Hmm. Does the BIOS report a Pentium at 100Mhz ? How did you read the cpu clock ?
Suggest Pull/ disconnect everything but the video card, and floppy drive. Run a DOS based system tool like HWinfo, Speedsys, Chkcpu, etc from floppy.
Could be a general board failure in the TX BGA chipset.... or could be something else....
One thing I mentally note is that the clock IC is in top right of pictures but the jumpers to set speed are way across the board at bottom right.
Not a design guru but that can be problematic no matter what board you work with (trace lengths, bad via's, cold solder joints, etc) with age.
Has the board ever worked proper in your possession ?

Pull all cards, go minimal with different cards.

I’m going to assume turbo button isn’t functional, but you could try a jumper on the pins (though that probably isn’t it)

Try a benchmarking comparison like speedsys. Are the benchmarks also lower?

Pull that cf card out. Bull the floppy cable out.

Hi Horun and Sphere478

Thank you very much for your suggestions. I tried running the benchmarks from a floppy drive with the CF removed and only a graphics card connected but this yielded the same results.

Horun wrote on 2024-04-14, 04:27:

Hmm. Does the BIOS report a Pentium at 100Mhz ? How did you read the cpu clock ? Suggest Pull/ disconnect everything but the vid […]
Show full quote

Hmm. Does the BIOS report a Pentium at 100Mhz ? How did you read the cpu clock ?
Suggest Pull/ disconnect everything but the video card, and floppy drive. Run a DOS based system tool like HWinfo, Speedsys, Chkcpu, etc from floppy.
Could be a general board failure in the TX BGA chipset.... or could be something else....
One thing I mentally note is that the clock IC is in top right of pictures but the jumpers to set speed are way across the board at bottom right.
Not a design guru but that can be problematic no matter what board you work with (trace lengths, bad via's, cold solder joints, etc) with age.
Has the board ever worked proper in your possession ?

Yes, the bios reports the correct setting on post. I also ran Speedsys (not sure why I didn't before) and this also reports the bus speed correctly at 66Mhz and the processor speed and multiplier correctly at 100Mhz @ 1.5. The board was faulty when I received it. The previous owner stated that it stopped working when he inserted 3.3V SDRAM without changing the jumpers from 5V DIMM to 3.3V SDRAM. After leaning expansion slot, CPU and ram sockets, I got it working to the point it is today. I measured the CPU clock with the CPU removed on the CLK pin of the socket (also checked multiplier this way). It could be that this isn't true any more once the CPU is inserted but speedsys still reports the correct values. I may try measuring this on the running board from the underside of the CPU socket. The clock chip is a PhaseLink PLL52C62-01XC for which I can't find a datasheet. As you mentioned the jumpers are on the other side of the board which could cause issues but speedsys still reports the correct values.

Sphere478 wrote on 2024-04-14, 04:54:

Pull all cards, go minimal with different cards. […]
Show full quote

Pull all cards, go minimal with different cards.

I’m going to assume turbo button isn’t functional, but you could try a jumper on the pins (though that probably isn’t it)

Try a benchmarking comparison like speedsys. Are the benchmarks also lower?

Pull that cf card out. Bull the floppy cable out.

I can toggle turbo on and off with CTRL + ALT + "+-". There doesn't seem to be a physical jumper on the board.

I got one more interesting result from the cache check on Speedsys. It looks like the L1 cache access has issues. I am not sure how the cache management works but this could explain the slowness. Below are two pictures with Level 1 cache disabled (first picture) and Level 1 cache enabled (second picture) . The issue is visible with L1 disabled as well which I don't fully understand. Any ideas what would cause such behavior?

What voltage RAM are you using now? Does it stay at the right voltage through the tests?

Those graphs very unusual I kind of suspect that you might have damaged your North Bridge.

I believe you said you tried other CPUs, and had the same result I hate to say it, but my hunch is that you have damaged silicone on that board. You can of course continue testing. I may be incorrect, but at this point that is what I am thinking.

Try different ram, maybe Simms if the board supports them

You can also try settings in the bios that deal with caching policies

The fact that this is even stable at all, does still make me wonder about some sort of turbo feature

I don’t bother reading your Clock pin while running, the probe may slip

Thank you for the additional suggestions. I tried both 5V EDO DIMM ram as well as 3.3V SD-RAM. Both work the same (still 25% speed). I am able to get through all benchmarks (doom, quake, etc), just at slow speeds. It is consistently at around 25%. So basically I have a nice 486 DX 33 here.

At this point I am also leaning to a damaged chip set. One indication for this is that the 1.15 kHz tone emitted from the PC speaker is present without anything on the board (no processor, ram, cards, EEProm), so it is being driven without code. The signal is a PWM like square wave oscillating between 4V and 5V. Quite strange.

Thank you again for all the suggestions. I think I will put this back on the shelve for now. I am not quite ready to move it into the donor pile yet though. I was hoping for potentially some damaged passive components but haven't found anything obvious and I don't know how passive components would lead to the observed L1 cache behavior.

Thank you again

If you have a donor north bridge you might try a transplant.

I had something similar with dead L2 cache chips.
Not that it's the case here. Just adding info.

sp3hybrid wrote on 2024-04-14, 20:10:

Thank you for the additional suggestions. I tried both 5V EDO DIMM ram as well as 3.3V SD-RAM. Both work the same (still 25% spe […]
Show full quote

Thank you for the additional suggestions. I tried both 5V EDO DIMM ram as well as 3.3V SD-RAM. Both work the same (still 25% speed). I am able to get through all benchmarks (doom, quake, etc), just at slow speeds. It is consistently at around 25%. So basically I have a nice 486 DX 33 here.

At this point I am also leaning to a damaged chip set. One indication for this is that the 1.15 kHz tone emitted from the PC speaker is present without anything on the board (no processor, ram, cards, EEProm), so it is being driven without code. The signal is a PWM like square wave oscillating between 4V and 5V. Quite strange.

Thank you again for all the suggestions. I think I will put this back on the shelve for now. I am not quite ready to move it into the donor pile yet though. I was hoping for potentially some damaged passive components but haven't found anything obvious and I don't know how passive components would lead to the observed L1 cache behavior.

Thank you again

It's always the RAM.
If possible try 2 times 72pin EDO RAM, like 2 times 8MB, 2 times 16MB or 2 times 32MB.
EDO DIMMs are quite exotic and normal SDRAM possibly too new or too high density for the board.