Hmm. Apparently there's a semi-rare Cyrix DX2-80 that runs at 5v, and a more common but still kind of hard-to-find ST DX2-80. Do these perform any better or worse than the Intel one? From what I can tell the Cyrix ones are a little behind but not by much, meanwhile the ST chips are completely unknown to me. Will they work with the "Intel" DX2 settings? I don't see why they wouldn't, but this board doesn't "support" other manufacturers chips since it's from before those really started popping up.

Hm, okay. It seems ST chips are just rebadged Cyrix designs, like TI or IBM ones. Now my question is, will a Cyrix 486 work on a board that predates the idea of having separate jumper configs for manufacturers other than Intel? The BIOS is very telling -- it doesn't even call it Intel because everything is just an "80486" to it.

AMD made a 5V DX2-80, but these run pretty hot. Even the AMD DX2-66 5V is hot.
The AMD should perform identically to the 5V Intel version (overclocked t0 80), but the Cyrix is a little better in some areas and a little worse in others.

I'd be fine with the Cyrix I guess, as long as it works in my board... at 25 bucks it's a little steep to buy the ST just on the offchance that it does work. I'd like it to be a sure thing, I'm told the Cx486s need BIOS support.

Ask around on CPU world. I think AMD or Cyrix 80MHz 5V should not be expensive.

After testing with the ISA super-IO card, it turns out the DX2-66 is plenty stable at 80MHz, it was just the VLB IO card that was already broken being even more so. I'm glad I held off on upgrading the CPU to one of the Cyrix designs, it likely wouldn't have worked with my board. I hear they need special BIOS/board support and while that's common on later boards, this particular one predates any of the "clone" chips and thus using anything not strictly Intel-compatible is a crapshoot. The DX2-66 at 80MHz is sufficient to play MP3s and it runs Duke3D more satisfactorily, but it's still not 100% ideal. I will likely hold out for a cheap DX4 OverDrive to pop up on eBay for cheap, meanwhile the DX2 @ 80 will tide me over for the moment. Hopefully the sample I end up with can cope with a 40MHz FSB...

Unfortunately this new IO card also has problems:

- With the SD-IDE adapter and CD-ROM drive on the same channel, what normally works fine (the SD-IDE seems to be hardwired to Master, the CD-ROM is set to slave) just doesn't; the "hard drive" isn't seen by the BIOS with the CD-ROM connected.

- The serial ports don't seem to work. I've double, triple, even quadruple-checked the jumper settings, I've wired the serial port every possible way on the headers, it just doesn't work. The BIOS says they're enabled at 3F8/IRQ4 and 2F8/IRQ3 respectively, but CuteMouse can't find my serial mouse at either port no matter what I do.

Since this card uses a Winbond chipset, and the busted VLB card uses a Winbond chipset, I've elected to go for something different. I've got yet another ISA IO card on the way, however, this one uses a UMC chipset. The VLB card I had at the start that worked fine used a UMC chipset, hopefully I'll get a repeat of the success I had with that card before I destroyed it in my ill-advised attempt to add functionality left off of that board.

Well, the new IO card has arrived, and it seems to work fine enough. It's a little finicky about what CD-ROM drive it'll accept as a slave device on the IDE channel, but I can't really blame it for that, even when it forces me to use a wildly inappropriate drive (a DVD+-RW burner, surely I won't be getting up to any of that on a 486...). Everything else works, or seems to -- even if it takes it a solid minute to, after seeking the floppies, actually start booting from the SD adapter. Serial ports both seem to work, the parallel port seems to work (even seems to have an ECP mode), and the game port was disabled in favor of the one coming from my sound card since it seems like you can't disable that one without also disabling MIDI output (at least on my CT2800).

The plan to acquire a DX4 overdrive and hope it copes with a 20% OC stands. The DX2 @ 80 is sufficient for the moment but I can definitely see it becoming an issue in the future -- Diablo is still rather extremely choppy at 80MHz, but it runs just well enough to suggest it'd be fine -- not perfect, but fine -- on a DX4 @ 120. MP3 playback is a mixed bag -- in the 486-optimized version of MPXPLAY it seems like with the spectrum analyzer and high quality mode off (I don't notice a difference with it or without it in output quality -- must just be some sort of filter) I can handle decoding up to 192kbps files so long as I don't touch the mouse and generate a ton of interrupts.

I ended up ponying up for the DX4 OverDrive, and initial testing seems great, it runs fine at 120MHz and the heatsink is barely bloodwarm. Benches complete well enough except for one thing!

I no longer have L2 cache! It does this at the 33MHz FSB too. Is this just a BIOS incompatibility, or some kind of issue with the CPU perhaps being write-back while my L2 is write-through? How would I fix this? Some benches are actually worse because while the CPU is 33 or 40MHz faster than the one it replaced, and it has double the L1 cache (and that is enabled, I checked with cachechk), it has no L2 cache. Do I have to just live with this? I'd rather not, Doom actually benches a little worse than the DX2 at 80.

Chipset is a UMC 480 and the BIOS date is 6/6/92. Would there perhaps be a newer BIOS I could have burnt to try? Maybe some other board that has the same chipset and keyboard controller that has a newer BIOS I could steal?

DX4 overdrives do not have writeback L1 cache. They only work as write through. There shouldn't be any problems operating it with writeback or write through cache. I usually experience this L2 cache behaviour on OEM boards with weird BIOSes, but yours is a Gigabyte, so it's kind of surprising. It sounds like the BIOS doesn't like your CPUID and automatically disables L2.
I've used DX4 overdrives on some earlier 486 boards (5V only) and never had issues. It usually seems to be the early 3.3V boards that have problems. Your board appears to be 5V only though, which is strange. Is a BIOS update available? Does yours have AMIBIOS? MR BIOS is available for UMC 480 chipset. I am willing to bet that will fix your problem.
It's pretty cool your chip works at 120MHz. I wish mine did.

How about a MR-BIOS ROM file repository?

Well, I jumped the gun a bit saying it works at 120MHz. It doesn't make it into Win95 or even reliably start Quake at that speed, and PCPBench is garbled... It posts, runs TOPBENCH (which with no L2 it loses to the DX2-80 in), runs Doom, and basically nothing else. So, no 40MHz bus, oh well -- it wasn't 100% problem-free at a 40MHz bus with the DX2 anyway. As long as I can get L2 working at a 33MHz with the DX4 I'll be satisfied. Sucks that it cuts my video performance by a quarter to use a 33MHz bus, but it is what it is I suppose.

Yes, this board has AMIBIOS, like I said dated 6/6/92. I wonder what ROMs are compatible with the socket it uses, and how easy they are to get ahold of -- and indeed if the programmer my friend has can even talk to them for flashing MR BIOS to one... I use AMISETUP to access hidden settings (mostly for memory/cache timings and waitstates) and I don't think it even knows what a CPUID is -- it now says 486DX or 487SX in the CPU field during boot, instead of 80486DX2. Misidentifies the CPU clock as being 80MHz no matter what the actual speed is, be it 100 or 120. It does make it into Windows and Quake at 100MHz, but with no L2 it performs significantly worse than the DX2, even at 66 in some cases.