kevmif wrote on 2021-07-24, 10:24:

I tried swapping the IDE cable as the factory one is very short and has a nice crease in it from how it is fed up through the ch […]
Show full quote

I tried swapping the IDE cable as the factory one is very short and has a nice crease in it from how it is fed up through the chassis to the hard drive.
I tried two different IDE cables but in both cases the drive would not spin up. The only difference is the newer cables were dual cables, not single. Do I need some sort of 'special' IDE cable for this thing.

I think the issue relates either to the cable, or the drive controller.

Can I put a separate ISA I/O card from a 486 in to this thing or will the bios just not acknowledge it?

Are you sure you connected the non factory IDE cable the proper way ? If the end was swapped the drive typically will not spin up.
If you can disable the onboard IDE then you should be able to try a basic ISA IDE controller.

kevmif wrote on 2021-07-24, 10:24:

I also noticed that directory listings seem to be slow to return to the prompt after listing the files if that means anything.

DOS displays the amount of bytes free when you run DIR. DOS is known to be quite slow on counting free clusters, so the first time you run DIR, it might take a couple of seconds for the free space to be counted. The value is cached inside DOS, so if you run DIR again, the "slow return to prompt" behaviour shouldn't repeat.

kevmif wrote on 2021-07-24, 10:24:

The only other things I have noticed is some very minor artefacts (a few brief fuzzy lines) when the entire screen changes in some games (had Jazz Jackrabbit and Wolfenstien 3D going on there)

If the fuzzy lines are at the top of the screen, this is a known shortcoming of some VGA cards: When you modify the color palette for fade-in / fade-out process, the game doesn't manage to change all 256 colors during the retrace period. On some VGA cards, adjusting the palette during the display period causes visible artifacts.

Find a good memory testing program and run it for a few hours.

I don't have a specific recommendation but you want something that is small, so that it can test as much of the memory as possible. On later systems I use memtest86 booted from a USB key. There should be plenty of old utilities for a 386 class system.

Even better is a standalone SIMM checker, like the old Innoventions RAMCheck and SIMCheck units. But those are not too common.

Bad memory can lead to all sorts of flakiness. You definitely want to test for it, even if it helps you eliminate the memory as the source of the problem.

Thanks everyone. I am reasonably sure the disk is solid as I put it in another PC and ran several iterations of NDD. Is it possible that I have selected an incompatible type in the set up program? My guess is no, as I would have expected NDD to totally crap itself if that were the case.

Horun: It seems so obvious in hindsight but the Compaq machine has the cable arse about because it is designed to be twisted. So I need to find a cable without the 'key notch' and without the blocked out pin (or just file down the key notch). Damn propitiatory crap. The pre 486 era really was the wild wild west of PC days wasn't it. Thank you for your help. I am confident I can now verify the IDE cable (with a bit of tinkering). Thanks for your comments about the IDE controller also. Yes the fixed disk controller can be disabled. That will be next step if the replacement cable doesn't sort it out.

mkarcher: Thanks for the insights on the directory listing issue. With the VGA issue, yes that seems to be what I am experiencing. I have a good old Trident card here (two actually) to test with when I hit the corruption issue again.

mbbrutman: Solid advice. Thank you. Will do that next. Your username sounds very familiar. Are you the mTCP guy? If yes, thanks for such a nifty little project. Keep up the great work.

Edit: hard lock in memtest86+ several hours in to testing. No memory failures reported as such but the hard lock is super concerning. I will remove the sound / network cards and try again.
Edit2: Not actually hard locked. Doing a 'random number sequence' test which is super slow. Gave the impression of a hard lock, especially because the keyboard lights (num / caps lock) didn't respond which gave the impression of a hard lock combined with the fact that this particular test takes over 30 seconds per percent. Will leave it running and see what happens
Edit3: 4+ hours of memtest86+ with no errors reported. I think it is safe to say the RAM is good (or at least doesn't appear to have any major faults).
Edit4: Replaced the IDE cable noting that the factory IDE cable has a twist in it and doing the same with the replacement cable. Running NDD now but so far so good. Guess we'll see how it goes.

Yep, that's me.

I'm glad to hear it is not memory, but that just means the mystery continues.

Another suggestion: replace the IDE cable with an 80 conductor one. While doing early testing and debugging on the original XT IDE (10+ years ago now!) I found some hard drives would not work with a normal 40 conductor cable, but the better 80 conductor cables were fine. They both have 40 pins, but 80 conductors in the cable reduces crosstalk between the signal wires. On a marginal circuit that can make a difference.