Maybe it does not have the needed instruction set, because IDT C6 is basically a 486. But a 486DX will run Quake. Maybe there are some math instructions missing, a 486SX will crap out, if you try to run Quake.

Cyberdyne wrote on 2021-06-08, 05:50:

Maybe it does not have the needed instruction set, because IDT C6 is basically a 486. But a 486DX will run Quake. Maybe there are some math instructions missing, a 486SX will crap out, if you try to run Quake.

So why on the other p5a it runs fine? Does the other p5a provide the missing instructions? 😀

Yep. That does not make any sense. And the BIOSes are the same, compared them, they have the same checsum and CRC, sometimes a BIT can make a difference? Or just switch the BIOS chips, and try again.

The process of elimination is your friend .

First, I would try swapping CPUs between working and non working machine and see whether fault follows the CPU or not.

If thay is non conclusive, I would try swapping parts until either

a) the issue follows the part
b) having swapped everything, the issue is confirmed to be the motherboard
c) the issue is no longer reproducible after a given part swap and can be ascribed to one of the great mysteries of life

EDIT: I was under the impression that you have 2 IDT CPUs, if you only have one and it works in one boards but not the other, that points to something being marginal on one of the boards. Hard to say what, but capacitors are always suspects on a board of this age .

darry wrote on 2021-06-08, 06:28:

The process of elimination is your friend . […]
Show full quote

The process of elimination is your friend .

First, I would try swapping CPUs between working and non working machine and see whether fault follows the CPU or not.

If thay is non conclusive, I would try swapping parts until either

a) the issue follows the part
b) having swapped everything, the issue is confirmed to be the motherboard
c) the issue is no longer reproducible after a given part swap and can be ascribed to one of the great mysteries of life

EDIT: I was under the impression that you have 2 IDT CPUs, if you only have one and it works in one boards but not the other, that points to something being marginal on one of the boards. Hard to say what, but capacitors are always suspects on a board of this age .

Actually I have 3 idt cpus, my first try was to swap all of them on the non working motherboard (noone of them worked, obviously), then I used one on the other and it was working, after that I kept using the same for testing both motherboards since the cpu is confirmed working.

I swapped all the possible components, ram and video card.

I also checked the voltages on the bios, they are all in spec.

What is weird is that appparently anything else works fine, doom benchmark works fine, windows 98 works fine.

Swapping the bios didn't work, quake (640 or 320 doesn't matter) goes to the half of the benchmark and then reboots

Nemo1985 wrote on 2021-06-08, 06:55:

[...] […]
Show full quote

[...]

Actually I have 3 idt cpus, my first try was to swap all of them on the non working motherboard (noone of them worked, obviously), then I used one on the other and it was working, after that I kept using the same for testing both motherboards since the cpu is confirmed working.

I swapped all the possible components, ram and video card.

I also checked the voltages on the bios, they are all in spec.

Which spec exactly? Winchip C6 CPUs are designed to work at VRE (3.52V) not VS (3.3V) levels. What is VID0 set to? Should be 2-3 for one of these.

If it's already at 2-3, you could try bumping up Vio by setting VIO1 to 2-3 (3.6V).

What is weird is that appparently anything else works fine, doom benchmark works fine, windows 98 works fine.

Swapping the bios didn't work, quake (640 or 320 doesn't matter) goes to the half of the benchmark and then reboots

That rules out firmware or settings and suggests that it's triggered when the Winchip's feeble litte FPU is heavily loaded. As the CPUs themselves work fine, that really does sound like a voltage issue. If it's not the level, it's probably the stability - although you'd need an oscilloscope to be able to troubleshoot that.

I have a pair of boards with a similar issue I've given up on eventually: two MSI MS-6168 rev2 boards, i440BX with onboard Voodoo3. Same board, same revision. Every active component is identical, date codes just a month or so apart. Yet one works fine with Coppermine and even Tualatin CPUs, where the other simply does not boot with any CuMine, it's Deschutes/Katmai only. Swapping BIOSs etc didn't help. Still on the lookout for an affordable scope for this sort of stuff, but too many other priorities for now.

dionb wrote on 2021-06-08, 07:21:

Which spec exactly? Winchip C6 CPUs are designed to work at VRE (3.52V) not VS (3.3V) levels. What is VID0 set to? Should be 2-3 for one of these.

If it's already at 2-3, you could try bumping up Vio by setting VIO1 to 2-3 (3.6V).

The cpu vcore is 3.5 while the vio is 3.5 too, I tried to raise it to 3.6 but nothing changed.
Also with the Winchip 2 even if uses the same settings works fine.

dionb wrote on 2021-06-08, 07:21:

That rules out firmware or settings and suggests that it's triggered when the Winchip's feeble litte FPU is heavily loaded. As the CPUs themselves work fine, that really does sound like a voltage issue. If it's not the level, it's probably the stability - although you'd need an oscilloscope to be able to troubleshoot that.

Unlucky I do not have an oscilloscope, that stuff it's quite expensive.

What I noticed is that the pll chip become hot, I will later try the p5a which works with the c6 to check if that happens too.
Do you think a recap could solve the issue, despite che caps are ok to a visual inspection?

dionb wrote on 2021-06-08, 07:21:

I have a pair of boards with a similar issue I've given up on eventually: two MSI MS-6168 rev2 boards, i440BX with onboard Voodoo3. Same board, same revision. Every active component is identical, date codes just a month or so apart. Yet one works fine with Coppermine and even Tualatin CPUs, where the other simply does not boot with any CuMine, it's Deschutes/Katmai only. Swapping BIOSs etc didn't help. Still on the lookout for an affordable scope for this sort of stuff, but too many other priorities for now.

You'll get that as caps degrade too, works with high speed lowest volt CPUs, then will only take middling ones, then will only take previous gen with higher core volts... partly it's regulation capacity and partly it's noise, because the higher core volt ones can cope with more ripple, whereas lower core volt ones it's exceeded the threshold for internal logic high states.

BitWrangler wrote on 2021-06-08, 13:20:

[...]

You'll get that as caps degrade too, works with high speed lowest volt CPUs, then will only take middling ones, then will only take previous gen with higher core volts... partly it's regulation capacity and partly it's noise, because the higher core volt ones can cope with more ripple, whereas lower core volt ones it's exceeded the threshold for internal logic high states.

Relevant detail in my case: those MSIs are notorious for cap plague and I've already re-capped both and subsequently re-re-capped the problematic one. Didn't change behaviour at all.

But here with this P5A... who knows, if you've excluded everything else, you could try - but I wouldn't be too hopeful unless you find positive evidence that the caps are at fault.

Usually caps during socket 7 era were not a problem.

But I have no idea how to troubleshoot the problem.
I tried the other board and the pll chip gets hot too but it works fine.