ao486: x86-compatible Verilog core implementing all features of a 486 SX (2014)
github.com54 points by csmantle 6 days ago
54 points by csmantle 6 days ago
It’s a neat core, but z386 is where the action is right now (https://github.com/nand2mario/z386_MiSTer). Interestingly this core doesn’t output a proper analogue signal. I always wondered why until I recently tried to implement a real EGA pixel clock on the PCXT EGA core. Basically, video cards with dual (or more) pixel clocks are not realistic on MiSTer. The PLLs are there, but once you get too many clock sources the routing becomes quickly impossible. MiSTer framework’s pipeline also forces pixel clocks tied directly to video output (for good reason) and switching that clock at runtime gets very tricky.
Thanks for mentioning z386. Much of its recent progress was made possible by the reverse engineering and disassembly of the original 80386 microcode (https://news.ycombinator.com/item?id=48247004), which z386 uses as its control program. That work has been especially valuable for reproducing obscure protected-mode behavior correctly.
With correctness helped greatly by that, development is now focused mainly on performance. In Doom, z386 is currently about 30% faster than ao486 on the same MiSTer setup, with benchmark performance roughly comparable to a 486DX2-66.
I'm pretty sure there was a more recent 386 clone than that
z386 is in development right now, and isn't even finished. Do you mean between ao486 and z386?
Curious: are these cores (ao486, z386) technically illegal because of Intel/AMD licensing/copyright/etc? Or has everything relevant expired? Surely the patents have but not sure what else comes into play.