, frustration, and lost productivity, which I’ve validated by using the various Linux desktops on secondary machines.
There were 1000 plugins created for Omarchy 4 in 2 days. That's why I don't feel it being hyper minimal anymore.
It's still not well integrated of course as those plugins are from different people, but I at least don't feel powerless as I know I can make any change easily.
It's interesting because I just haven't felt the polish.
For example when using PyTorch I wanted to try to speed up my NN kernel by 2x by just using half precision and haven't noticed any speedup at all. Also I was missing the easy to use GNU tools that had to be mixed with Apple's tools.
I loved using Arc browser as well, and I'm missing it, but I guess I will do without it somehow (Chrome's vertical tabs are just not the same).
My main program missing from going back to Linux was ChatGPT Desktop, but now it's there.
I just checked out Hammerspoon, I'm happy for you that you wrote it, and looks great, but it has the same problem that I had: for security reasons Apple stopped allowing the window APIs to get all important information on other workspaces. You can only do it with Accessibility API. I was trying to fight with it but have up.
If the browser being Chromium-based isn’t a hard requirement, it may be worth checking out the Firefox-based Zen Browser[0]. Its UI is very similar to that of Arc, to the point that I’d call it Arc’s spiritual successor.
How are you liking Omarchy? I saw a video on it recently, and it looks 'pretty' but still looks like it's a lot of memorization of shortcuts and feels like the 40% keyboard of OS's. Like some people it's absolutely amazing, but lets be honest, it's going to be really difficult to be as productive as a full fat keyboard.
I ordered an ASUS Zephyrus G16 with 5090 NVIDIA card + 1.9kg (quite an overkill, and I know that I will have to limit power output), but hasn't arrived yet.
But what's fun is that I love QML+QuickShell with its hot reloading, Hyprland with its Lua support.
With AI nowdays it's just so easy to do deep UI changes that wasn't possible a year ago.
The 512GB Ultra is amazing, sure. But who is it for exactly? VC funded big spender founders? In that case why would they need local AI? The ultra rich enthusiast? But there can't be too many of those. So who actually buys these?
I think there's some mental inertia around what a computer is and what it's worth. This thing can build custom software for you, mostly autonomously. It can monitor things happening on the internet that are relevant to you, in a holistic and flexible way. We have one crawling the web for local events we'll like, and it judges them based on what it knows about us, and it tells us about the best matches every weekend, which has yielded some awesome outings we wouldn't have known about. It reads the literature on a subject in seconds and uses it as context to help in decision support. It's not the same value proposition as a computer 3 years ago, where most people are mentally anchored on what a computer should cost. Having it at home means that you can use it as a personal agent that always puts your interests first, regardless of what ad model the commercial providers decide to put in, and you can stash in it your medical data, what you buy, what you make, your worries, hopes, and dreams, without worrying about that being used as training data, or worse, something to exploit you commercially. I think it'll become considered totally reasonable to consider spending the cost of a small car on a computer, for many families.
Also, a lot of companies are looking at how to run capable models locally to cut some of their (massive) cloud AI bills. An easy answer is worth a lot to them.
By this kind of logic we'd paying something like ~$10,000/month for our internet connections. The perceived value needs to exceed the cost but that does not make it the only factor to consider price with.
What makes this expensive & sell well is it's not very fungible at the moment. Where else are you going to get 512 GB of high speed memory with a well supported accelerator attached that you can throw in the corner of anyone's home and not really have them notice? There are plenty of lesser options, plenty of noiser/power hungry options, plenty of harder to support options, but not really something in direct competition at the moment. Even the next rounds of the integrated AMD/Nvidia solutions are only targeting 196 GB of much slower memory and compute.
I think the difference besides the supply crunch there is that everyone connected gets the ~same internet, just faster or slower. Quantitative, not qualitative difference. On the other hand, a computer that can run Gemma 4 8B versus one that can run DeepSeek Flash are different enough experiences that I'd say they're effectively a difference in kind. It's been a bit since we had such serious stratification in outright capability in computing, rather than just how long it takes to get something done, or how many of something it can serve at once. In the early 90s, I think there were a lot more of those "this computer can do this thing, this one just can't" scenarios.
Closest competition I see right now are stacks of 2-4 connected DGX Sparks, similar lowish speed high mem, and about the same cost/gig.
I like this train of thought. The inverse is saying that the cost of this computer is the value we give away to AI companies by doing compute on their servers with our data. And to take it another way, is the value to you, the cost of a small used car?
> And to take it another way, is the value to you, the cost of a small used car?
For me personally, not quite that valuable yet, but I think it's getting there quickly. Deepseek V4 Flash massively increased the value of local AI to me, to the point where it's displaced most of my Claude Code usage, its upcoming vision enabled version should bump it further, and it's only going to get better from there.
It's a lot faster, but a lot of it is also feeling free to discuss things I wouldn't be comfortable sending to Claude, with the idea that that info is now theirs in perpetuity. I got my genome fully sequenced recently (it's cheap now!), and I get a battery of blood tests every year. Wouldn't do processing on any of that with Claude, but local AI? Totally great.
And if I was running a company with a large cloud AI bill, I'd probably buy a wheelbarrow full of these macs. Cheaper, but also a more solid/predictable base to build on.
Enthusiasts buying these for fun are not the target market. These aren’t big sellers to begin with but a lot of the sales are going to companies where people have budgets for gear like this and can make a business case for it.
This is, sadly, probably a foreign concept to a lot of people who have only worked at companies where hardware purchases are viewed as something to minimize and everyone is stuck with the same low spec laptops that the finance department picked out. At companies where someone might have a legitimate use for a $20K machine, their fully loaded costs (not their salary) are $300K or more, and other teams like sales are spending thousands of dollars per week on things like travel and hotels for their job, spending $20K on a computer that’s going to last several years is not a hard choice.
If I can get Sol level capabilities on a $20k machine, then it is well worth it for my employer to buy me that machine for work as a workstation. When you start paying in tokens vs subscription costs due to enterprise agreements, you really start to see how much cash utilizing frontier models at the frontier costs (and I'm efficiently using luna and other models where possible!)
Yeah, I'm not sure people realize how expensive ZDR/Zero Data Retention is, and how important it is to a lot of businesses, this kind of thing starts looking really cheap really fast if it's a reasonable substitute.
Even using multiple windows in parallel for as many as 5-10 hours per day, I find that I am not fully using my claude max (20x) and chatgpt pro (20x) accounts. I can for sure use up the claude max account, but chatgpt either gives me a free reset before I run out of tokens or I just fail to use the full quota. The quota for Sol seems like 10x that of Claude Opus at the same level, and forget Fable, you can use a 5 hour quota in 20 minutes.
But lets do the math:
Lets say a 20k workstation can run 1 inference at a time at the same speed you get with Sol hosted by openai (big assumption) and run an equally capable model (big assumption).
Each month this gives you about 100-170 inference hours on a Sol 20x Pro account, and 720 hours (if you utilize 24/7) on the workstation.
Assuming a 36 month amortization before the workstation has to be replaced due to no longer being able to run frontier models or is too inefficient due to electrical costs or what have you:
The monthly workstation cost is about $550 capex and $150 electricity -> $700/month
You would need about 6 Pro accounts to reach that capacity, which would cost you $1200 a month.
But this fails because:
- You most likely can't utilize the workstation 24/7. Your work hours will be concentrated into 6-10 hours per day.
- During work hours you are capable of utilizing more than 1 concurrent session. 6 Sol accounts would support as many as 20-30 during working hours, not all the time but if you could burst to that many (don't forget sub-agents and agent directed parallel agent workloads).
- In 1 year the cost of Sol level models is likely to cost a fraction of what it does now.
this leads to:
Workstation 1 Sol Pro 2 Sol Pro
Monthly cost $700 $200 $400
Raw capacity (hrs) 720 120 240
Usable capacity (hrs) 100-130 120 240
Concurrent sessions 1 3-5 6-10
$ per usable hour ~$6.00 $1.67 $1.67
Usable hours per $700 ~115 ~420 ~420
Subscriptions are, and will likely remain, the best deal in town. Unfortunately, larger companies aren't able to do that. When your monthly token costs are in the $5-10k range, the local inference starts to look a lot more attractive
In the case where you pay for tokens without a subscription, the analysis is still very much not in favor of buying hardware.
The assumption previously used was that you can run a Sol level model on an M6 or whatever hardware $20k gives you. That is not true, it was an assumption made to show that even giving your own hardware every reasonable advantage it still loses.
Lets compare buying tokens of the best model you might run on your own hardware (still being unrealistic in favor of your own hardware) vs that same class of model on the market. I think one of the best models you might be able to run is GLM 5.4, but lets just look at chinese models generally:
$20k workstation, best case: $15k M5 Ultra 512GB, 36-month amortization, ~$440/mo. Runs a GLM-5.3-class model at ~30 tok/s. Saturated 24/7 it produces roughly 58M output tokens/month.
Buying those tokens:
DeepSeek V4 Pro @ $0.87/M $50
Kimi K2.6 @ $4.00/M $232
GLM-5.3 @ $4.40/M $255
Kimi K3 @ $15.00/M $870 (does not fit on the box)
The economics can never work in your favor for buying your own hardware here, unless you can utilize it or sell excess capacity and you have access to nearly free electricity. The reason is someone else can buy the same hardware at scale (or realistically more efficient hardware), park it somewhere with very cheap electricity, and sell tokens. They can get very high utilization that you are not likely to get.
And keep in mind I am giving 'your own hardware' no overhead or maintenance cost, despite your condition that it's in a large corporate environment. In reality corporate IT would make it almost impossible to set up and your would need huge lead times to buy the hardware and get it installed.
I think "ultra rich enthusiast" is in the right ballpark. There are people betting on being able to create their own revenue generating products and services with their own local hardware and very little operating costs. That may or may not make sense as a business idea. But people with wealth and risk appetite trying a new kind of business model and cost structure has a strong tradition.
Put another way: If $25k is the full extent of the start up capital costs, and operating costs are very low, that is a much cheaper business to start than most! The question is whether this is actually a useful model for a revenue generating business. I think that remains to be seen.
My guess is that there will be a few hits (which we'll hear a lot about - especially when someone actually pulls off "the first single-person unicorn", which I do suspect will happen someday) and a huuuge number of misses, which we won't hear much about.
The competition is a custom multi-GPU NVIDIA RTX pro desktop, which go for much much more. $20k is cheap for 512GB addressable memory. The old mac pro could easily be configured to cost that much.
Other than the local AI crowd which is much recent it is professionals using Final Cut Pro for video editing, Logic Pro as a DAW and music production, Video transcoding, Photoshop and other tasks for high performance computing that don't need Laptops but want above 128GB of ram and prefer a Mac. Then there is the obvious group of developers that are making Apps for all of their products. Also, these are great for the workplace. AI is much more recent thing that Apple products were used for.
If Apple didn't sold these things they wouldn't make them but, also the level of marketing that Apple is talking about for AI is basically the new group they need to capture because the ones I just listed are already buying Macs and or easily to motivate with the other obvious CPU / GPU performance upgrades for code compilation, faster memory and video transcoding.
this comment has always existed behind every apple release, most especially anything vaguely pro-ish.
to answer your question : looking at the aftermarket availability of Apple's prior best and brightest : practically no one buys them.
"people here buy them" , well, 'here' is one of the most affluent groups of people in the world.
They're available as movie and television set pieces (undoubtedly disappearing into the home of someone close to the staff post-production), and for administrative/boss types that can slip the cost into a ledger somewhere that few will ever see.
It has been a hobby of mine every few years to check out the apple site and see how big I can option a machine. My record was when I was in high school years ago and was able to option some pro studio-ish apple desktop thing to like 61,000 usd out the door.
Movie set pieces, as a motivation for Apple making these high-end configs available? That makes no sense.
For one thing, you can’t tell from a movie what the specs are. A $999 Mac Studio looks exactly the same as a $20,000 one.
For another, Apple updates the industrial design on their products so rarely, a 6-year-old Mac, iMac or MacBook also looks nearly indistinguishable from a brand-new one.
Local AI is the future, and a lot of people want the first mover advantage or to toy around with it. I know a guy with a small rack of Nvidia Spark machines that he uses for that purpose; it's as much as a decent used car.
But is it? If it’s cheap enough latency doesn’t matter. Unlike say cloud gaming, where latency does matter. I’ll take my games local and my text bots cloud
Leasing now an option, only $50/month (cheaper than inference subscription?), so even cash-poor can go the amortized-investment route.
I've often felt there is tremendous value locked up in underutilized old computers. It would be interesting to see Apple in 3 years offering compute as a service using lease returns (or more likely, partnering with someone else to operate it (perhaps exclusively in secondary markets like China or India, to address political demands for local siting or jobs). Apple is in the best position to work around or even gap-fix older software/hardware limitations in a controlled environment, and now they can do so without cannibalizing new hardware sales.
I looked at multiple configurations, and mathed it out. With leasing, you pay ~75% of the capital cost (excl. tax) over 3 years, but end up with no asset.
Apple computers tend to have excellent resale value, and Mac Minis/Studios have the least depreciation of them all. I understand the benefits to both taxes and cash flow, but boy is Apple winning big on those lease offers for Studios.
I'm not sure who that line is supposed to impress. Gamers focusing on graphically demanding AAA games would laugh at this. People who don't game much probably won't know whether this is good or not.
I somehow find it better to give 2 frontier model companies 100-200/month than dropping 10 grand on a hardware that will get old in no time with bad TPS. I really want to have a fully local model but seems like one more generation wait and we will be there?
Lots of people use the Mac Mini to run the frontier models over night or while traveling. I have a rack in my basement and have thought about throwing one in. You can get a cheaper machine but Apple feels a little more, “rack and forget,” if you have less price sensitivity.
Mac Mini + MacBook Neo w/ ssh can be a better setup than MacBook Pro for many people.
The page says 170G/s memory bandwidth for the NPU and 1.2T/s for the GPU. Why the discrepancy if it's all "unified memory"? The former is nothing to write home about as far as AI compute is. The latter is really nice.
Which one is it you can run local models on? I suppose the NPU only.
For me it's be Strix Halo, 128gb machine, especially running Qwen models. Except when I bought it, it was $1,900, now it's $4,600 for the same box. (Wow that's insane)
For tinkering and learning, it's been great. Tie it into something like Hermes and you have a pretty powerful AI assistant in a box. And when you need to step up your model, you just do something like OpenRouter and it makes it pretty easy.
I bought a 128GB M4 Max Mac Studio a while back, and for a while I thought like I had done really well to buy it when I did.
The problem I'm having now is that no models are targeting RAM of that size. Everything is either much smaller, targeting laptops, or much larger, targeting hardware well out of reach of enthusiasts.
Please, AI people, start making models targeting 128GB machines again. The last interesting one was Qwen 3.5 122B.
I have had a difficult time with running 120b models on my 128gb setup, especially with any larger context size. The 6bit of Qwen 3.5 is already just over 100gb, and when you go down to 4bit it seems a bit lobotomized.
I use both the base M4 Mac mini and an M4 MacBook Air for Final Cut Pro, Photoshop, and Fusion, and while they're not as almost-always-perfectly-smooth as my Mac Studio, they're about 100x better than the experience I used to have on my old Intel MacBook Pros in the 2010s.
I edit 4K ProRes and H.265 footage, sometimes with multicam (up to 4 streams) and color adjustments, titles, etc. It's only after stacking 3-5 effects before things can stutter, really.
Or if you try doing something CPU-intense in the background _while_ running some heavy creative software. I just don't do that.
Damn. I just bought a maxed out MacBook Pro M5 Max 128GB 8TB, still waiting for it to be delivered. I could get 256GB RAM M5 Ultra 1TB for roughly the same price, and it's double the memory bandwidth. Which one would you recommend? I do plan to run local LLMs.
Contact Apple care and see what they can do. They have been known to do upgrades if you buy around the same time. But with the ram situation that courtesy might be gone now.
Sick. Particularly stoked for the 10gb network card ($100 option) when using the Mac Mini as a server. Just wish the memory + NVMe prices could come back down to pre ai-goldrush prices. As $2999 for the M5 Pro with 64GB RAM feels painfully over-priced.
As long as everything is using 10gb, faster transfer speeds when moving data between computers. For downloads, it wont help unless you have 10gb fiber, but for most folks, 2.5gb is quite fast. Hell my spinning media NAS has a hard time saturating when moving files between internal servers.
My example is I'm looking for a new media creation machine to replace my homebuilt PC from 2015. Since that old machine can't run Windows 11 and also because Apple storage is so expensive, my idea is to turn the PC into a Linux storage machine with a 10G NIC. Then I should just be able to edit off of the storage instead of worrying about caching it locally.
The prices are ridiculous though. I may just keep rolling with my Windows 10 setup.
RAM and SSD in apple gear has always been way over-priced. There was a short blessed period in March where the M5 Max macbook pro was out, but the general 30% price hike had not yet happened. In this period, given the insane inflated RAM prices, the price apple was charging for the M5 Max with 128 GB RAM was actually _reasonable_.
I have Mac M1 Max and I'm quite happy with it. But these advances make me think that maybe I should upgrade to Mac M6 (or something) when it is released.
Every time Intel/AMD gets close enough, Apple just crushed competition in terms of SoC. I don’t know who’s on that chip team but they’re world-class. Hope they get paid more than a bunch of AI idiots Meta hired to do absolutely nothing (but I know they are not even close).
M5 Pro in a Mac Mini with 64GB RAM and 10Gbit Ethernet seems like the perfect Jellyfin server and Ollama test server. All for just over $3K (I specced with only 1TB local nvme).
This may depend on the size of your library. I tried installing Jellyfin on a Synology NAS, which runs Plex just fine, and it ran so poorly it was basically unusable. It “worked”, but it was painful.
not sure which cpu that Synology has, but I run jellyfin on ugreen nasync dxp8800 plus which has intel with QSV and it breaks no sweat in both transcoding (if needed) and serving over 10GBE.
It's a Synology DS720+ with a Celeron J4125, 2GHz, 4 cores, 2GB of RAM.
I thought it would be fine, because Plex has no issues, but it was painful. Every client I tried on the AppleTV was equally painful, and I didn't even try using a client until making sure all the metadata was downloaded and setup via the web UI. I was very deliberate and did one thing at a time, spending a whole day on it (mostly waiting and browsing to different screens to force metadata to get downloaded and cached).
0% APR for 12 months (24 for iPhones only?) from Apple Financial Services for a device that can approach or even exceed what we were paying for new cars just a few years ago. Apple is definitely making bank off these financing offers, and with very little risk as unlike a car these Mac Studios don’t lose 20% of their value when you drive them off the lot.
My friend, the same way everyone else does. It goes from 0% to 28% interest when you miss a payment. Those are rates normal lenders fall asleep dreaming of.
Maybe in a year or so or maybe never. It depends on how 14a turns out and if it is comparable to TSMC 2NM. They may also choose to utilize 18a-p / 14a for other chips and not the M-series.
From what I’ve noticed, Apple products have been getting worse in quality year after year. Sometimes they even ruin their own devices with updates... I guess it’s all because of marketing.
I’m mostly talking about their iPhones, where new updates sometimes make older models worse. I know a lot of people whose iPhones started lagging after iOS updates.
Is Apple just going to announce everything silently from now on? No more getting excited for the big events to see what's new -- it just appears on the blog one day?
Also: "a staggering 1.2TB/s of unified memory bandwidth" -- yay, the GPU has reached the year 2020! (I'm a bit bitter that my M4 Max is near useless for local LLMs because of its low memory bandwidth.)
Aren't you kind of stuck with smaller models on the mini though? Even with the pro you'd be stuck with a max of four daisy chained over thunderbolt and with the 160 gig memory bandwidth you'd probably be far better off with other configurations.
> M5 Ultra features a massive amount of high-bandwidth unified memory, up to 512GB, and delivers a staggering 1.2TB/s of unified memory bandwidth that is 50 percent higher than M3 Ultra.
Apple never needed to participate in the AI race to zero. Because they were already at the finish line years ago building their own chips that can run large >100B parameter AI models locally.
As someone who works in AI now, I have found it pretty amazing that Apple basically didn't do much with AI software, and focused more on the hardware side. I think this is what the future of AI is going to look like, local models run on your mac for your workflow.
It's possible that they're working on their own LLM that's going to work very well on their chips, and possibly outperform anything out there when they do release it.
10 years ago 32GB ram laptops sounded too much. 8 was enough. These days even I would get that much ram since it’s soldered. 64GB is higher end.
In a few years we should see such high end hardware commonplace. Working with a local LLM to get work done is the ideal way to go which has mostly hardware limitation as of now that gets solved in due time.
> 10 years ago 32GB ram laptops sounded too much. 8 was enough. These days even I would get that much ram since it’s soldered. 64GB is higher end.
Ten years ago I got 64gb of ram in my laptop, same as I have now. I bought both for business and personal use. System ram capacity hasn't changed much in 10 years.
It makes me curious how old you were 10 years ago.
We were definitely outliers that long ago. I put 64 GB in a MacBook Pro back in 2019, and that was (a) overkill for everything I ever ran on that machine, and (b) stupidly expensive by 2019 standards (albeit almost affordable by 2026 standards)
>I have found it pretty amazing that Apple basically didn't do much with AI software
The iPhone 15 was almost entirely marketed based upon AI (I would say fraudulently so, advertising features they still haven't delivered), and a huge portion of the OS work was on local AI or AI integration.
And for that matter Apple has been dumping enormous sums into their own AI development. Their failure to have a lot to show for it doesn't void the fact that they tried really, really hard.
It's bizarre how often this "Apple sat on the sidelines and let the AI people fight...so smart!" narrative appears on HN. Apple hasn't gone down the path of spending hundreds of billions on nvidia GPU data centres, but they absolutely tried really hard to matter in AI.
Same here. Sadly I think the voices like ours won't be heard, though, because Apple's looking for someone who's going to buy in on the whole ecosystem, and I think we're not it. Or at least I'm not.
It's worth noting that the 5090 (or the RTX Pro 6000 big brother with 92GB VRAM) will run rings around the Mac when it comes to compute.
My old 3090 is typically significantly faster (almost 2x token/s) than my M4 Max 128GB machine, as long as the model fits in the 24GB of VRAM.
In most situations it's a better idea to just buy tokens. But there are definitely cases when that's not an option. And then a machine like the M5 Ultra can allow you to do things locally for a fairly limited budget. And in a simpler package to manage than a machine with multiple GPUs.
There is no magic, if the data you compute as atomic chunk don't fit in cache then memory bandwidth R/W limit kicks in and architecture does not matter. On contrary - having multi gpu setup of same price and same memory size with even slower memories may give you effectively much higher bandwidth but at the cost of power consumption.
I was so blown away at all the discourse surrounding "Apple fumbling on models". They should never have been in the model game to begin with. Apple crushes hardware over the last decade and that's a huge advantage today. In the end, massive models have proven to be very strong, but small models have proven to be good enough (especially with the recent Qwen 2.8 27B drop) and that's where I imagine the future will lie for consumers.
I think this is a bit of a crazy statement. Everyone expects Apple to somehow build a category leading product every year. I'd expect something innovative every couple of years
* the iPhone
* the iPad
* apple watch
* airpods
* unified memory laptops and computers
Those are all products that either created a category or changed that industry.
They did participate early on with Apple Intelligence and failed miserably. Really good move to not double down and let the others explore the space first
Is there anything comparable that runs Linux, doesn't necessarily look as good, but is perhaps (a lot) cheaper/fixable? Or is this really pretty optimal?
I mean this is not nvidia based right? It's all custom? So we can use it under Asahi perhaps?
I want to get something for my company to run local models, wondering what would be a good option.
I love linux and would be using it if the ARM support was better. It's just not there and most distros that support ARM do it a little poorly. I just haven't seen anything even remotely comparable to Apple Silicon and unfortunately Linux is struggling very hard to support it.
It's not quite that ARM support isn't good on Linux, it's that there aren't high-performance ARM chips with strong general-purpose software stacks. Like the Raspberry Pi is very well supported, but otherwise the only upmarket devices are things like Ampere workstations and hyperscaler server chips.
You can't run Linux directly on these. Asahi Linux supports up to M2 only.
Linux runs very well in a VM on macOS. There are many good options for this, some free and open source (QEMU, UTM, Lima, Colima), some proprietary (VMware Fusion, Parallels).
But Linux in a VM doesn't get access to the real GPU, so model performance is limited. Those running on the CPU perform well, and those needing the GPU don't.
However, macOS on M-series macs is excellent for local models. (Maybe not as excellent as a box full of the best nVidia GPUs, but still excellent).
So if you're getting Apple hardware, like Linux, and want to run all of it locally, a fine setup for a machine to run local models, with agentic characteristics:
- macOS running one of the many local model runners. I used to use Ollama and Whisper, and now use llama.cpp instead of Ollama. Others use LM Studio, oMLX, etc. Provide HTTP endpoints to access the models.
- Linux in a VM for overall control and orchestration, with standard VM settings, and bridged networking so it appears as its own machine on your network. Also, in here provide a robust shared file server for shared state. Use this VM as your desktop and primary access to the machine, if you like Linux.
- Linux in a VM to launch ephemeral, volatile containers, with the containers using a memory-only tmpfs overlay on top of a read-only Linux filesystem in a VM disk image, with tools in this filesystem. Alternatively, a writable Linux filesystem in a VM disk image, with disk buffering set to use macOS host buffering and discard fsync requests. These settings optimise for container disk performance for data that's only ephemeral which will be deleted soon or on system shutdown. (You can combined both VMs, but need to use two VM disks to get equivalent behaviour, and be careful about VM disk configuration of the two disks.)
- Containers spawned within that second Linux VM can be spawned very quickly and run quickly, so are ideal for LLM agents that need a quick sandbox. These sandboxes generally run faster than a macOS sandbox, despite being on the same machine with VM overhead, because Linux is faster at some things. Teach the LLMs to store files and memories they want to keep in the shared file server.
I guess, what I mean is: Why are these tiny aluminum boxes so optimal?
I just want my butt ugly repairable beast machine to do the same trick. Why is my ram not unified? I have an iGPU in my server, but it can't access the 64 GB ram (I got last year for 150 euro) directly or something? It's on the CPU right? Why did only Apple go for this architecture? So many questions...
The PC platforms have anemic memory bandwidth in comparison. Eg, Strix Halo is 256GB/s max. If money is a bigger limiter than performance it can be an option though. As can Nvidia DGX Spark machines. (Also limited to 128GB memory and comparatively low bandwidth, but higher compute than Strix Halo.)
AFAIK, apple does not release drivers open source, asahi is a reverse-engineering endeavour and does not support GPU.
For nvidia, there are both proprietary and open-source linux drivers. CUDA and inference works on linux with nvidia.
I would recommend checking out this video of Alex Ziskind to shop for a computer to run local LLMs: https://www.youtube.com/watch?v=mevUEQcumzU&t=224s.
TL;DR besides Apple he recommends, DGX Spark, Tenstorrent Wormhole N300, AMD Radeon 7900 and NVIDIA RTX 5090.
will be great fun if one M5 Ultra with 512GB memory at 1.2T bandwidth capable of doing 3x smallish local model inferencing each at Opus 4.5 level of intelligence.
The memory bandwidth and size seems to be there, but what is the tokens per sec on like a qwen model? And you can basically do 3x opus 4.5 on the $100 a month claude plan. Your payback will be near infinity years after electricity.
"M6 also introduces a Dual 16-core Neural Engine, providing up to 2x the peak compute over previous generations to make on-device AI workflows run even faster" .
Apple has the mlx framework. Most/all major software for running models locally support it. Apple also has RDMA for interconnecting multiple machines across Thunderbolt connections.
Uhhh, they are? They’re a hardware co for sure, but to say they aren’t focusing on on-device models is absurd on its face. They’ve spent over 2 years on Siri AI which is (mostly) local.
I know, this is a bit of a meaningless comment, but it's funny in a way. Feels like late 90s again:
Reading Infinite Games, they tell an anecdote about a Microsoft exec on a flight telling an Apple exec that the Zune was a way better portable music player than the iPod. The Apple exec was just “yup, you’re probably right” and then soon after the iPhone dropped
To be fair, the Zune actually was a better media player than the ipod in almost every way. That wasn't a false statement.
The issue was the abysmal marketing and "me too" attitude Microsoft had (and still has).
Technical capability pales in comparison to human desire.
I had the first iPod, and later the first Zune. The Zune had a beautiful UI, but by then Apple came out with the iPod Nano 2nd gen and that took the cake.
It was better in every way you can quantify in a tech specs listing, and not in any way that actually matters to the customer.
I had the OG white Zune. It was the same price as an iPod ~$250. The reason I chose Zune was the Zune Pass. A precursor to Spotify. Plus growing up in the Seattle area, I knew other people who had them and we would trade songs.
100% - the iPod wasn't just good enough, it was more than enough for almost everyone. The Zune was firmly in the diminishing returns category by the time it came out.
me still thinking about the iRiver H320 I was lusting after...
Actually, Microsoft came with "modern" smart phone much sooner. Actually too soon, technology wasn't there (price, computing power, size, weight, battery life).
Apple never came first ... but often just at the right moment and had marketing skills to make it a new trend.
I'm the opposite of an Apple fanboy (typing this on a Windows box), but this is some grade A nonsense.
Sure, MS launched PDAs and phone-ish devices long ago, running Windows CE and whatnot but they were awful. It's absolute bollocks that the iPhone was a splashing success because of Apple marketing. It was a splashing success because it worked spectacularly well. Random non-tech people would randomly pull out their newest purchase to show their friends. "And now it's a notepad!" "Look and now suddenly it's a calculator!" Sure, your awful HP Tablet had all that, and a call function, well before. But it sucked. It felt like using a computer while squinting, and not like a magic calculator that can turn into a notepad and then into a phone and then into an iPod.
The iPhone was a success because it worked so well. And it worked so well because the technology was there - in part because they invented it and in part because they had the taste to not bring out a shit product but wait a bit instead.
Windows CE (1996) was garbage compared to Palm OS (also 1996.)
>Apple never came first ... but often just at the right moment and had marketing skills to make it a new trend.
Newton.
Simon Sinek - Apple vs Microsoft: https://m.youtube.com/watch?v=jEOftmUJ6a4
But it had nothing to do with the iPod/iPhone release in his story.
(story starts ~1:15, but I highly recommend the entire talk)
I loved the book The Infinite Game. Really changed how I looked at work, personal research, and being an author. Recommended!
Might want to edit, as this would make sense/be amusing if the exec in the last sentence was Apple.
That, or I can’t read.
You got it, thank you!
Hell hath no fury like Apple Global Security scorned...
Zune was probably a better player. It was too late, and arrived when MS didn't much care.
There are emails unearthed in various lawsuits where you can read Bill Gates screaming at his subordinates: "why the hell can't our partners like Sony and Creative create a similar device? Give them all, give them early access to everything, work with them". In the end MS felt compelled to make their own.
Sony can't be bothered to compete with Apple in that era. They were trying to recover from their DRM dreams, and their devices were already sounding great with in-house software and hardware.
Creative's Muvo^2 already was the poor man's iPod with surprisingly good audio quality as well.
I feel like this completely misses the mark. Audio quality was never the compelling feature of the iPod and people weren't clamoring for it because it sounded good.
When the iPod came out you largely had two options for carrying your music collection on the go. You either carried a binder of CDs, or you had some niche player like Mini-Disc or an MP3 player. Both alternatives were expensive and had limitations similar to a CD in terms of number of tracks you could carry.
I had an MP3 player on either side of 2000 that was slightly smaller than a deck of playing cards that could use Smart Media flash memory cards. The largest card at the time was either 16 or 32mb and was enough to hold 1 album at near CD quality.
Creative's Muvo was a weird form factor that was larger than an iPod. It had a horrid interface both on device and for loading music. It's only grace was that it was slightly cheaper than an iPod and didn't need a Mac with FireWire. Although iircc this was pre USB 2.0 so not having FireWire would mean loading music took forever and a day.
The iPod allowed you to carry most, if not all, of your music collection in a package slightly larger than a deck of playing cards. And it had a fantastic interface for navigating music on the device.
This was at a time before most people had laptops and if you had a PC it was at home and used sparingly. The iPod was such a compelling mobile computing device that it drove adoption of the iMac. Apple would eventually release iTunes for Windows and USB support but that was many years later.
> I had an MP3 player on either side of 2000 that was slightly smaller than a deck of playing cards that could use Smart Media flash memory cards. The largest card at the time was either 16 or 32mb and was enough to hold 1 album at near CD quality.
I had something similar. The storage was the iPod's killer feature, along with iTune's $0.99 songs. Suddenly you no longer had to buy whole albums, and you didn't have to swap out what was on your MP3 player every day when you wanted a different playlist. A 5GB hard drive in your pocket was a huge innovation then.
"1000 songs in your pocket" was the entire driving force behind the thing. It was constantly bellowed in the marketing and it's what Steve Jobs demanded of the engineering team from day one. The size and the storage were paramount and they knew no one else could match it.
I had a Rio Volt which somewhat bridged the gap. 700MB mp3 CD/RWs (about 10 hours of music per disc) and a CD player when traveling and picking up new music. Not as small as an iPod, but no book of CDs was necessary, either.
Yep this was really it. The amount of storage on the ipod was amazing and just opened up the idea that "yes you put your full music collection on it".
> Zune was probably a better player.
In some ways, anyway. Never owned a Zune myself, but a university classmate did and I was shocked by how poorly it handled non-Latin languages… she had a ton of Japanese and Korean songs loaded onto it, and their metadata all displayed as "missing character" blocks. She used it a lot like one might use an iPod Shuffle despite it having a nice screen because the only way to tell what was playing was by hearing it play.
By contrast my 4th gen B&W iPod which was about 5-6 years older handled unicode just fine.
IIRC the Zune had a much better DAC on the device than any ipod. I own two of them and the sound quality was always notably great.
It let you squirt over a song to your friend
https://youtu.be/ud6rwVkbovA
And they could listen to it 3 times in 3 days
“Squirting” was ridiculed at the time and the word itself scaring away women. Jobs killed it then in an interview with a classic quote: Microsoft = Cold tech and Apple = Humanity. MS scares her away, Apple gets the girl.
> QUESTION: Microsoft has announced its new iPod competitor, Zune. It says that this device is all about building communities. Are you worried?
> Steve Jobs: In a word, no. I’ve seen the demonstrations on the Internet about how you can find another person using a Zune and give them a song they can play three times. It takes forever. By the time you’ve gone through all that, the girl’s got up and left! You’re much better off to take one of your earbuds out and put it in her ear. Then you’re connected with about two feet of headphone cable.
I just want to take a minute and be grateful for getting the chance to live through the 90s again. I always felt a bit sorry for myself being a child through the og 90s. Now I feel like in a decade or so into the future I will look back and be happy I got to live in the naïve days of windows 95 again, as an adult this time. I really appreciate it.
Cheap RAM in 500 yards. ->
[apparent tunnel to cheap RAM painted on rock face]
...and Apple zooms through it like it's a real tunnel.
Seems most computer companies are still boasting that they can beat the M1 and it's like...congrats on beating a 6 year old chipset?
I still have zero urgency in upgrading from my M1 MBP.
I don’t get that sentiment. I have an m4 16gb and it’s such a sluggish machine. and that’s not even doing development, just browsers, word, excel, PowerPoint. (And so many eternal bugs… I feel like I’m on windows)
I have an M2 Pro MBP 16GB, compared to the latest MBP issued by my employer a couple months ago, I see zero difference in real world performance. Can’t speak for Office on Mac, I have not used it in over a decade, but for typical full stack web dev I literally don’t need anything more.
Apple is like the Billy Mitchell of computing. Just waited for them to beat it before announcing the gains they were sitting on.
Does this imply they faked all their benchmarks?
Echos of the ai race as well haha
It's been really weird reading laptop reviews over the last few months. I've seen a bunch of reviews where they have their usual bar graphs comparing a bunch of laptops, and the laptop that is at the bottom is an Apple. It's the really inexpensive Neo, of course, but even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops.
I assume the M6 will take the crown back and then a few months later Intel/AMD will release a new chip and take that crown back again. That's the state of the world we used to expect, but it's a state that has been missing ever since the release of the M1 in 2020 until Intel finally caught up again this year.
> even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops.
When plugged in... This caveat is so enormous it should almost be legislated. If your computer use is at all portable, a computer that scales down to 20 - 40% of GPU power when unplugged is an enormously significant factor. So far as I'm aware (could be wrong about arm devices?) there's no non-apple laptop that operates at 100% speed on the road.
Newer Intel Panther Lake chips perform the same or very similarly on battery as they do when plugged in - as do Qualcomm chips. However, I don't know where they're seeing "consistently beat on performance and battery life by Intel Windows laptops". Multi-core performance can definitely beat M5 in some configurations but single core performance is still fairly far behind and battery life is comparable again depending on the exact specifications and design of the laptop. I've seen analysis showing M5 is still the perf/watt king though regardless of configurations.
Sorry, poor wording. It's the Neo that's on the bottom in many reviews, but the M5 is regularly beaten by Windows laptops. My usage of "consistently" was in the sense of regularly beaten, not in the sense of always beaten.
So…inconsistently?
So consistently beats the M5 on something, just not on everything.
Have you tried Panther Lake or Wildcat Lake laptops?
Jesus, they're still calling their chips Lake? As in Skylake? As in 2012?
Jesus, why does it matter? Don't you still have your name since birth? How lame!
Because the naming convention was used for chips based on Skylake microarchitecture (but with other differences such as process node). Not changing the name implies they've either changed the naming convention (why?) or they haven't improved the microarchitecture since 2012.
They keep the legacy processing of content around. But the originating teams and founders are long gone.
Apple also is faster when plugged in
Citation needed. I've owned Apple laptops for many years - and I'm a video editor (amongst other things). If this is the case it's new behaviour.
This is a difference between selecting Highest or Automatic for performance under the battery menu item.
> but even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops
The real question is whether all these Intel Windows laptops spin their fans at full speed when doing absolutely nothing. That’s something I can never go back to. I do some light gaming on my M2 Pro MBP and it gets hot when trying to push 120 fps. But my Lenovo Legion (that’s now collecting the dust) is so loud I could hear it through headphones.
> [...] but even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops.
Hold up, in what metric/benchmark? Feel personally like we life in an age where, no matter the SOC vendor, something high performant and efficient is offered, so seeing a claim that any vendor, be it Intel, AMD, Qualcomm or Apple, is consistently outperforming another, I'd like to get more context on that.
It seems logical in this case.
Intel were x86, Apple Silicon is ARM-based. ARM-based chips are more power-efficient.
Also, it's built on a much smaller process. 3nm, if I'm not mistaken, older Intel was something bigger than 10nm. Heck, if you take a really old Intel Mac, you have something like 65nm process, which is much less efficient than 3nm.
Here's a random benchmark I found on the internet (literally the first thing on Google, you can find more if you want) https://www.cpu-monkey.com/en/compare_cpu-intel_core_i7_1065...
> It's the really inexpensive Neo, of course, but even the M5 laptops are consistently beat on performance and battery life by Intel Windows laptops.
Of course you can beat the entry level MacBook Neo by comparing it to larger, more powerful, more expensive laptops.
The M5 is an entire family with a range of performance. Intel/AMD have done a lot to improve performance but they’re not beating the high end M5 chips on performance or battery life yet.
You can also beat the Neo on every measure by comparing it to $700 Windows laptops. https://www.tomshardware.com/laptops/dell-xps-13-2026-review
It's also not hard to find a Windows laptop that beats the M5 on battery life. Single-thread performance is the only remaining measure where the M5 is king.
x86 is way more complex when compared to ARM processors, as a result their TDP is way higher when you request performance from them.
Intel had to reduce the frequency of their processors when running AVX2 instructions and the AVX2 frequency of the processors were non-disclosable to anyone.
Also, benchmarking Intel processors and publishing these numbers were forbidden in some cases. I don't know whether this ban is still in effect.
x86 processors can't keep up with the ARM processors TDP and thermal profile wise. So they slow down a ton when running on battery. See Jeff Geerling's last video on Apple Neo vs. some Intel laptop. It's as "efficient", but slow as a newborn tortoise learning to walk when unplugged and trying to get the most endurance out of the battery.
My M1 Mac gets almost 2 days of low-intensity use after ~6 years of use, and it got warm once or twice because something ran away in the background for tens of minutes.
Being mistaken on the CISC vs RISC debate (besides, modern x86 is closer to RISC via micro-ops then old school CISC) is understandable. There is a lot of misinformation out there and myths, plus, it just feels right to consider CISC overly burdened, etc. TLDR: Intel x86_64 is closer to RISC then you likely think and Apple Silicon arm is closer to CISC then you likely think. These lines are blurry and have been for decades, very much for good reason.
But talking this authoritatively on something without doing the reading, that's grating: https://chipsandcheese.com/p/arm-or-x86-isa-doesnt-matter
> But talking this authoritatively on something without doing the reading, that's grating...
Thanks for your prejudice on me without knowing anything about me. In short, I'm a HPC sysadmin and programmer who works in a HPC center, separated from the actual hardware by a couple of floors.
We can discuss how transistors' heat generation doesn't discern about ISAs or being in a DAC or a cutting edge microprocessor, and we can even discuss how implementation of some functional blocks generate heat regardless of the ISA being involved. If you want we can discuss how saturating memory controllers affect pipeline saturation in processors even...
But talking this authoritatively on something with that amount of prejudice, that's grating.
But still, your point was AFAIK already dismissed long ago.
After that comment, why does it matter where you work?
Because being too confident about someone you don't know is a bad habit. Where I work doesn't matter though, but what I do is.
Pointing me to C&C is a nice touch though. Not only I read the site and very article you sent me before, I used to consume Anandtech before that.
As a mere mortal, I can make mistakes and gladly accept them, but I can't accept rude replies. Pardon my French, but being called a low-key liar or smoke blower gets me a little upset.
> Pointing me to C&C is a nice touch though. Not only I read the site and very article you sent me before, I used to consume Anandtech before that.
All I'll say is, that's worse then. Presuming you had not read up before promoting a long disproven myth, that was an assumption by me, I'll admit that and maybe I should not have done that, my mistake. But it was a gracious mistake, it was done in your favour, it was giving you credit.
In the ‘90s it was the opposite.
- Apple: We have matched Intel in CPU performance thanks to this new PowerPC! (Shows ad that uses carefully handpicked benchmarks to suggest that the PowerPC is actually faster when it really isn’t on average)
- Intel: Oh, we just found a 30% clock rate increase in the pocket of our other fab pants.
- AMD: Hold my beer, I have the DEC Alpha guys making an x86 CPU… How about 64-bit while at it.
I said late 90s though. The insanity which poured through early 2000s.
5.25GHz Pentium 4s, intentionally lower binned Athlons, the era your CPU got obsoleted the moment you booted it for the first time.
I don't remember early 90s much. I was too young back then. I don't remember much stuff from that era. But late 90s, early 2000s.
Oh, boy.
P.S.: AMD64 was a great sucker punch though. One of the professors in our university rejected to believe and got mad when he learnt that Intel licensed AMD64 from AMD, heh.
Except in the 90s, I could wield my OS of choice on the hardware I bought.
That's sadly true though. Also booting something was really simple.
Now we boot an embedded microcontroller (or CPU) which boots the main CPU which boots another OS semi-persistently to boot the main OS (if it's allowed).
Sometimes there are other processors needs to be up to allow processor to continue booting as well (these are mostly servers, but eh).
To be fair, this kind of dominance is not unprecedented. Intel was even further ahead in the early 2000's. Every new competitor process was further behind the leading edge and not closer. TSMC started launching half nodes like 28nm just to have something in the market that would sell.
But then you started to see the cracks. New competitors would launch new products with very slightly better metrics than Intel's older stuff, just to be, heh, meep-meeped at the next press conference. But the overlap was real, if small. And it grew over time until everyone looked up around the 5nm node and realized Intel had lost.
That's where we are right now with Apple. "Funny in a way", sure. But history says this is more likely to be the beginning of the end. Everything goes in cycles.
Haha, late 90s wasn't really like that though
Apple levelled up in the middle of a fight.
I do not find it funny tbh.
I'm very very surprised that Xiaomi matches Apples speed even with the newest release, its not diminishing Xiaomis success.
I think it's great but it's important to remember that we haven't seen Xiaomi's chips in actual devices under real world tests. We don't know if the speeds are sustainable, under what wattage, etc. Competition is still great and I look forward to learning more, of course.
It's interesting to see how defensive some of these pro-China posters get on HN and elsewhere. I'm genuinely curious why there seems to be an inferiority complex here with respect to America/American companies.
China follows foot-in-the-door tactic per the Art of War. (done in HK, Korea, Singapore, and Malaysia)
This in-turn later on, AIs will train to make pro-China comments as AIs train on these.
They got the sheer man-power, and with AIs it's even easier.
I'm not a pro-china poster, i'm from germany and didn't find this 'funny'.
Apple is the second richest company on the world (which doesn't need help/protection?!) and they have experts in chip design.
Xiamoi is some random chinese company not known for high end chips and was able to catch up impressivly in a short period of time.
This fact doesn't get funny or wahtever just because apple brought out a new chip today.
I'm not a fanboy for any of it and do not care.
> Xiamoi is some random chinese company not known for high end chips and was able to catch up impressivly in a short period of time
Xiamoi is a huge company and is a commonly known brand. They’ve been making chips for a long time. They didn’t start a few months ago and catch up with Apple on their first try.
> random chinese company
Oh i have soooo many news for you!
Are they well known in China? I'm seeing that Chinese tech has decoupled from the West. They have all this cool stuff that we do not hear about because they aren't selling it to us because they don't need to. It's usually been stuff with better price to performance or ultra low prices though, rather than ultimate performance stuff. I wouldn't be surprised if they made a top end chip that everyone in China knew about, and we didn't.
> Xiamoi is some random chinese company
That might be the funniest thing I read all day. Xiaomi is a massive company that makes all sorts of things, including a lot of pretty high-end mobiles, and has an annual revenue in the order of 75 billion US.
It’s no Apple, but it’s not exactly “some random Chinese company” either.
> catch up impressivly in a short period of time.
They're using ARM designed cores. So it's more like Apple just isn't as far ahead of ARM as some people claim.
I find pro apple comments funnier; but you know, everyones got their own rose colored glasses.
I find it funny not because I support Apple. I find it funny because it feels like the leapfrogging happened in early superscalar CPU evolution. The era when the Moore's Law was working.
I enjoy it because of progress, not because of Apple.
Even w/the pricing spike, inflation adjusted we are back to roughly the prices of a new Mac SE/30 for something that can beat a turing test w/o sweating.
I yield the floor to no one when it comes to pessimism, but that's incredible.
The DRAM market is cyclical. I don’t think anyone truly knows when, but it will happen.
Fab capacity is being bought online; there’s just lead time.
Noticeably greater intelligence is being achieved at the same number of parameters (see: Qwen3.8).
I think the future will be bright, it might be a matter of time. And for tinkers, a used Epyc + DDR4 server can be great fun and epic value.
100%. We're probably on the cusp of over-capacity, a glut, cheap RAM, bankruptcies, and shortages.
The memory companies report that they're sold out through 2027... so it might be a while
That's only if the AI companies continue their build outs.
How many people actually use fable over opus? How far are we up the diminishing returns curve, and will their customers even care?
I don’t think the effects on the hardware market of open weight models triumphing has been reflected upon enough. It’s not clear that it will be less impactful if every enterprise decides to build for predominantly on-premise inference. In fact, before the question is ultimately decided, we’re probably heading towards a few years where hobbyists, enterprises and ‘hyper scalers’ are all competing for certain parts in common. Ram booked through 2027 sounds about right.
Looks like I'll be sporting my 5800X3D + DDR4 + 9070 XT gaming box for a little while. Too bad the CPU's ST is slower than my MB Air m4.
I built a new PC about two years ago, and I probably got it at the last possible opportunity for a while. CPU and motherboard have come down by maybe £100 in between the two of them, but a 7900 xtx (or any other 24GB GPU) for under £1,000 now seems like a bargain, and £180 for 64GB of DDR5 makes me feel like an old man talking about the halcyon days.
I built my current PC the day that the AM5 platform released, for about $6k not including the 3090 I moved over from the previous rig.
If I sold just the two sticks of RAM in it right now, it’d pay for nearly half of the total cost.
Welcome to the club. If you're _really_ competitive in cs2, I'd swap out to a 9800x3d setup, but it's still a maybe. Very little reason to upgrade right now other than to run LLMs.
It will never make sense to me to run Llms locally unless I had 50k. I don't even think that would compete with price perf of a remote llm and getting business done. And that's completely ignoring that sol/fable level is not local
I'm expecting it to somewhat collapse. I don't know if it'll go back to pre bubble prices (here's to hoping), but I do expect a pretty sharp decline around 2030... probably not before then.
Basically everyone that makes memory is building new fabs, meanwhile I'm not sure how much longer AI datacenter demand for ram will last. I think the decrease in AI ram demand and the new fabs will likely coincide leading to a collapse in pricing.
That is, of course, assuming the memory manufacturers don't pull their favorite trick and collude.
If there's a decrease in AI ram demand, it will not be because the models get better. Models getting better will increase RAM demand, because it grows the part of the economy that models are useful for. Classic Jevon's Paradox.
If it will happen in 100 years it will practically never happen (for us). Even 25 years would be a lot, it's half of a career.
Could you provide more details about the Epyc + DDR4 server?
Cycles tend to be 5-10 years long not 25. Without knowing anything else I would expect a fab you seriously start planning today will be at full capacity in about 5 years. Nobody serious likes delays - in particular the banks don't like loaning money that won't at least start paying off. They know it takes some time to design a building - but factories typically are standard buildings so once you know about the size you can get it done fast - I expect 1 year to have the building done is the worst case (and it can be done in 3 months possibly if your project management is good - after interest this is cheaper than the 1 year). It takes time to build and install the specialized machines that go inside - this is the largest problem, but you typically order them first and then plan the building around the needed space and when they will arrive. Then you need 6 months to setup the inside of the building. From there it is just ramp up time.
The above is a standard project management problem. We do this for lots of industry all the time. There is every reason to think you can get a new factory running in 5 years.
Note that I said 1 factory above. Some of the special machines we don't have the ability to make them fast enough to do 2 (I don't know the real number!) new factories in 5 years. Existing factories are using most of the special machine capacity to replace machines that wore out on the way - this can be corrected as well, but it adds another year and the expenses are much larger. Realistically though 1 new factory is likely enough.
If you don't understand that the AI driven memory boom has totally broken the cycle you are going to lose a lot of money. There is infinite demand for Intelligence and that translates directly to chips.
> The DRAM market is cyclical.
...
> I don’t think anyone truly knows when, but it will happen.
Do you know what cyclical means ... ?
An additional £1,000 for a 2TB drive is crazy though, rapidly takes the new Mac mini from a good price to a nuts price
Considering hard drives were $10k / GB in the Mac SE/30 days, that feels like a bargain too.
But software also fit in 640kB instead of 640GB.
It still can, but until very recently the motivation for keeping software small wasn't there. I'm still hoping the tide is coming in.
On the higher end upgrade we're more like back to roughly SGI prices.
Yes. I commented elsewhere that it's only twice as expensive as my first mac which has 128kb.
You should compare with competition, not what was decades ago.
E.g how is the perf/$ vs Wildcat lake
when was the turing test beaten?
1966
https://en.wikipedia.org/wiki/ELIZA_effect
It turns out the limiting factor isn't how sophisticated algorithms are, it's how gullible humans are.
What does that have to do with the Turing Test? The TT has clear rules: There are judges that have a dialogue with anonymized AI/humans. The humans cannot cheat and impersonate a machine, they have to act normally. The AI obviously should try to sound human.
No AI would pass this test with experienced judges.
You’ve moved the goalposts.
You can always say “oh well these judges don’t have the experience to catch this type of AI.
The fact that you have to insert this qualifier, to ensure you always have a way to discredit the test, pretty much shows to me that we’re beyond it.
You think current frontier models couldn't pass for a human on an online chat? You and I have very different perceptions of reality.
I think that if you have a long enough chat, yeah, I think you can figure out who's meat. The original rules for the TT specified a short interaction, but I can probably accelerate it by pasting in large code snippets to force early compactions.
Sounds like something we could settle right here and now.
Ha ha! Fool! You've been talking to an LLM all this time! Your wife is actually Haiku 4.5.
Huh, should've known it was odd for my wife to always say I'm right (as the boomers would say)
The test doesn't say that the judge has to be experienced. But I also don't care if some random gullible person can't tell the difference. Nothing passes the Turing Test for me yet.
Edit: Also doesn't say anything about who the human test subject is
How can you be certain you haven’t failed a Turing test?
Of course, and I'm sure OP wouldn't disagree with you, it was clearly a joke for emphasis. Some people round here need to clean and calibrate their humour detectors more often.
I wasn't responding to OP. I get the emphasis, the M6 Mac is very powerful.
I think of this and the book the author wrote Computer Power and Human Reason every time I try to talk to product about the short comings of LLMs
It was always a bad test, despite the greatness of Turing. The human organism is built to 'project' humanity onto anything available; apart from this none of the peculiar phenomena of the so-called 'modern human' is even intelligible, even the possibility of science. I bring all that is in me onto you as soon as you seem to be saying something, and reciprocally. We do this at the drop of a hat, and all specifically human life depends on it. But this power shows its 'gullibility' with 'gods' as also with Eliza. I am not snide about it because it is overreach by something the significance of which is overwhelming , but one is indeed amazed by the failure to reflect on the part of the ones eg giving LLMs rights - to take extreme case of a very widespread cultus - as if /they/ were the rational party, not ancients placating the storm god.
Yeah...but in context, "gullible" seem a bit pejorative. Humans are also hopelessly incapable of sensing radioactivity, methanol in their alcoholic drinks, carbon monoxide, and a great many other things that our ancestors just didn't encounter much.
Though we're pretty good at sizing up a person's emotional balance/maturity and competence at familiar tasks. So maybe have an old blacksmith watch the AI/robot interact with horse owners for a while, then shoe their horses, and see how well it does.
The EU just had to pass a law to force companies to disclose if a customer service agent is AI or Human. It is beaten.
https://commission.europa.eu/news-and-media/news/safer-and-m...
Also the endless online debates of ‘is this post made by ai? What about those images, that video or that music?’.
Nowadays, I prefer AI CS agents to humans. I just had a chat with an AI yesterday, it understood me perfectly even when I made mistakes, I was impressed.
In contrast, humans tend to paste me the same barely-relevant macro over and over, no matter how much time I spend explaining my issue.
Yeah, at least LLMs read everything you write (for now). Human first level support agents are incredibly frustrating if you have to explain anything with more than one logical step.
Customer service is very different. Crappiest audio quality possible & scripted answers all the way down. Almost like humans are forced to behave like machines.
According to Psychology today, April this year by GPT 4.5
https://www.psychologytoday.com/ca/blog/the-digital-self/202...
April last year
I think it was determined that the Turing test is too easy because humans are too easily fooled.
Yes, which is exactly and entirely the point of the whole paper. Somehow missed still -- despite how important AI has become, shockingly few people actually read the short, layperson-accessible paper that started the whole field.
I kinda have to link it now, so uhh here's a random PDF: https://www.hec.edu/sites/default/files/documents/Computing%...
The Turing test is more complex than what gets suggested.
And the "popularized" version is faulty also since it uses an ideal, abstract human judge (like the "spheroidal economic agent").
But if you want to add declinations to the said popularized image of the Turing test, you may add Maxim Lott's IQ tests at trackingai.org . Between the end of 2024 and the beginning of 2025 LLMs reached an equivalent IQ of 100, for example.
Or can we reframe it: when did humans start losing the (so-called) "Turing test".
I think there are elements showing lowering of performance and expectation.
~1960
ELIZA beat the Turing test and then everyone forgot about it. Humans are just really terrible at recognising robots.
2001 according to Wikipedia
https://en.wikipedia.org/wiki/Turing_test
It depends who takes the test. I am not yet, to my knowledge, fooled by AI.
I've tried [1] and I almost 100% detect which is the AI. I really want to convince myself I have failed, does anyone know of a better site/resource for this?
I know it might be moving goalposts but I would consider AI to have passed in a well and truly undisputed manner when [2] is resolved.
But in a more practical sense, if AI can impersonate humans so well today then why are state of the art frontier models so obviously AI when they create PRs, commit messages, documentation, etc. Are the companies deliberately making them unnatural?
[1] https://turingtest.live/
[2] https://www.metaculus.com/questions/11861/date-when-ai-passe...
we might need to bring back the Voight-Kampff test. anthropic at the very least is introducing a water making system to Claude which might make them more identifiable to humans as well as much easier to detect for machines.
“advanced LLMs like GPT-4”
> "advanced LLMs like GPT-4"
Not sure where you're quoting from but if it's the metaculus question comments, many of them are from 2023. The consensus is it will resolve in 2029. I believe it will not resolve before 2035.
From the home page of turingtest.live.
Yeah. That's why I asked if there was something better
2050 I'm guessing
are you living under a rock?
Sigh..
Is there anything better now though?
All I see from AI, is an amplification of the enshittification of the internet.
And people being even more alone.
Sure, when you look a little wider, since 2000 we have seen the following major improvements:
- Extreme poverty has dropped from 30% to under 10% globally. - Child mortality rates have dropped in half - Internet access has exploded from 10% to 70% - Solar energy costs have dropped 90% - Cancer death rates have declined by 30%
All of these massive improvements in less than 30 years.
While there certainly are issues to solve, and if you simply follow journalism you may think the world is worse off, but for many, their lives have been significantly improved.
On that note, I'd like to share Fix the News: https://substack.fixthenews.com/
It's a Substack that reports good things happening around the world, divided into sections like "Conservation and Restoration," "Climate and Energy," "Medicine," etc. And they also give part of their profits directly to projects in those categories.
(I'm not affiliated with them, I'm just a subscriber.)
These improvements are showing that we're producing plenty so that the cost is driven down and distributed to wider and wider portions of humanity.
But my personal observations of AI is that it's producing more and more of the same stuff and not moving the front forward much. The human innovation and invention seems to be lost.
Thanks for sharing this sentiment and including data. So few people seem aware of the wonderful progress humanity keeps making. Makes me worry that the progress will stall or even reverse because people don't even know it's happening.
Is that AI / compute driven?
In part, sure. From drug discovery to crop yields to education, compute and AI have material benefits. It's kind of shocking that anyone could doubt this.
Evidence:
* https://arxiv.org/abs/2402.09809
* https://phys.org/news/2016-12-mobile-money-access-percent-ke...
* https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3893351
This being HN, I hasten to add they also have massive downsides, we're all doomed, nobody programs the right way anymore, those poor people just think compute & AI are improving their lives, etc, etc.
AI in the sense of LLMs is too new to really make an impact yet. But it is compute driven. Modern science and engineering would be impossible as we do it now without high-end compute.
Just like the computer revolution, it will make a small number of people richer and put everyone else at their mercy. In real terms the average person is far poorer than in the 20th century. Used to be able to buy a home and support a family on a single income. Now it can take two just to survive.
In real terms, the median human globally today is vastly richer than at any point during the twentieth century.
And the median American is also richer in real terms, both in terms of wealth and in terms of income.
Of course, that all assumes that you use a reasonable measure of inflation that’s stable, well-designed, and applied methodically and consistently over many decades.
Alternatively, you can cherry pick data points and go based on vibes, which lets claim whatever you want!
I wonder if doubling the workforce had anything to do with that..
You're right, probably nothing to do with moving all manufacturing out of the country or Reagan's union busting.
Well, there is more renewable energy than ever before powering the world. It's all being built and added to the grid faster and faster each year.
That's one big plus.
Then you’re either spending time in the wrong parts of the internet or doing things that don't fulfill you.
These times are exciting and rough seas make good sailors. Find your path forward.
What is exciting about getting pablum spoonfed by a data center owned by trillionaires?
I'd rather go to the library and read a book.
You can do both. I still find time to read a couple books a month.
You have agency and can choose your own adventures. If you don't like something, don't do it.
Why in the world are you choosing to live that way?
I'm writing the best music of my life, realizing games and art projects I never had time for, and writing higher quality software in addition to dramatically more of it. Who has time for pablum?
How exactly are you using these tools that you have that experience?
You are actually not doing much of that.
You have become a spectator.
Who is stopping you?
Your legs broke? Go do that
What does it have to do with legs? This site has turned a mental hospital with AI patients.
It's an expression. "You're legs broke?" here means "What's stopping you from going to the library and reading a book?"
It's a grovel-for-investor-dollars site that we've long pretended is for serious technical discussion. Comments should always be filtered through that lens.
False dichotomies make for terrible conversation.
"Find your path forward!" he shouted with glee, as he ran toward the cliff.
I run a lot of local models (I am always experimenting) on my 32G M2-Pro MacMini - I would love to upgrade.
The financial aspects don’t work however: I can learn and experiment with what I have for local models, and I pay as I go on FireWorks.ai for open model inferencing and no matter how much I use this service my monthly bill is between $10 and $40 and much faster than any reasonable home rig.
Hybrid ‘small local’ and buying inference is the way I choose.
Apple Studio with maxed out M5 Ultra, 256GB RAM and 16TB storage is 18,299$. The 512GB RAM version apparently is coming in October, considering that the difference between 96GB and 256GB is priced at 4000$, the 512GB upgrade must be eye watering.
So, on the mini the RAM upgrade runs at 25$ per GB on all tiers, the same as the Studio therefore the upgrade to 512 will probably cost 6400$.
The fully maxed out Apple Studio then will be 24699$. It's 17199$ if you don't upgrade the storage(1TB).
Nevertheless I itch to have one :)
So is downpayment on a house. I would buy the house and just pay for tokens as needed. The house will get more valuable and that wealth would buy a lot of tokens in the future - which will probably get cheaper.
EDIT: or buy AAPL. If I had bought Apple stock instead of buying a Mac LC II in 1992, then I would have about $2 million in Apple stock.
Or more simply – $25K (+ tax) put in a savings account will earn about enough interest to pay for a $100/month AI subscription indefinitely. And at the end of it you still have the $25K.
Not 'more simply', there are basically zero savings accounts that are going to net you a 5%+ interest rate to give you that $100 a month. And that $25k becomes less valuable over time. $25k is now only worth $19k because inflation.
Money market/treasuries (even ETF like SGOV) gets pretty close to 5% when typical savings rate is a bit under
If treasuries “fail” we have a different class of problem.
It’s basically impossible to compete on economic terms with deeply subsidized hardware that is widely available to rent or as a service with zero commitment.
For general inference there’s no ROI that makes this work vs subscriptions.
25k for computer now, plus 9-10% sales tax, plus operating cost, plus time and cost for R&D tinkering with models, harnesses, and infra (assuming highly capable engineering talent that can get paid for your human inference) vs a HEAVILY subsidized subscription at 200 per month with free R&D has a pretty long ROI (15 years?)
At API costs, it’s like 6 months if you’re heavy on inference. For training, specialized models will have their own ROI that makes this worthwhile. Then debate renting capacity and the platform to choose
Isn't there something to be said for owning your own hardware though?
This is the real answer.
Unless you need privacy for your inference this instant, paying for credits can get 80 to 90 percent of people everything they need.
Of course if you do need that privacy, then forking the $25K over to Apple is a no brainer.
I don't need privacy, so it would be financially imprudent for me to spend 20 grand on such a machine. But I have a financial management client who does need such privacy, and if I get more fully engaged with them then I would be able to justify getting a loaded Mac.
Why is this a no brainer?
There are both cheaper and faster options out there.
where are downpayments so cheap? I'll move there ^^'
There are 70 houses for sale in Pittsburgh for $25K or less
Coshocton, Ohio, plenty of houses for <100k
On second thought...
Had I done that as well, then maybe I wouldn't have gotten intrigued by HyperCard, then Director/Authorware, then Flash, then HTML, then...
With the way RAM prices are going up, you could expect to make 20% profit on any purchase.
I tripled my money on my RAM purchase of three years ago. So, yes, for short-term appreciation that's hard to beat. But I don't think it's something that will continue.
M6 Mac mini maxes out at 32GB—if you want 64GB you have to go with the M5 Pro (just priced it out on Apple's store page).
I wish you could drop $20k and get a house! Down payment here store like $100k+ (AUD). So "only" 5~ Max Studios.
Ha yeah that was my thought. How is this a down payment? This would only be 20% of a $100k house...
Well a substantial proportion of home buyers in Canada put down only the minimum 5% down payment.
Apple hasn't been selling just ram for a long time, they sell vram. Try getting 512 gb of HBM on current Nvidia cards - it's gonna cost way more than $ 24k. And here you get the same amount of memory for weights right in a quiet unit under your desk
Put together a similar build with a couple of rtx 6000 Ada cards and Apple's price tag suddenly looks pretty damn reasonable
HBM itself is very expensive but it’s not really fair to compare to LPDDR or GDDR
They’re very different things.
The more logical argument to me is that Apple uses its upgrade price points as more than just direct BOM and rather as a proxy for things that are amortized across all their sales like support/warranty/etc so higher SKUs subsidize the costs of the lower ones.
i remember when SGI boxes were $50k and then literally worthless just a couple bears later. i remember my university had a pile of them for free outside the deans office.
I figure most electronics are worthless after a couple of bears.
But every second bear doubles the transistor count.
I think the reason they offer these options in the first place is the discontinuation of the MacPro and their remaining need to offer high end solutions.
I don't think 16TB storage is a right choice. Going 2TB and it's 11,299. Probably, if you buy the storage and install it yourself you can go higher and quite cheaper.
Ten years ago a bought an expensive MBP because I do a lot of stats in R, Python etc that benefited from it. But the next Mac I’ll buy will be a much lower-end model, because it’s just easier these days to do that work in notebooks in the cloud.
More validated by the day that my $450 M4 Mac Mini (16GB) was the best deal in computing for a long, long time.
I live right next to a micro center and remember when they started offering that deal... still so pissed at myself for not buying one. I ended up just buying a raspberry Pi for what I was doing, but seeing as where the prices are now, I messed that up a bit. Also my worst sin was not buying 64GB of DDR5 when I was doing my computer upgrades back in August last year.
I returned an M4 Mac mini, 64GB, unopened... because I thought it was excessive for my needs then. I swear it'll be one of the things flashing before my eyes when this all ends.
Should have grabbed two.
Should have grabbed 100.
Rumors say that Apple will only release M6 base variant and skips M6 Pro, M6 Max and M6 Ultra variants to concentrate all efforts to create a good AI capable M7:
"According to reports from Bloomberg, Apple will be skipping its M6 Pro, M6 Max, and M6 Ultra chips to accelerate development of the M7 chip. That means the only chip to be released from the M6 family will be the base M6.
The reason for this break with tradition: AI. Apple had been planning major neural-processing upgrades for the M7 family and ultimately decided those improvements were important enough to justify accelerating the next generation rather than completing the M6 lineup." https://9to5mac.com/2026/08/08/apple-m7-chip-heres-why-it-ma...
I'd skip M5 and M6 chips for LLM work and wait for a year for M7.
The M6 doubled the neural engine from 16 to 32 cores. I would expect that the M7 doubles that again to 64 from 32? That would make sense.
I believe that the CPUs are actually limited by ram bandwidth more than the neural engine right when it comes to LLM processing?
Maybe the M7 introduces something new to get around the current ram bandwidth problems on the non-Ultra chips.
LPDDR6 is coming.
Do current models run on the NPU or GPU? Wondering if Apple will have something like a TPU.
I'll upgrade M3 Air only when Mx Pro/Ultra can run Opus level perf locally. Otherwise what's the point.
> I'd skip M5 and M6 chips for LLM work and wait for a year for M7.
Please say more? Is it because it is a one-time cost, unlike a recurring subscription of Claude/Codex?
The true-local AI chip, codename "buddy", will be the M8
Apple still has the best hardware so I moved to it for the last few years, but the closed software ecosystem is terrible for taking advantage of it.
I wasn't able to debug network errors (restartin my Mac worked), Metal was missing low level disassembly / debugging tools (there is some hard to use UI), but the worst thing was the inflexible windowing system.
Even getting all the window handles on all screens/desktops with their titles and programs is impossible.
I just decided that I move to Omarchy 4 (basically Hyperland + QuickShell) + NVIDIA GPU, and I already was able to customize it more than my Mac in years.
I will miss Apple's hardware for sure, but not MacOS and the missing hardware documentation
I had a chance to try Omarchy past few days and it's just very different vs macOS. I honestly never had a problem with the windowing system ever since I built my own customization scripts (i.e. Hammerspoon). I can see the appeal for someone who wants ultimate customization though but macOS still wins overwhelmingly when it comes to polish, ecosystem, user experience, and apps (nothing comes close).
It’s not just Omarchy, there’s really not much out there in the desktop Linux sphere for those who are mostly happy with how macOS works out of the box. Everything is either in a similar vein to the Omarchy setup (hyper-minimal tiling WM), Windows-like (KDE, Cinnamon, most other DEs), or a chimera with a grab bag of design bits from every desktop and mobile platform (GNOME, Pantheon, COSMIC).
It’s a bit depressing because it means that if I ever feel forced to switch my daily driver, it won’t come without a dump truck load of friction, frustration, and lost productivity, which I’ve validated by using the various Linux desktops on secondary machines.
There were 1000 plugins created for Omarchy 4 in 2 days. That's why I don't feel it being hyper minimal anymore.
It's still not well integrated of course as those plugins are from different people, but I at least don't feel powerless as I know I can make any change easily.
It's interesting because I just haven't felt the polish.
For example when using PyTorch I wanted to try to speed up my NN kernel by 2x by just using half precision and haven't noticed any speedup at all. Also I was missing the easy to use GNU tools that had to be mixed with Apple's tools.
I loved using Arc browser as well, and I'm missing it, but I guess I will do without it somehow (Chrome's vertical tabs are just not the same).
My main program missing from going back to Linux was ChatGPT Desktop, but now it's there.
I just checked out Hammerspoon, I'm happy for you that you wrote it, and looks great, but it has the same problem that I had: for security reasons Apple stopped allowing the window APIs to get all important information on other workspaces. You can only do it with Accessibility API. I was trying to fight with it but have up.
If the browser being Chromium-based isn’t a hard requirement, it may be worth checking out the Firefox-based Zen Browser[0]. Its UI is very similar to that of Arc, to the point that I’d call it Arc’s spiritual successor.
[0]: https://zen-browser.app/
How are you liking Omarchy? I saw a video on it recently, and it looks 'pretty' but still looks like it's a lot of memorization of shortcuts and feels like the 40% keyboard of OS's. Like some people it's absolutely amazing, but lets be honest, it's going to be really difficult to be as productive as a full fat keyboard.
On what hardware you running Omarchy?
Just Beelink, but it doesn't matter at all.
I ordered an ASUS Zephyrus G16 with 5090 NVIDIA card + 1.9kg (quite an overkill, and I know that I will have to limit power output), but hasn't arrived yet.
But what's fun is that I love QML+QuickShell with its hot reloading, Hyprland with its Lua support.
With AI nowdays it's just so easy to do deep UI changes that wasn't possible a year ago.
The 512GB Ultra is amazing, sure. But who is it for exactly? VC funded big spender founders? In that case why would they need local AI? The ultra rich enthusiast? But there can't be too many of those. So who actually buys these?
I think there's some mental inertia around what a computer is and what it's worth. This thing can build custom software for you, mostly autonomously. It can monitor things happening on the internet that are relevant to you, in a holistic and flexible way. We have one crawling the web for local events we'll like, and it judges them based on what it knows about us, and it tells us about the best matches every weekend, which has yielded some awesome outings we wouldn't have known about. It reads the literature on a subject in seconds and uses it as context to help in decision support. It's not the same value proposition as a computer 3 years ago, where most people are mentally anchored on what a computer should cost. Having it at home means that you can use it as a personal agent that always puts your interests first, regardless of what ad model the commercial providers decide to put in, and you can stash in it your medical data, what you buy, what you make, your worries, hopes, and dreams, without worrying about that being used as training data, or worse, something to exploit you commercially. I think it'll become considered totally reasonable to consider spending the cost of a small car on a computer, for many families.
Also, a lot of companies are looking at how to run capable models locally to cut some of their (massive) cloud AI bills. An easy answer is worth a lot to them.
Nvidia is definitely working on a AI Rack machine fully built for on-prem uses for companies.
By this kind of logic we'd paying something like ~$10,000/month for our internet connections. The perceived value needs to exceed the cost but that does not make it the only factor to consider price with.
What makes this expensive & sell well is it's not very fungible at the moment. Where else are you going to get 512 GB of high speed memory with a well supported accelerator attached that you can throw in the corner of anyone's home and not really have them notice? There are plenty of lesser options, plenty of noiser/power hungry options, plenty of harder to support options, but not really something in direct competition at the moment. Even the next rounds of the integrated AMD/Nvidia solutions are only targeting 196 GB of much slower memory and compute.
I think the difference besides the supply crunch there is that everyone connected gets the ~same internet, just faster or slower. Quantitative, not qualitative difference. On the other hand, a computer that can run Gemma 4 8B versus one that can run DeepSeek Flash are different enough experiences that I'd say they're effectively a difference in kind. It's been a bit since we had such serious stratification in outright capability in computing, rather than just how long it takes to get something done, or how many of something it can serve at once. In the early 90s, I think there were a lot more of those "this computer can do this thing, this one just can't" scenarios.
Closest competition I see right now are stacks of 2-4 connected DGX Sparks, similar lowish speed high mem, and about the same cost/gig.
I like this train of thought. The inverse is saying that the cost of this computer is the value we give away to AI companies by doing compute on their servers with our data. And to take it another way, is the value to you, the cost of a small used car?
> And to take it another way, is the value to you, the cost of a small used car?
For me personally, not quite that valuable yet, but I think it's getting there quickly. Deepseek V4 Flash massively increased the value of local AI to me, to the point where it's displaced most of my Claude Code usage, its upcoming vision enabled version should bump it further, and it's only going to get better from there.
It's a lot faster, but a lot of it is also feeling free to discuss things I wouldn't be comfortable sending to Claude, with the idea that that info is now theirs in perpetuity. I got my genome fully sequenced recently (it's cheap now!), and I get a battery of blood tests every year. Wouldn't do processing on any of that with Claude, but local AI? Totally great.
And if I was running a company with a large cloud AI bill, I'd probably buy a wheelbarrow full of these macs. Cheaper, but also a more solid/predictable base to build on.
Enthusiasts buying these for fun are not the target market. These aren’t big sellers to begin with but a lot of the sales are going to companies where people have budgets for gear like this and can make a business case for it.
This is, sadly, probably a foreign concept to a lot of people who have only worked at companies where hardware purchases are viewed as something to minimize and everyone is stuck with the same low spec laptops that the finance department picked out. At companies where someone might have a legitimate use for a $20K machine, their fully loaded costs (not their salary) are $300K or more, and other teams like sales are spending thousands of dollars per week on things like travel and hotels for their job, spending $20K on a computer that’s going to last several years is not a hard choice.
Remember core customers of Apple studio type products is content creation/video editing, etc.
AI based tools are very useful here - thinks like object removable or cleanup etc, not just AI generation.
For example Apple mentioned performance increases for https://learn.foundry.com/nuke/content/reference_guide/air_n...
If I can get Sol level capabilities on a $20k machine, then it is well worth it for my employer to buy me that machine for work as a workstation. When you start paying in tokens vs subscription costs due to enterprise agreements, you really start to see how much cash utilizing frontier models at the frontier costs (and I'm efficiently using luna and other models where possible!)
Yeah, I'm not sure people realize how expensive ZDR/Zero Data Retention is, and how important it is to a lot of businesses, this kind of thing starts looking really cheap really fast if it's a reasonable substitute.
I don't think this is accurate.
Even using multiple windows in parallel for as many as 5-10 hours per day, I find that I am not fully using my claude max (20x) and chatgpt pro (20x) accounts. I can for sure use up the claude max account, but chatgpt either gives me a free reset before I run out of tokens or I just fail to use the full quota. The quota for Sol seems like 10x that of Claude Opus at the same level, and forget Fable, you can use a 5 hour quota in 20 minutes.
But lets do the math:
Lets say a 20k workstation can run 1 inference at a time at the same speed you get with Sol hosted by openai (big assumption) and run an equally capable model (big assumption).
Each month this gives you about 100-170 inference hours on a Sol 20x Pro account, and 720 hours (if you utilize 24/7) on the workstation.
Assuming a 36 month amortization before the workstation has to be replaced due to no longer being able to run frontier models or is too inefficient due to electrical costs or what have you:
The monthly workstation cost is about $550 capex and $150 electricity -> $700/month
You would need about 6 Pro accounts to reach that capacity, which would cost you $1200 a month.
But this fails because:
- You most likely can't utilize the workstation 24/7. Your work hours will be concentrated into 6-10 hours per day.
- During work hours you are capable of utilizing more than 1 concurrent session. 6 Sol accounts would support as many as 20-30 during working hours, not all the time but if you could burst to that many (don't forget sub-agents and agent directed parallel agent workloads).
- In 1 year the cost of Sol level models is likely to cost a fraction of what it does now.
this leads to:
Subscriptions are, and will likely remain, the best deal in town. Unfortunately, larger companies aren't able to do that. When your monthly token costs are in the $5-10k range, the local inference starts to look a lot more attractive
In the case where you pay for tokens without a subscription, the analysis is still very much not in favor of buying hardware.
The assumption previously used was that you can run a Sol level model on an M6 or whatever hardware $20k gives you. That is not true, it was an assumption made to show that even giving your own hardware every reasonable advantage it still loses.
Lets compare buying tokens of the best model you might run on your own hardware (still being unrealistic in favor of your own hardware) vs that same class of model on the market. I think one of the best models you might be able to run is GLM 5.4, but lets just look at chinese models generally:
$20k workstation, best case: $15k M5 Ultra 512GB, 36-month amortization, ~$440/mo. Runs a GLM-5.3-class model at ~30 tok/s. Saturated 24/7 it produces roughly 58M output tokens/month.
Buying those tokens:
The economics can never work in your favor for buying your own hardware here, unless you can utilize it or sell excess capacity and you have access to nearly free electricity. The reason is someone else can buy the same hardware at scale (or realistically more efficient hardware), park it somewhere with very cheap electricity, and sell tokens. They can get very high utilization that you are not likely to get.And keep in mind I am giving 'your own hardware' no overhead or maintenance cost, despite your condition that it's in a large corporate environment. In reality corporate IT would make it almost impossible to set up and your would need huge lead times to buy the hardware and get it installed.
I am very much in the 'buy your own hardware' camp. Here is a picture of my 2x 3090 + 2x 4090 server: https://photos.app.goo.gl/kDWgzDXQJq4Jwegk6
And here is my cluster of orangepi's: https://photos.app.goo.gl/Yg1najPG8d8qrKqP9
I think "ultra rich enthusiast" is in the right ballpark. There are people betting on being able to create their own revenue generating products and services with their own local hardware and very little operating costs. That may or may not make sense as a business idea. But people with wealth and risk appetite trying a new kind of business model and cost structure has a strong tradition.
Put another way: If $25k is the full extent of the start up capital costs, and operating costs are very low, that is a much cheaper business to start than most! The question is whether this is actually a useful model for a revenue generating business. I think that remains to be seen.
My guess is that there will be a few hits (which we'll hear a lot about - especially when someone actually pulls off "the first single-person unicorn", which I do suspect will happen someday) and a huuuge number of misses, which we won't hear much about.
The competition is a custom multi-GPU NVIDIA RTX pro desktop, which go for much much more. $20k is cheap for 512GB addressable memory. The old mac pro could easily be configured to cost that much.
For some , the idea that some proprietary information could potentially leak is enough to justify any price.
Other than the local AI crowd which is much recent it is professionals using Final Cut Pro for video editing, Logic Pro as a DAW and music production, Video transcoding, Photoshop and other tasks for high performance computing that don't need Laptops but want above 128GB of ram and prefer a Mac. Then there is the obvious group of developers that are making Apps for all of their products. Also, these are great for the workplace. AI is much more recent thing that Apple products were used for.
If Apple didn't sold these things they wouldn't make them but, also the level of marketing that Apple is talking about for AI is basically the new group they need to capture because the ones I just listed are already buying Macs and or easily to motivate with the other obvious CPU / GPU performance upgrades for code compilation, faster memory and video transcoding.
this comment has always existed behind every apple release, most especially anything vaguely pro-ish.
to answer your question : looking at the aftermarket availability of Apple's prior best and brightest : practically no one buys them.
"people here buy them" , well, 'here' is one of the most affluent groups of people in the world.
They're available as movie and television set pieces (undoubtedly disappearing into the home of someone close to the staff post-production), and for administrative/boss types that can slip the cost into a ledger somewhere that few will ever see.
It has been a hobby of mine every few years to check out the apple site and see how big I can option a machine. My record was when I was in high school years ago and was able to option some pro studio-ish apple desktop thing to like 61,000 usd out the door.
Movie set pieces, as a motivation for Apple making these high-end configs available? That makes no sense.
For one thing, you can’t tell from a movie what the specs are. A $999 Mac Studio looks exactly the same as a $20,000 one.
For another, Apple updates the industrial design on their products so rarely, a 6-year-old Mac, iMac or MacBook also looks nearly indistinguishable from a brand-new one.
Movie production (editors, sound design, and a portion of fx work).
You can connect up to 4 of them via RDMA so 2TB total RAM.
Its cheaper than Nvidia AI hardware.
also way slower if we're just going against 'nvidia hardware'.
Even at high announced pricing 4 of them still much more accessible than any Nvidia solution with same amount of VRAM,
A number of people in these comments, it would seem.
So, rich enthusiasts it is, then
I'm going to buy one to watch youtube videos and surf social media.
If you want to run reasonably big, local AI models, what are your alternatives?
That may not be many people, but there certainly will be some people who want to do that, and are willing to pay big bucks to do so.
I know people who would get it for local LLMs for use in their company.
you can run open source models in the privacy of your own home :)
Local AI is the future, and a lot of people want the first mover advantage or to toy around with it. I know a guy with a small rack of Nvidia Spark machines that he uses for that purpose; it's as much as a decent used car.
But is it? If it’s cheap enough latency doesn’t matter. Unlike say cloud gaming, where latency does matter. I’ll take my games local and my text bots cloud
It's as much as a new car
Twitter users.
Companies used to have local server rooms in their office, like mini centers, and buy all the equipment end to end.
It’s when self hosting and local hosting was the norm, and why it’s also starting to come back.
There will be workloads that can never touch a public cloud, and for it solutions like this are an option.
Leasing now an option, only $50/month (cheaper than inference subscription?), so even cash-poor can go the amortized-investment route.
I've often felt there is tremendous value locked up in underutilized old computers. It would be interesting to see Apple in 3 years offering compute as a service using lease returns (or more likely, partnering with someone else to operate it (perhaps exclusively in secondary markets like China or India, to address political demands for local siting or jobs). Apple is in the best position to work around or even gap-fix older software/hardware limitations in a controlled environment, and now they can do so without cannibalizing new hardware sales.
I looked at multiple configurations, and mathed it out. With leasing, you pay ~75% of the capital cost (excl. tax) over 3 years, but end up with no asset.
Apple computers tend to have excellent resale value, and Mac Minis/Studios have the least depreciation of them all. I understand the benefits to both taxes and cash flow, but boy is Apple winning big on those lease offers for Studios.
> Leasing now an option, only $50/month (cheaper than inference subscription?)
For which configuration, though?
>fluid frame rates in demanding games like Mixtape.
not getting on that bandwagon but wasn't that not the most demanding game as its a just a nonstop cutscene.
it's not super well optimized I think. unreal engine 5 is taxing even without a lot of gameplay or stuff on screen
I'm not sure who that line is supposed to impress. Gamers focusing on graphically demanding AAA games would laugh at this. People who don't game much probably won't know whether this is good or not.
The M6 is useless for AI. Is there any model which is actually useful and fast on 32GB?
I somehow find it better to give 2 frontier model companies 100-200/month than dropping 10 grand on a hardware that will get old in no time with bad TPS. I really want to have a fully local model but seems like one more generation wait and we will be there?
You just described why the datacenter business is hard and as a corollary why space datacenters will not be economically viable.
Lots of people use the Mac Mini to run the frontier models over night or while traveling. I have a rack in my basement and have thought about throwing one in. You can get a cheaper machine but Apple feels a little more, “rack and forget,” if you have less price sensitivity.
Mac Mini + MacBook Neo w/ ssh can be a better setup than MacBook Pro for many people.
The page says 170G/s memory bandwidth for the NPU and 1.2T/s for the GPU. Why the discrepancy if it's all "unified memory"? The former is nothing to write home about as far as AI compute is. The latter is really nice.
Which one is it you can run local models on? I suppose the NPU only.
I think you misread, it’s 170gb/s for base M6 model and 1.2tb/s for M5 ultra.
Unified memory is about address space. The bandwidth is still determined by bottlenecks to the processor. CPU/RAM links are still fairly narrow.
96GB -> 256GB upgrade costs 4000 GBP in UK or $5460. $34 for GB.
In US its $4000 upgade so $25 for 1GB.
Also:
> 512GB memory option for M5 Ultra coming late October
That US price is before sales tax no?
Correct. It is better to go to delaware and purchase it.
Delaware? For people in California, Oregon is way closer.
Or just use Privacy.com and use an address in Delaware. Then you can buy it where ever.
How does this even work? You need to get the item delivered, and sales tax will incur in the state where it is delivered.
VAT?
I think the $34/GB figure might be inclusive of VAT and the $4560 not, which would be $28.5 otherwise. Not sure.
Oops sorry its just typo. Its 5460 not 4560.
AFAIK UK VAT is 20% and it's 27% higher price. Its just what you get for living in UK I guess.
I was just comparing this to an rtx6000 96gb build and when the f*ck did nvidia double the price?
If you want to comfortably afford this gen you had to trade options on memory stocks...
Can anyone recommend the perfect sweet spot for someone who wants to run their own inference?
Thinkstation PGX maybe?
Got the recommendation from these articles: https://www.xda-developers.com/qwen-3-8-27b-reverse-engineer... https://www.xda-developers.com/lenovo-thinkstation-pgx-revie...
But haven't had a chance to try it myself.
For me it's be Strix Halo, 128gb machine, especially running Qwen models. Except when I bought it, it was $1,900, now it's $4,600 for the same box. (Wow that's insane)
For tinkering and learning, it's been great. Tie it into something like Hermes and you have a pretty powerful AI assistant in a box. And when you need to step up your model, you just do something like OpenRouter and it makes it pretty easy.
I bought a 128GB M4 Max Mac Studio a while back, and for a while I thought like I had done really well to buy it when I did.
The problem I'm having now is that no models are targeting RAM of that size. Everything is either much smaller, targeting laptops, or much larger, targeting hardware well out of reach of enthusiasts.
Please, AI people, start making models targeting 128GB machines again. The last interesting one was Qwen 3.5 122B.
Great news. Qwen 3.8 Flash Next (125B A6B) is coming out tomorrow. 4 or 6-bit should run nicely on 128GB.
Should bench better than Opus 4.7.
I have had a difficult time with running 120b models on my 128gb setup, especially with any larger context size. The 6bit of Qwen 3.5 is already just over 100gb, and when you go down to 4bit it seems a bit lobotomized.
Well the upside is that you can run a laptop-sized model and still have enough memory left over to run a couple of Electron apps.
Call me when Linux will be officially supported. After that I can look and see what else can I get for the price and maybe then ...
FWIW scaling up from https://huggingface.co/avlp12/Qwen3.8-27B-Alis-MLX-6bit and some other sources:
You might expect the M5 Ultra to produce 50 t/s from Qwen 3.8 27B with a good context length.
Tangential, but what would be the ideal Mac option for home movie editing, casual gaming and amateur CAD fiddling in Fusion?
I plan on maximizing my residual student benefits, and taking advantage of education pricing.
The regular Mac mini will blow you away, search YouTube for video editing reviews using different Mac mini’s.
I use both the base M4 Mac mini and an M4 MacBook Air for Final Cut Pro, Photoshop, and Fusion, and while they're not as almost-always-perfectly-smooth as my Mac Studio, they're about 100x better than the experience I used to have on my old Intel MacBook Pros in the 2010s.
I edit 4K ProRes and H.265 footage, sometimes with multicam (up to 4 streams) and color adjustments, titles, etc. It's only after stacking 3-5 effects before things can stutter, really.
Or if you try doing something CPU-intense in the background _while_ running some heavy creative software. I just don't do that.
Very well timed for John Ternus's first quarter.
Interesting that "coding" is now part of the marketing brochure as one of the use cases, while that was historically kind of missing. Is this new?
Damn. I just bought a maxed out MacBook Pro M5 Max 128GB 8TB, still waiting for it to be delivered. I could get 256GB RAM M5 Ultra 1TB for roughly the same price, and it's double the memory bandwidth. Which one would you recommend? I do plan to run local LLMs.
Contact Apple care and see what they can do. They have been known to do upgrades if you buy around the same time. But with the ram situation that courtesy might be gone now.
Gonna go sell a kidney, should just about cover a base Mac Studio. Guess I'll need a payday loan for the power cable
~12k for 80 core gpu with 256gb, 14k in October for 512gb. Seems like that could make for a very descent on prem inference server.
It cant be 14k for 512GB because 96 -> 256 upgrade alone cost $4000
Sick. Particularly stoked for the 10gb network card ($100 option) when using the Mac Mini as a server. Just wish the memory + NVMe prices could come back down to pre ai-goldrush prices. As $2999 for the M5 Pro with 64GB RAM feels painfully over-priced.
As someone who is a beginner at home networking and have a Mac Mini running as a Plex server at home; what is the use case for the 10gb network card?
As long as everything is using 10gb, faster transfer speeds when moving data between computers. For downloads, it wont help unless you have 10gb fiber, but for most folks, 2.5gb is quite fast. Hell my spinning media NAS has a hard time saturating when moving files between internal servers.
My example is I'm looking for a new media creation machine to replace my homebuilt PC from 2015. Since that old machine can't run Windows 11 and also because Apple storage is so expensive, my idea is to turn the PC into a Linux storage machine with a 10G NIC. Then I should just be able to edit off of the storage instead of worrying about caching it locally.
The prices are ridiculous though. I may just keep rolling with my Windows 10 setup.
It feels stupid having my internet be faster than my computer can connect to it
RAM and SSD in apple gear has always been way over-priced. There was a short blessed period in March where the M5 Max macbook pro was out, but the general 30% price hike had not yet happened. In this period, given the insane inflated RAM prices, the price apple was charging for the M5 Max with 128 GB RAM was actually _reasonable_.
If I compare to like, January 2024, the prices for RAM these days make me want to weep.
Usable ram amounts in late October
Usable, not affordable.
Subjective
Maxed out Studio is $30,000+ tax in Canada if financed through Apple.
That's wild!
The storage is pretty silly. Bring that down and the max isn’t nearly as bad, more like 12k
I would love to see real LLM performance benchmarks for these machines. Apple statement regarding LLM performance seem little vague.
I have Mac M1 Max and I'm quite happy with it. But these advances make me think that maybe I should upgrade to Mac M6 (or something) when it is released.
Every time Intel/AMD gets close enough, Apple just crushed competition in terms of SoC. I don’t know who’s on that chip team but they’re world-class. Hope they get paid more than a bunch of AI idiots Meta hired to do absolutely nothing (but I know they are not even close).
M5 Pro in a Mac Mini with 64GB RAM and 10Gbit Ethernet seems like the perfect Jellyfin server and Ollama test server. All for just over $3K (I specced with only 1TB local nvme).
Probably worth it for Ollama but you can run Jellyfin off a raspberry pi.
Yeah, if you're transcoding for clients a lot I'd just spend the cash on upgrading the clients to support HEVC at minimum.
> you can run Jellyfin off a raspberry pi.
This may depend on the size of your library. I tried installing Jellyfin on a Synology NAS, which runs Plex just fine, and it ran so poorly it was basically unusable. It “worked”, but it was painful.
not sure which cpu that Synology has, but I run jellyfin on ugreen nasync dxp8800 plus which has intel with QSV and it breaks no sweat in both transcoding (if needed) and serving over 10GBE.
It's a Synology DS720+ with a Celeron J4125, 2GHz, 4 cores, 2GB of RAM.
I thought it would be fine, because Plex has no issues, but it was painful. Every client I tried on the AppleTV was equally painful, and I didn't even try using a client until making sure all the metadata was downloaded and setup via the web UI. I was very deliberate and did one thing at a time, spending a whole day on it (mostly waiting and browsing to different screens to force metadata to get downloaded and cached).
You don't have enough RAM. Not sure why you wouldn't use a cheap PC to run jellyfin on and then just use your NAS as the media pool.
Why would I buy and manage a whole PC for Jellyfin when Plex runs off the NAS without issue?
With everyone saying Jellyfin can run on a Pi, I would think it would be better optimized for low-end hardware.
True re: the Raspberry Pi, but I was thinking the 10Gbit Ethernet would allow more concurrent streaming in my household. But it's probably overkill.
If your household consists of more than 15 people consuming 4K content at the same time, the 10G upgrade may be worth considering.
Edit: never mind, 2.5G is now the default, so you'll probably need more than 40 people to saturate that with streaming video.
Wouldn’t it be amazing for Apple to give us a MacBook Air 15” M6 with a 15W sustained passive TDP capability?
Just amazing engineering push, the competition got the message and we benefit.
0% APR for 12 months (24 for iPhones only?) from Apple Financial Services for a device that can approach or even exceed what we were paying for new cars just a few years ago. Apple is definitely making bank off these financing offers, and with very little risk as unlike a car these Mac Studios don’t lose 20% of their value when you drive them off the lot.
Interesting times, to say the least!
How are they "making bank" by giving out 0% loans?
My friend, the same way everyone else does. It goes from 0% to 28% interest when you miss a payment. Those are rates normal lenders fall asleep dreaming of.
A Macbook Neo with an M6 and 16 GB RAM at $699/€699 would be a killer feat
Saw a 768GB RAM mac coming soon, would wait for that.
When will Apple's Mx CPUs use Intel's 18A/18A-P/14A node(s)?
Maybe in a year or so or maybe never. It depends on how 14a turns out and if it is comparable to TSMC 2NM. They may also choose to utilize 18a-p / 14a for other chips and not the M-series.
From what I’ve noticed, Apple products have been getting worse in quality year after year. Sometimes they even ruin their own devices with updates... I guess it’s all because of marketing.
Extraordinary claims require extraordinary evidence.
I’m mostly talking about their iPhones, where new updates sometimes make older models worse. I know a lot of people whose iPhones started lagging after iOS updates.
When is the m6 air coming though
> M6 supports up to 32GB of unified memory to multitask across demanding apps
I can't believe that Apple still comes with this bullshit like 32 GBs is a lot. It's a lot for video memory - vRAM, but not RAM.
The M6 is a base level chip, for normal users. Anyone needing more than 32GB of RAM is likely going with higher end chip that supports more RAM.
Is Apple just going to announce everything silently from now on? No more getting excited for the big events to see what's new -- it just appears on the blog one day?
Also: "a staggering 1.2TB/s of unified memory bandwidth" -- yay, the GPU has reached the year 2020! (I'm a bit bitter that my M4 Max is near useless for local LLMs because of its low memory bandwidth.)
I assume they'll have events for the MacBook Ultra and M7.
I should have bought a 100 m4 mac minis when I had the chance. Thanks hyperscalers for buying all the supply and renting it back.
A m4 mac mini is better than al of these per dollar, msrp adjusted.
Hopefully by the end of the decade China figures out manufacturing at scale and fixes this.
How much ram would each of those have 100 had?
16gb unified
Aren't you kind of stuck with smaller models on the mini though? Even with the pro you'd be stuck with a max of four daisy chained over thunderbolt and with the 160 gig memory bandwidth you'd probably be far better off with other configurations.
> M5 Ultra features a massive amount of high-bandwidth unified memory, up to 512GB, and delivers a staggering 1.2TB/s of unified memory bandwidth that is 50 percent higher than M3 Ultra.
Apple never needed to participate in the AI race to zero. Because they were already at the finish line years ago building their own chips that can run large >100B parameter AI models locally.
As someone who works in AI now, I have found it pretty amazing that Apple basically didn't do much with AI software, and focused more on the hardware side. I think this is what the future of AI is going to look like, local models run on your mac for your workflow.
It's possible that they're working on their own LLM that's going to work very well on their chips, and possibly outperform anything out there when they do release it.
>local models run on your mac for your workflow.
10 years ago 32GB ram laptops sounded too much. 8 was enough. These days even I would get that much ram since it’s soldered. 64GB is higher end.
In a few years we should see such high end hardware commonplace. Working with a local LLM to get work done is the ideal way to go which has mostly hardware limitation as of now that gets solved in due time.
> 10 years ago 32GB ram laptops sounded too much. 8 was enough. These days even I would get that much ram since it’s soldered. 64GB is higher end.
Ten years ago I got 64gb of ram in my laptop, same as I have now. I bought both for business and personal use. System ram capacity hasn't changed much in 10 years.
It makes me curious how old you were 10 years ago.
> Ten years ago I got 64gb of ram in my laptop
We were definitely outliers that long ago. I put 64 GB in a MacBook Pro back in 2019, and that was (a) overkill for everything I ever ran on that machine, and (b) stupidly expensive by 2019 standards (albeit almost affordable by 2026 standards)
>I have found it pretty amazing that Apple basically didn't do much with AI software
The iPhone 15 was almost entirely marketed based upon AI (I would say fraudulently so, advertising features they still haven't delivered), and a huge portion of the OS work was on local AI or AI integration.
And for that matter Apple has been dumping enormous sums into their own AI development. Their failure to have a lot to show for it doesn't void the fact that they tried really, really hard.
It's bizarre how often this "Apple sat on the sidelines and let the AI people fight...so smart!" narrative appears on HN. Apple hasn't gone down the path of spending hundreds of billions on nvidia GPU data centres, but they absolutely tried really hard to matter in AI.
|It's possible that they're working on their own LLM
Yep, Siri AI; they’re doing it in public.
‘Apple Foundation Model’
nth mover advantage.
1.2TB/s is 2/3 the speed of an nVidia 5090.
But you get a generic computer and much more RAM.
And you lose a couple of organs.
The real downside for me is not having Linux support.
It would take Apple one or two engineers to make Linux life much easier on macs. But Linux is outside their walled garden so it's ignored.
Same here. Sadly I think the voices like ours won't be heard, though, because Apple's looking for someone who's going to buy in on the whole ecosystem, and I think we're not it. Or at least I'm not.
I’m done with macOS.
My Mac Mini is strictly a headless server for llama.cpp.
I use a Linux workstation.
If I were limited to use Mac hardware , I would install Linux in VMware Fusion and work from there.
It's worth noting that the 5090 (or the RTX Pro 6000 big brother with 92GB VRAM) will run rings around the Mac when it comes to compute.
My old 3090 is typically significantly faster (almost 2x token/s) than my M4 Max 128GB machine, as long as the model fits in the 24GB of VRAM.
In most situations it's a better idea to just buy tokens. But there are definitely cases when that's not an option. And then a machine like the M5 Ultra can allow you to do things locally for a fairly limited budget. And in a simpler package to manage than a machine with multiple GPUs.
is it still effectively 2/3rds? Don't know enough to compare a discrete GPU/CPU setup to something like this where it's more integrated
There is no magic, if the data you compute as atomic chunk don't fit in cache then memory bandwidth R/W limit kicks in and architecture does not matter. On contrary - having multi gpu setup of same price and same memory size with even slower memories may give you effectively much higher bandwidth but at the cost of power consumption.
How much memory does that come with?
32 GB GDDR 7
I was so blown away at all the discourse surrounding "Apple fumbling on models". They should never have been in the model game to begin with. Apple crushes hardware over the last decade and that's a huge advantage today. In the end, massive models have proven to be very strong, but small models have proven to be good enough (especially with the recent Qwen 2.8 27B drop) and that's where I imagine the future will lie for consumers.
I suspected Apple would let everyone else blow all their money then, when the dust settles, deliver a better experience to end users and clean up.
Agreed. Apple doesn't innovate anymore, but they're generally pretty good at adapting once other people have.
I think this is a bit of a crazy statement. Everyone expects Apple to somehow build a category leading product every year. I'd expect something innovative every couple of years
* the iPhone * the iPad * apple watch * airpods * unified memory laptops and computers
Those are all products that either created a category or changed that industry.
They did participate early on with Apple Intelligence and failed miserably. Really good move to not double down and let the others explore the space first
Is there anything comparable that runs Linux, doesn't necessarily look as good, but is perhaps (a lot) cheaper/fixable? Or is this really pretty optimal?
I mean this is not nvidia based right? It's all custom? So we can use it under Asahi perhaps?
I want to get something for my company to run local models, wondering what would be a good option.
I love linux and would be using it if the ARM support was better. It's just not there and most distros that support ARM do it a little poorly. I just haven't seen anything even remotely comparable to Apple Silicon and unfortunately Linux is struggling very hard to support it.
It's not quite that ARM support isn't good on Linux, it's that there aren't high-performance ARM chips with strong general-purpose software stacks. Like the Raspberry Pi is very well supported, but otherwise the only upmarket devices are things like Ampere workstations and hyperscaler server chips.
You can't run Linux directly on these. Asahi Linux supports up to M2 only.
Linux runs very well in a VM on macOS. There are many good options for this, some free and open source (QEMU, UTM, Lima, Colima), some proprietary (VMware Fusion, Parallels).
But Linux in a VM doesn't get access to the real GPU, so model performance is limited. Those running on the CPU perform well, and those needing the GPU don't.
However, macOS on M-series macs is excellent for local models. (Maybe not as excellent as a box full of the best nVidia GPUs, but still excellent).
So if you're getting Apple hardware, like Linux, and want to run all of it locally, a fine setup for a machine to run local models, with agentic characteristics:
- macOS running one of the many local model runners. I used to use Ollama and Whisper, and now use llama.cpp instead of Ollama. Others use LM Studio, oMLX, etc. Provide HTTP endpoints to access the models.
- Linux in a VM for overall control and orchestration, with standard VM settings, and bridged networking so it appears as its own machine on your network. Also, in here provide a robust shared file server for shared state. Use this VM as your desktop and primary access to the machine, if you like Linux.
- Linux in a VM to launch ephemeral, volatile containers, with the containers using a memory-only tmpfs overlay on top of a read-only Linux filesystem in a VM disk image, with tools in this filesystem. Alternatively, a writable Linux filesystem in a VM disk image, with disk buffering set to use macOS host buffering and discard fsync requests. These settings optimise for container disk performance for data that's only ephemeral which will be deleted soon or on system shutdown. (You can combined both VMs, but need to use two VM disks to get equivalent behaviour, and be careful about VM disk configuration of the two disks.)
- Containers spawned within that second Linux VM can be spawned very quickly and run quickly, so are ideal for LLM agents that need a quick sandbox. These sandboxes generally run faster than a macOS sandbox, despite being on the same machine with VM overhead, because Linux is faster at some things. Teach the LLMs to store files and memories they want to keep in the shared file server.
I guess, what I mean is: Why are these tiny aluminum boxes so optimal?
I just want my butt ugly repairable beast machine to do the same trick. Why is my ram not unified? I have an iGPU in my server, but it can't access the 64 GB ram (I got last year for 150 euro) directly or something? It's on the CPU right? Why did only Apple go for this architecture? So many questions...
Strix platform maybe?
The PC platforms have anemic memory bandwidth in comparison. Eg, Strix Halo is 256GB/s max. If money is a bigger limiter than performance it can be an option though. As can Nvidia DGX Spark machines. (Also limited to 128GB memory and comparatively low bandwidth, but higher compute than Strix Halo.)
Asahi was stuck at M3 last time I checked it out.
Development on m3 is ongoing, m2 is supported
M2 even.
AFAIK, apple does not release drivers open source, asahi is a reverse-engineering endeavour and does not support GPU. For nvidia, there are both proprietary and open-source linux drivers. CUDA and inference works on linux with nvidia. I would recommend checking out this video of Alex Ziskind to shop for a computer to run local LLMs: https://www.youtube.com/watch?v=mevUEQcumzU&t=224s. TL;DR besides Apple he recommends, DGX Spark, Tenstorrent Wormhole N300, AMD Radeon 7900 and NVIDIA RTX 5090.
yesterday someone posted a link saying xiaomi "matched" apple's latest M series performance. Was that for less than 24 hours?
will be great fun if one M5 Ultra with 512GB memory at 1.2T bandwidth capable of doing 3x smallish local model inferencing each at Opus 4.5 level of intelligence.
The memory bandwidth and size seems to be there, but what is the tokens per sec on like a qwen model? And you can basically do 3x opus 4.5 on the $100 a month claude plan. Your payback will be near infinity years after electricity.
Well said
fucking awesome
> Apple’s developer frameworks and tools — including Core AI, Core ML, Metal, and Xcode
Can we please kill the xcode. It is worst pile of garbage I have to use just to develop ios app.
>M6 supports up to 32GB of unified memory
Is this a joke?
>Additionally, M5 Ultra features a massive amount of high-bandwidth unified memory, up to 512GB
Now we're talking. But at what cost?
The Pro and Ultra variants are the ones with the higher RAM amounts. They didn’t announce the M6 Pro yet.
Yeah, but they still write regarding plain M6:
"M6 also introduces a Dual 16-core Neural Engine, providing up to 2x the peak compute over previous generations to make on-device AI workflows run even faster" .
They've announced there won't be a M6 Pro in favor of getting M7 out the door.
At the price points they are hitting, can’t afford over 32 GB on the cheaper model.
256 memory gets you to like 11k. So like 15-20k.
512GB late october sounds lame.
and 512GB is so 2025.
Give us 1TB version. Where is the competitive spirit?
Yeah, for developers it should be 128GB base version.
>Where is the competitive spirit?
To be fair, where is the competitor at this form factor?
Cannot believe "moar transistors" is still the only idea they have.
They pionneered unified memory a few years ago.
No. Apple did not invent or pioneer the concept of unified or shared memory, but they did create an exceptional implementation of it.
I know they are a phone company, but I think they should focus on local models software, not only hardware.
They aren't a phone company, and haven't been ever. They're a hardware company first
More like an "ecosystem company".
It's the Hardware, Software and Services in combination. None would work without the other (to reach the scale apple is)
> They're a hardware company first
More specifically, they're a hardware dongle company first
Software wise there's plenty to choose from already. Ollama/llama.cpp, LM Studio, Lemonade, vllm etc. Anything Apple would bring to the table?
Maybe https://mlx-framework.org
FWIW, Ollama, LM Studio and Lemonade (and oMLX) also wrap Apple's MLX framework.
Apple has the mlx framework. Most/all major software for running models locally support it. Apple also has RDMA for interconnecting multiple machines across Thunderbolt connections.
False, they are a Marketing Company.
their strength has been hardware for over 30 years now
Uhhh, they are? They’re a hardware co for sure, but to say they aren’t focusing on on-device models is absurd on its face. They’ve spent over 2 years on Siri AI which is (mostly) local.
My 100k company only buys Mac laptops and you're calling them a phone company. Such an odd comment.