Nothing obvious I can think of. The multiplier must be set to 2.5 instead of 3 or the FSB has been turned down to 55 from 66. I've turned the cache on and off on several boards/cpus and never had the clock speed change.

TandySensation, thanks. I was editing while you were posting.

Zark. It's a heat issue.

When the unit warms up enough, the POST shows 166 on the next boot. Let it cool a few minutes -- like the time it takes to haul the box out, up, & open to short the CMOS -- and it goes back to 200.

Jumpers are correct for voltage, clock, and multiplier. CPU fan is working, and it's got a proper heatsink. Nothing is obviously wrong with the board like a spoiled cap, or filth that needs removing.

Not sure what next. Maybe dig through storage to see if I've got a similar CPU to try.

Seems it was the chip. I can't imagine why though. Ever heard of a chip being unstable about holding its mhz?

Doesn't look fake. It's a SL26Q.
http://www.cpu-world.com/CPUs/Pentium/Intel-P … BP80503200.html

I'm having no problems with a P133 clocked at 166. Have run Phil's VGA Benchmark set five times, including L1 OFF and L1+L2 OFF.

--------
today's bonus trivia

This P133 is an SY022. Word is(was?) many SY022 have disabled BF1, which is multipliers 2.5 and 3.0. A theory is the disabled chips may be the ones imprinted "i133" on the back, whereas "iPP" are good. This one's an "iPP", and is running 2.5x66.6. It wouldn't boot with 3.0x66.6.

Haven't played with further yet. The 1/4" high copper sink is completely cool around the edges. It's not even a fraction warmer than the surroundings. Just using a generic period fan on it.

[EDIT: added image.]

Perhaps the trouble with the 200mmx it is something wrong with the 3.0x multiplier, but I can't uncover anything definitive. All I can find with the P133 is 3.0x doesn't work as well as 2.5x.

2.5x 66.6 - shows 166MHz POST (167.0MHz CPU Utility)

2.5x 75 - won't boot at default 3.2v, or at next & final step 3.52v.

3.0x 55 - 166 (165.0) - benchmarks 10% slower than 2.5x66.6, 20% slower for L1 OFF.

3.0x 60 - 188 (180.0) - unstable; requires 3.52v. benchmarks slightly slower than 2.5x66.6 despite listing as faster MHz.

3.0x 66.6 - won't boot at 3.2v or 3.52v.

pewpewpew wrote:

today's bonus trivia

This P133 is an SY022. Word is(was?) many SY022 have disabled BF1, which is multipliers 2.5 and 3.0. A theory is the disabled chips may be the ones imprinted "i133" on the back, whereas "iPP" are good. This one's an "iPP", and is running 2.5x66.6. It wouldn't boot with 3.0x66.6.

Haven't played with further yet. The 1/4" high copper sink is completely cool around the edges. It's not even a fraction warmer than the surroundings. Just using a generic period fan on it.

Interesting, I've never seen that stuff about the "i133" and "iPP" markings before. I have an unlocked SY022 but it's installed in a board somewhere so I can't compare how it's marked. I just know it was a tray CPU because it came from a Packard Bell.
Based on the dates of the machine my SY022 came from, I think that stepping is very late production for 133MHz. As such, they probably all run cool at that speed.
According to my old notes, 166/66 (2.5x) was stable on the original board.
On some super-7 board, it was stable in memtest86 at 180/90FSB still at default voltage.
200/66 (3x multiplier) was memtest86 stable on an Asus SP97-V but only after raising Vcore to 3.5v.

As far as the multiplier issue on the 200mmx chip, maybe it's a corroded pin on the CPU? This document:
http://download.intel.com/design/pentium/data … ts/24199710.pdf
says that the multiplier is controlled by 2 pins, labeled BF0 and BF1. If BF0 flips, it would make the difference between 2.5x and 3x. Perhaps the CPU isn't getting a consistent read on that pin at startup.
There's a pinout diagram on section 2.0 of that file, BF0 must be on there somewhere, I'll let you find it. 😀
That PDF is apparently for P54C, not P55C, but I don't think that matters. The multiplier pins would have to be compatible.

Since it only happens with one CPU, the issue is *probably* on the CPU side. I say probably because maybe, just maybe, this CPU just happens to be more prone to provoking an electrical fault that exists on the board (because of more heat, or whatever). I guess if you want to rule that out, you could try the same CPU on a different board to see if it behaves any better. You could also just try it again on the current board - the simple act of reseating it might be enough to fix it.

The pins all look alike to me. Swabbed the area with isopropyl alcohol anyway. Then I stupidly ran the first two sets of benchmarks at 3.2v before noticing... The recommended is 2.8 with max 2.9.

Ran two Phil VGA sets on each; normal and L1 OFF. (actually also L1+L2 OFF on first set to really warm things up.)

3.0x 66.6 - 200 (200.4)
3.5x 66.6 - 233 (233.8 )
3.5x 68.5 - 240 (239.8 )
3.5x 75 - won't boot
3.5x 75 - increase one step to 2.9v - still won't boot
3.0x 75 - 225 (227.6) - almost... Fine except Quake on the L1 OFF test. Crashes on load. Had to increase to 2.9v to run that.

Oddity: the first two had indentical results for the L1 OFF tests. Um, why?
3.0x 66.6 - 200 (200.4)
3.5x 66.6 - 233 (233.8 )

pewpewpew wrote:

All I can find with the P133 is 3.0x doesn't work as well as 2.5x.

This is empty-headed nonsense, of course. Those differences were due to bus speed.

shamino wrote:

you could try the same CPU on a different board to see if it behaves any better.

It's so odd having only the one pentium now. They were common as weeds just a little while ago. I suppose it'd feel less strange if I hadn't kept any. It's this unit's bald existence that makes that me think my memory of its brethren filling the back closet is still correct.