I know little about this space but the idea of “reverse-hyper-threading” (my favorite description from the article) is interesting for sure with so many workloads being single-core-bound. Kind of lets you have the best of both worlds with cores joining and splitting as needed to adjust to the current workload.
Of course I’m sure there are a ton of caveats but the idea alone is interesting, and somewhat surprising coming from Intel. Obviously filing a patent is way easier than shipping so Intel still has plenty of room to disappoint but cool idea nonetheless.
I know little about this space but the idea of “reverse-hyper-threading” (my favorite description from the article) is interesting for sure with so many workloads being single-core-bound. Kind of lets you have the best of both worlds with cores joining and splitting as needed to adjust to the current workload.
Of course I’m sure there are a ton of caveats but the idea alone is interesting, and somewhat surprising coming from Intel. Obviously filing a patent is way easier than shipping so Intel still has plenty of room to disappoint but cool idea nonetheless.
More discussion at https://news.ycombinator.com/item?id=45091921