Z386 FPGA – The Z386 project is rebuilding an Intel 80386 on FPGAs by executing original microcode—an approach aimed less at beating faster CPUs and more at reconstructing how the original hardware behaved. It currently runs on Altera Cyclone V and Gowin GW5A boards at pe
On an FPGA lab bench, an Intel i386 isn’t just being approximated—it’s being reenacted.
The Z386 project. developed by [nand2mario]. takes a deliberately specific path: it implements an Intel 80386 CPU on programmable hardware using the original microcode. A recent blog post from the project lays out how the FPGA design is coming together. and why the choice matters. There are many ways to implement an i386-class CPU on an FPGA. but using original microcode is presented as one of the most compelling approaches for getting close to the real hardware’s behavior.
The project is built as a sibling to [nand2mario]’s earlier z8086 effort. which did something similar for the Intel 8086. The shared idea is straightforward—execute original microcode for compatibility—but the leap from 8086 to 80386 brings a different kind of complexity. The microcode isn’t just larger; the internal state and the instruction set balloon in ways that ripple through the entire design.
Z386’s blog discussion emphasizes that the jump isn’t cosmetic. The 80386 era adds a much larger instruction set paired with correspondingly more complicated internal state to manage. Newer capabilities include memory management, paging, and register debugging. It also includes extensions to protected mode that began with the i286.
That complexity shows up in the machine’s practical footing, too. Z386 currently runs on a handful of FPGAs, including the Altera Cyclone V and the Gowin GW5A. In real-world terms, it lands at performance “equivalent to a ~70 MHz i386,” though with slightly worse cycle efficiency. The project points toward one plausible reason for that gap: limited cache size—16 kB on these configurations—compared with the 32+ kB cache found in the fastest i386 CPUs.
Still, the point of getting the CPU running isn’t purely theoretical. The performance is described as more than enough to run software, with the project specifically naming games like DOOM.
What makes Z386 feel different from the usual race toward speed is the stated goal. The aim isn’t to outperform modern FPGA-friendly cores such as ao486. Instead. it’s closer to an archaeological reconstruction—bringing back the original hardware’s interaction with its microcode so developers can study what made the old system tick.
For now. Z386 is operating as a living museum piece: an i386-class processor recreated in silicon logic. powered by the very microcode that once drove the real chips. And judging by the ability to run full software like DOOM. the reconstruction is already convincing enough to leave the lab and make noise on a screen.
Z386 FPGA Intel 80386 microcode i386 open source CPU Altera Cyclone V Gowin GW5A DOOM reverse engineering retro computing
So it’s basically a 386 for people who hate progress?
I don’t even get it. Like they’re rebuilding a 80386 but on an FPGA? Wouldn’t a normal computer just run Doom way better anyway. Sounds like a nerd museum thing lol.
The 70 MHz thing is confusing me because my cousin’s old Dell had like 1 GHz or whatever and it still lagged sometimes. If it’s only 16k cache too then how is it anywhere near the original experience. Also they said microcode like it’s the same from back then but how do they know it’s exact? Probably close enough though.
Wait so they’re using the original microcode to make it act like the old chip, but it’s still not beating modern cores? Kinda feels like they’re doing this just to prove a point. Also FPGA boards are always expensive right? Meanwhile I’m just trying to stream Netflix, not run paging and protected mode on a lab bench.