I've been experimenting with making an IBM AT (the "braindead" 286) compatible with a real CMOS 25 MHz 286 and a real 287XLT with 16 MiB of RAM while shoving all of the BIOS ROM, bus, keyboard controller, and glue logic in an FPGA. This seems to be a more flexible and genuine approach rather than total emulation that's bound to be flawed in subtle, inauthentic, and incompatible ways. It's not completely authentic compared to a model 5170 type 3 but it aims to be most likely to offer the best performance, flexibility, and faithful reproduction of historical bugs. I try to stick to known good BIOS dumps and schematics to be sure the functionality is 1:1 identical, but may have to tweak things like wait states and such to make it work. 6 16-bit ISA and 2 8-bit ISA slots, fingers crossed. I'm thinking about using an external battery-backed RTC that's better than an MC146818 but the FPGA presents the functionality of it.
I've been experimenting with making an IBM AT (the "braindead" 286) compatible with a real CMOS 25 MHz 286 and a real 287XLT with 16 MiB of RAM while shoving all of the BIOS ROM, bus, keyboard controller, and glue logic in an FPGA. This seems to be a more flexible and genuine approach rather than total emulation that's bound to be flawed in subtle, inauthentic, and incompatible ways. It's not completely authentic compared to a model 5170 type 3 but it aims to be most likely to offer the best performance, flexibility, and faithful reproduction of historical bugs. I try to stick to known good BIOS dumps and schematics to be sure the functionality is 1:1 identical, but may have to tweak things like wait states and such to make it work. 6 16-bit ISA and 2 8-bit ISA slots, fingers crossed. I'm thinking about using an external battery-backed RTC that's better than an MC146818 but the FPGA presents the functionality of it.