June 23, 2026
386 and furious
80386 Early Start Memory Access
Old-school PC fans are freaking out as this homemade 386 suddenly gets way faster
TLDR: A homemade version of a classic 386 PC chip just got a big speed boost, making old games like Doom run much better without increasing its clock speed. Fans are excited, but the comments quickly turned into a mini-drama over why Windows still won’t boot and how the older design managed to work at all.
A retro computing project just had a full-on glow-up, and the comment section is treating it like a surprise comeback tour. Developer nand2mario says his FPGA-made version of Intel’s old 80386 chip now performs dramatically better after adding a clever trick the original chip used to start memory work a little earlier. Translation for normal humans: the fake old chip got a lot snappier without even raising the clock speed. The headline stat that made people sit up? Doom jumped from 16.6 to 23.0 frames per second, beating another well-known FPGA core in that test.
And the community mood is a delicious mix of hype, curiosity, and nerd-detective energy. One commenter basically summed up the wholesome excitement with, “So proper 386 on an fpga? How cool is that.” That’s the fan reaction in a nutshell: this isn’t just a speed patch, it feels like a love letter to classic PCs. But naturally, the comments didn’t stay purely celebratory. Someone immediately poked the sore spot with “I wonder what exactly stops windows from booting” — the kind of question that turns a victory lap into a mystery thread. Suddenly it’s not just “wow, faster Doom,” it’s “okay, but what’s still broken?”
Then came the armchair engineering sleuthing. Another commenter wanted to know how the older version even worked before this change, asking if it was rushing calculations or stretching extra cycles. That’s the real drama here: the project is winning praise, but the crowd is already demanding the director’s commentary. In retro tech land, success doesn’t end the debate — it starts a new one.
Key Points
- •The article explains that Intel’s 80386 used an Early Start mechanism to begin the next instruction’s memory-address work during the final cycle of the current instruction, reducing effective memory latency.
- •The author added Early Start and other CPI optimizations to the z386 FPGA core, improving benchmark performance without changing the 85 MHz clock rate.
- •Reported z386 performance increased from 16.6 to 23.0 FPS in Doom, from 33.7 to 44.5 in 3DBench, and from 147 to 170 in Landmark, reaching or surpassing ao486 in some tests.
- •The article uses microcode and an `add eax, 16` followed by `mov ebx, [eax+4]` example to show how Early Start overlaps writeback, effective-address calculation, relocation, and memory-read issuance.
- •The article states that Early Start requires forwarding to handle register write hazards, and that a corner-case flaw in the 386DX forwarding network caused the POPAD bug.