With the 4x4MB and DRAM refresh set to standard I now passed the 33% test successfully as well. Those are fresh sticks made by Tiido, so I have very high confidence those are good. I'll let it run completely and hopefully it does a full pass with no error, and if that's the case I'll try some games again to see if I see no/less random freezes. 16MB takes a lot of time to test, though 😁 and is way more than the 4MB I wanted to use on this board. But heh...if it works, it works and I won't complain!

Fail -_-

really wondering if that's some memtest bugs. I guess I'll try to play some games again with those settings and the 4x4MB sticks at this stage...

I *may* have found the issue. It's very possible memtest really threw me off and made me waste a lot of time... trying more games using conventional memory only, I couldn't reproduce any freeze, which made me realized that my standard DOS menu I copied off my 486 with some emm386 customization may have been the culprit from the beginning. I have modified it to a very basic setup and so far so good...will update if something new comes up in my testing

So it works now even with memory manager enabled?

I thought so but was too quick. I still get random problems. Even worse, I thought I narrowed it down to memory manager and everything was fine without it, but I just had another similar freeze in that configuration 🙁 back to step 1...

I would try removing the cache chips and try without them. Just to be sure.

Id look at all ETEQ pins too

Good idea, didn't suspect anything in regards to cache. I will test that. Thanks for the tip!

rasz_pl wrote on 2024-02-25, 20:47:

Id look at all ETEQ pins too

Thanks, if cache ends up a dead end I'll get the microscope and give it a good look tomorrow.

Looks like the cache is a good lead! I removed 128KB last night, and a leg from a chip came right off...I suspect it may have been fragile already.
I disabled all cache in the BIOS and ran memtest again overnight, no error! It made a full pass and a half and was still running. So I re-enabled the cache in the BIOS and ran memtest again before going to work. I'll see the results tonight.

Well if one of the SRAM chips has been broken I think we found the culprit. I have a good feeling about this, hopefully all will be right with this board from now on.

Yep, that does look like a culprit indeed but I can't 100% guarantee that the leg was already broken and it did not happen during the extraction process..they were quite stuck, I used a chip extractor and had to apply non-trivial force to get them out but did it as cleanly as possible, but you never know, those legs are tiny. In any case, I'd rather be at home working on it than in the office but it'll have to wait for tonight 😁

Boohyaka wrote on 2024-02-26, 07:16:

Yep, that does look like a culprit indeed but I can't 100% guarantee that the leg was already broken and it did not happen during the extraction process..they were quite stuck, I used a chip extractor and had to apply non-trivial force to get them out but did it as cleanly as possible, but you never know, those legs are tiny. In any case, I'd rather be at home working on it than in the office but it'll have to wait for tonight 😁

The good news is those 28-pin, 32KB cache chips are readily available and even still being made, although I have found some of my UMC 481/491 chipset boards to be picky about brand

Indeed. But if it works I'll probably keep it running with 128KB. Not a build I'm trying to max out or anything, actually quite the opposite 😀 I have enough other machines if I need more performance!

what do you know, came back from work to close to 5 successful passes with no errors in memtest 😁 looks like that little cache chip bastard really was after it all in the end.

thanks Paar for putting me on the right track, I really overlooked cache, doh.

I may try to get back to the original 4x1MB sticks and replace the faulty cache chip with some spare I probably have around to get back to the 256kb, but at this stage I'd rather play some games to confirm overall stability 😀

That's good news. I'm glad I could help. Am in a process of fixing damaged motherboard myself (486) so we will see if I meet the same positive ending as you.

OK, so the computer is now rock stable even with the 4x1MB sticks. Even the RAM refresh BIOS setting back to "Concurrent" instead of "Standard" works fine. I can only assume this setting makes the RAM interact in a different way with the cache?

All in all it's a happy ending 😀 it will stay at 128KB, not like it makes a dramatic difference.