I doubt that something gone wrong with CPU, but some Tualatin ready board could be handy to narrow this down to one component.

Was the system stable before and, with no changes in hardware, has become unstable overnight?

Do you have the modded BIOS installed? My unit works flawlessly with the modded bios and a Powerleap + 1,4/512/133. shortsightedness...

I have put back the Coppermine 1000EB on Asus Slotket and everything is stable again. The only recent modification to the system, other than the CPU change was the Promise Ultra133 TX2 . I had previously been running a SIL3114 with no issues and then a Promise SATA150 which died (which is why I had put in the Promise Ultra133). I have since reverted to the SIL3114 and am now running informal and formal (Prime95) stability tests under Windows 98 SE (not Linux) with the Coppermine 1000EB (at 750MHz), to establish a baseline .

As an interesting side note, the system will refuse to boot from a hard disk connected to the Promise Ultra133 TX2 or from a floppy if the Ultra133 TX2 is installed . The Ultra133 TX2 detects the connected disk without issue and booting from a Linux DVD works without issue and gives access to the Ultra133 TX2 connected HD . This is why I never noticed the issue with the Ultra133 TX2 in this board as I was running all my tests under Linux . I have replaced the Promise with a SIL3114 (which I had been using before in this board), which works fine . This is quite strange as the Promise SATA150, which is more recent, worked fine before its demise (not detected anymore).

EDIT:

The Serpent Rider wrote on 2020-07-08, 13:37:

I doubt that something gone wrong with CPU, but some Tualatin ready board could be handy to narrow this down to one component.

The only natively Tualatin capable board I have is currently working fine with another Tualatin 1.4GHz 512K CPU that had a broken pin which I "fixed" . I would rather not touch that if I do not absolutely need to .

The 512K PIII-S chips are much harder on aging VRMs than Coppermine and even Tualatin-256 chips. I had the exact problem with an Abit Coppermine board years ago. It worked fine with Coppermine chips, even at 150MHz FSB. It could also run a PIII-1200 and Celeron-1400 (both Tualatin-256) fully stable. However, with a PIII-S 1400 inserted, it only POSTed around 10% of the time at 133MHz FSB; 50% if I clocked the FSB down to 100MHz.

I also had an IBM BX board that would happily run a Tualeron-1400 all day, but hang at the hard drive detection stage of POST with a PIII-S 1400.

It's really too bad, as the doubled cache on the PIII-S really does wonders on an FSB-constrained CPU.

Standard Def Steve wrote on 2020-07-08, 15:58:

The 512K PIII-S chips are much harder on aging VRMs than Coppermine and even Tualatin-256 chips. I had the exact problem with an Abit Coppermine board years ago. It worked fine with Coppermine chips, even at 150MHz FSB. It could also run a PIII-1200 and Celeron-1400 (both Tualatin-256) fully stable. However, with a PIII-S 1400 inserted, it only POSTed around 10% of the time at 133MHz FSB; 50% if I clocked the FSB down to 100MHz.

I also had an IBM BX board that would happily run a Tualeron-1400 all day, but hang at the hard drive detection stage of POST with a PIII-S 1400.

It's really too bad, as the doubled cache on the PIII-S really does wonders on an FSB-constrained CPU.

The PowerLeap adapter, which has its own CPU VRM, last worked fine with the same type of CPU on a different board about 2-3 years ago and has been in storage since . I suppose something might have gone wrong with it in the meantime .

Having found no issues with the Coppermine setup, I plugged the PowerLeap with Tualatin 512K back in . This time it seems to POST every time, but the P3B-F hardware monitor complains about the CPU voltage being 2.1V (even when board set at 1.55V) and CPU temp being 127C !!! It did not do this at the last attempt . The values returned are obviously either irrelevant (PowerLeap adapter has its own VRM) or false (CPU obviously is not at 127C) . I will clean the contacts on the PowerLeap and MB slot .

That said, having ignored the warnings, I pressed on into Windows and am running a Prime95 run (at 1050MHz). We'll see what happens.

EDIT: I already tried re-seating the PowerLeap to no avail .

I'll just say this: oof!

Good luck, I've sold my Ms6905 MASTER, and really regret doing that, as it is easily moddable for Palpatine!

Seventeen minutes of Prime95 on the Tualatin 512K passed by without issue .

I tried cleaning the contacts on the PowerLeap and also tried a Tualeron 1300 I managed to find on the PowerLeap and, with both Tualatin CPUs, I get the same bogus temp and voltage values .
I put back the Coppermine and CPU temp and voltage are normal in hardware monitor .

I put back the PowerLeap and Tualeron 1300, set the BIOS to ignore CPU temp and voltage and started running Prime95 .

At this point, I think both Tualatin CPUs are likely OK as is the P3B-F .

I believe the PowerLeap adapter is having issues or maybe dying .

I put the Tualatin 1400 512K back in my spares drawers and am going to see if I can get the flaky PowerLeap stable with the Tualeron . If the Tualeron dies, I shouldn't cry too much .

kitten.may.cry wrote on 2020-07-08, 17:41:

I'll just say this: oof!

Good luck, I've sold my Ms6905 MASTER, and really regret doing that, as it is easily moddable for Palpatine!

If worse comes to worse, I could probably just order a modded Tualatin 1400 512K from the dude in Korea and use it with my Asus Coppermine slotket . That could be an option for you too, along with a cheap Coppermine capable slotket .

EDIT: Just saw that prices have gone up significantly and that P3B-F is not listed as compatible . I will have to check the motherboard's VRM and decide if I want to take a chance .

EDIT2: I still find it strange that the PowerLeap seems to be stable now that hardware monitor reports an issue with voltage and temp.

What do you think has better performance on for a 100MHz FSB board, a Tualatin Celeron 1.4 or the P3-S @ 1050MHz? I'm not sure which one to buy. The P3 will be nice if I ever upgrade to a 133MHz board and maybe running it at 100MHz (+ undervolt?) will mean a lower TDP than the Celeron.

The 512K PIII-S chips are much harder on aging VRMs than Coppermine

I very much doubt it. Additional cache is nothing, compared to heavily reduced voltage.

Oetker wrote on 2020-07-08, 17:54:

What do you think has better performance on for a 100MHz FSB board, a Tualatin Celeron 1.4 or the P3-S @ 1050MHz? I'm not sure which one to buy. The P3 will be nice if I ever upgrade to a 133MHz board and maybe running it at 100MHz (+ undervolt?) will mean a lower TDP than the Celeron.

I already tested something similar. I had compared a P3-S @ 1050MHz and a Tualatin Celeron 1.3 in 3DMARK2001SE . The P3-S @ 1050MHz came out on top .
See this post Re: Tualatin Celeron vs Williamette Celeron and the rest of its thread .

My PowerLeap with Tualeron 1.3 was Prime95 stable for over 4 hours before I stopped the test manually, as I considered that to be enough to gauge stability .

I am now running 3DMARK2001SE in a loop . It has completed 163 single test iterations with no issue so far .

My very uneducated guess is that whatever was causing the instability with the PowerLeap adapter had to do with the voltage and temperature reporting being in the process of failing and, now that this specific component is dead, the adapter works fine . I can live with that .

EDIT: I am going to classify this one as solved and will keep my eyes open for another PowerLeap adapter at a sane price . Thanks to everyone for all the comments .

VRM on my Asus P/B-F 1.03 is a HIP6020ACB . Anybody ever try running of those with a modded slotket while using the board's VRM to output 1.5V ?

EDIT: YOLO! I ordered one of the modded CPUs from the dude in Korea and I will test it .

darry wrote on 2020-07-09, 00:07:

VRM on my Asus P/B-F 1.03 is a HIP6020ACB . Anybody ever try running of those with a modded slotket while using the board's VRM to output 1.5V ?

Mine works with a LinLin and a normal Coppermine Slotket, and with an upgradeware Slot-T. But the two mosfets behind the Slot get very hot, so I put heatsinks on those + a large fan on the whole area. No issues, even at 150 FSB. See also viewtopic.php?f=46&t=73362

flupke11 wrote on 2020-07-09, 04:54:

darry wrote on 2020-07-09, 00:07:

VRM on my Asus P/B-F 1.03 is a HIP6020ACB . Anybody ever try running of those with a modded slotket while using the board's VRM to output 1.5V ?

Mine works with a LinLin and a normal Coppermine Slotket, and with an upgradeware Slot-T. But the two mosfets behind the Slot get very hot, so I put heatsinks on those + a large fan on the whole area. No issues, even at 150 FSB.

Thanks for the info . I will be sure to consider some extra cooling options .

EDIT: Having another YOLO moment, I decided to put my "spare" 1.4GHz Tualatin 512K back into the PowerLeap and give it whirl at 133MHz FSB (AFAIK Fx5900XT should be fine with 89MHz AGP) . Currently running Prime95 overnight .

darry wrote on 2020-07-08, 22:55:

Oetker wrote on 2020-07-08, 17:54:

What do you think has better performance on for a 100MHz FSB board, a Tualatin Celeron 1.4 or the P3-S @ 1050MHz? I'm not sure which one to buy. The P3 will be nice if I ever upgrade to a 133MHz board and maybe running it at 100MHz (+ undervolt?) will mean a lower TDP than the Celeron.

I already tested something similar. I had compared a P3-S @ 1050MHz and a Tualatin Celeron 1.3 in 3DMARK2001SE . The P3-S @ 1050MHz came out on top .
See this post Re: Tualatin Celeron vs Williamette Celeron and the rest of its thread .

Thanks for taking the time to test this!

Ever the undecided experimenter, I interrupted the Prime95 run before bed and decided to loop 3DMARK2001SE. Unfortunately, sometime in the last few hours, it crashed out of the loop complaining about not being able to init 3D hardware . Windows did not crash and I was able reboot cleanly and set BIOS back to 1050MHz for CPU (and normal AGP speed) .

Now running 3DMARK2001SE again in a loop while I get some more sleep. If that works, I will run Prime95 again for a few hours.

If that works, I may experiment with 133MHz FSB again with a longer Prime95 run to be sure that 3DMARK2001SE crash was due to 89MHz AGP and not an unstable CPU or RAM (PC133 RAM was running at 133MHz at CL3).

EDIT: 3DMARK2001SE crashed out of the loop again at 1050MHz with a d3derr_notavailable in its log . I am using an FX5900XT on a BX board (driver 77.72) . That may be part of the issue or the video card is flaky .

EDIT2: I ran 4 hours of Prime95 to test the CPU at 1.4GHz and it passed . I also switched to 56.64 Nvidia drivers (I don't remember the reason I was not using those to begin with) and 3DMARK2001SE is looping fine so far (with AGP at 89MHz) .

EDIT3: I ran through 458 single tests of 3DMARK2001SE in a loop without issue before I interrupted it. I think it's stable enough .