Something is corrupting the BSP file it's loading immediately. Different CPUs shouldn't cause this. Perhaps it's your memory timings?

I can't configure any memory timings so just tried different modules (several)... but swapping the CPU out fixes it immediately...

Have you tried running Memtest86? It's probably not a bad stick of RAM but the problem might show up as such. I believe Memtest86 will also test the L1 and L2 caches on the CPU which could also be where the problem arises.

Might just be unfortunate that the CPU got a minor static electricity shock at some point in its life from someone who handled it improperly and now a couple of L2 cache cells are bad as a result.

EDIT: L1 cache, thanks dionb. Spending too much time around PII chips 😀

L1 you mean... a Pentium doesn't have L2 cache on-die, and the cache on the motherboard is fine as other CPUs show.

InbetweenDays wrote:

and finally it worked with a different CPU.

The CPU could be just faulty, or it can no longer comply with its timings due to age (switching time of semiconductors increase with power-on hours).

I had a similar case not long ago. Had built a perfectly working P1, no problems whatsoever. Well, apart from one: copying files from USB would always corrupt data. Tried everything incl. a different USB card. What fixed it in the end? Replacing the CPU...

derSammler wrote:

The CPU could be just faulty, or it can no longer comply with its timings due to age (switching time of semiconductors increase with power-on hours).

Interesting, thanks.

derSammler wrote:

I had a similar case not long ago. Had built a perfectly working P1, no problems whatsoever. Well, apart from one: copying files from USB would always corrupt data. Tried everything inkl. a different USB card. What fixed it in the end? Replacing the CPU...

That's troubleshooting for you... 🙁

Malvineous wrote:

Have you tried running Memtest86? It's probably not a bad stick of RAM but the problem might show up as such. I believe Memtest86 will also test the L1 and L2 caches on the CPU which could also be where the problem arises.

Thanks, I didn't think of doing that. Is that Memtest86 or Memtest86+ (or both)?

I tried Memtest86+ 5.01 and... well, see attached. I stopped the test after taking this pic because I decided that over 166000 errors was probably enough to prove a point. 😀

The errors all started in test #7, and only with that CPU with a 66MHz clock rate. When I bumped it down to the 60MHz clock there were no errors. I tried the 6x86 again but Memtest wouldn't work with it, and since I couldn't be bothered taking my P200 machine apart again, I just put in a P75 overclocked (66MHz x 1.5)... no errors in test #7 (didn't test further).

Interesting. Can you configure the tests so that only the L1 cache tests are run? Would be interesting to know whether it's really the cache at fault or whether Memtest thinks the cache is good.

In your screenshot the errors are at the 30MB mark, do they start earlier than this? Just thinking that as a curiosity exercise, if the errors are all above the 16MB mark, then if you run it with only 16MB total it would be interesting to see whether you still get any errors.

Malvineous wrote:

Interesting. Can you configure the tests so that only the L1 cache tests are run?

No idea. It doesn't say anything specifically about the L1 cache.

Malvineous wrote:

In your screenshot the errors are at the 30MB mark, do they start earlier than this?

Yes, the errors start immediately (too many lines scroll past but I could see they started at 0.something)

I've just done a heap more testing, and the end result is this: With one stick of RAM there's no problem, but as soon as I put two sticks in, I start getting Memtest errors with test #7 -- regardless of the total amount of memory. (I tried 16MB, 8MB and 4MB modules).
But ONLY with the P166 CPU running with the 66MHz clock rate. I tested it with the P200 and Memtest completed with no errors (well, I just did one pass -- I've left it running now).

Interestingly, Quake was working fine today, with nothing having changed except the ambient temperature (so I tried letting the CPU get hot but that didn't make a difference).

If it's the CPU then I'd expect the problem to follow it to another machine... but it didn't. It's only that CPU in that machine with a 66MHz clock rate and more than 1 stick of RAM.

Weird.

I take it you pressed "c" and looked through the menus? It's been a while so I can't recall if there are specific cache testing options.

Apparently some Athlons used to fail on only a couple of the tests (5 and 7 I think) and cause was default memory timings that were too aggressive. Perhaps if you can adjust memory timing in the BIOS to be a bit more conservative that could be a possible workaround? It looks like it's just some part of the system that's reached its operating limits just at the edge of the spec. What else is different between the P166 and P200? Core voltage, cache size, etc. Perhaps that might give a hint as to what could be at fault, if everything else in the system is otherwise the same between success and failure.

It could be a board/chip combo issue.... unlikely but an option.
I had this with 486 and 586 hardware. Some cpus do not like to be driven by too high a fsb on the "wrong board" - this can even be the case regarding frequencies where it is taken as granted they work with.

Does the CPU work at 75 x 2?

Malvineous wrote:

I take it you pressed "c" and looked through the menus? It's been a while so I can't recall if there are specific cache testing options.

Yep... I didn't see any cache options.

Malvineous wrote:

Perhaps if you can adjust memory timing in the BIOS to be a bit more conservative that could be a possible workaround?

There's nothing I can change in the BIOS unfortunately... no settings for memory timing, only parity/ECC.

Malvineous wrote:

What else is different between the P166 and P200? Core voltage, cache size, etc.

Very little IIRC, but I'd have to check the actual codes on them and look it up.

amadeus777999 wrote:

Does the CPU work at 75 x 2?

It works at 50x3, 60x2.5 and 60x3. Problems start with the 66MHz clock rate, but that's as high as the board supports. (50/60/66 only).

What's so hard to accept that the CPU is just done and can't work at 66 MHz FSB any longer? Replace it or let it run at 60x3 until it dies completely.

derSammler wrote:

What's so hard to accept that the CPU is just done and can't work at 66 MHz FSB any longer? Replace it or let it run at 60x3 until it dies completely.

Haha... nothing... except that it works fine in 2 other computers I tried.
If it had shown the same problem in another machine I'd be more satisfied... 😀

I would have called it solved when you didnt experience any problems at 180mhz. 😀