Apparently, the memory modules were to be blamed (they're PC-100, not PC-133).

Once I've set DRAM clock to 100 MHz, everything seems back to normal:

Or so I thought...

The next day, setting DRAM clock to 100 MHz has no effect, the system freezing even before POST.

Coppermine-T is slightly different from usual Coppermine. I've never know it before your post 😀
So a PCI debug card (POST card) would be handy when the system stopped booting.

Thank you for your response, Roman.

After some consideration, I can't but come to a conclusion there's indeed something wrong with the 133 MHz FSB and PC-133 memory modules which are marked as such but are not PC-133 capable.

So now I'm ordering a 100 MHz FSB CPU and some extra PC-133 DIMM modules (Kingston or Hynix), since those I have currently are nonames.

You could also try to run the cpu/ram at fsb 100 right now.
On page 1o in the manual you can find the setting for fsb 66 and 100 via jumper JP9 & JP10,
and on page 11 you can find jumper JP15 settings for automatic or manual selection.

PARKE wrote on 2021-05-24, 15:03:

You could also try to run the cpu/ram at fsb 100 right now.
On page 1o in the manual you can find the setting for fsb 66 and 100 via jumper JP9 & JP10,
and on page 11 you can find jumper JP15 settings for automatic or manual selection.

Thank you.

I've actually tried a bunch of other pre-Coppermine-T Coppermine-based CPUs (SL4P2, SL5DV), and all of them proved quite unstable at both 66, 100, and 133 MHz FSB clock.

So I ended up blaming the motherboard and replaced it with a VIA Apollo Pro133T-based Shuttle AV18 v3.1 (AV18v31) which turned out to be so much better than the ECS motherboard I originally had. The new motherboard accepted each and every Socket 370 CPU I had available, from a 500 MHz Celeron Mendocino to a 1400 MHz Pentium III-S.

Thanks, everyone.

Case solved.

I have a guest system built around the ECS P6VAP and it runs perfectly fine. I'm using a 1.13Ghz Pentium III CPU I had in a dual socket 370 build and I haven't encountered any stability or performance issues with it.

When I got the board it had 3 bloated - no - exploded capacitors (they looked kind of like half-popped popcorn) and would post but would lock up in bios or after booting (if I pressed F1 at the CMOS error message) just like you're describing. I replaced the four caps next to the CPU socket with panasonic low-esr and it's been running great for over a year now. Before the 1.13GHz coppermine I had a 933MHz P3 in it.

The rest of the system is composed of a Geforce 3 Ti500, a AWE64 value and 256MB of PQI PC133 SDRAM (1 stick).

Socket3 wrote on 2021-07-12, 18:45:

When I got the board it had 3 bloated - no - exploded capacitors (they looked kind of like half-popped popcorn) and would post but would lock up in bios or after booting (if I pressed F1 at the CMOS error message) just like you're describing. I replaced the four caps next to the CPU socket with panasonic low-esr and it's been running great for over a year now. Before the 1.13GHz coppermine I had a 933MHz P3 in it.

Socket3, thank you for your response.

My own ECS P6VAP-A+ motherboard is in almost "new" condition -- i. e., no blown capacitors, nothing (otherwise, I would have replaced them already -- Panasonic low-ESRs are good indeed). The problem I was describing above is indeed purely CPU-specific: Mendocinos work fine, Coppermines don't.

Still, the new Shuttle replacement is one of the best ISA-capable P-III motherboards -- it even has the built-in SoundBlaster chip which can be turned on/off from within BIOS. This is absolutely perfect in case you need not only a DOS-compatible sound card (enabled with TSRs under DOS), but a full OS-agnostic SB-compatible one.

I just came across some capacitors last week, Gloria brand, that had straight sides, flat top, and had blown out at the bottom. It was very hard to see when they were installed, but if you inspect around bases very carefully and see any sign of anything that might be capacitor juice then change them. Funnily enough those were on a Shuttle.