Changing behavior of Over-Clock Settings?

[BonzaiDuck]

I would appreciate any useful observations on this matter.

Without increasing temperatures significantly, I was able to push my 3.0C Pentium4 to 3.70. I had set VCore voltage to 1.525V -- noting that it worked reliably, was lower than P4C warranty-spec, and seemed to operate with what had been the CPU load temperature at a slightly less ambitious over-clock setting. And I ran PRIME95 with this voltage setting for some eight to twelve hours without any error whatsoever.

I wanted to see if I could trim the VCore voltage just a tad more -- to 1.5125V. This generated some rounding error under PRIME95, so I rebooted, entered BIOS and reset the the VCore to 1.525V.

But now, the same setting that proved rock-stable under PRIME95 before shows an occasional rounding error under PRIME95. Certainly, it was fortunate that I chose to re-test the setting at all!!

Anybody have the slightest clue as to what is happening here? I don't get it . . .
1.5375V seems to be just fine, but I am puzzled by the inconsistency I've experienced here . . .

ASUS P4P800 standard BIOS 1019.005
Intel P4 3.0C @ 3.70
1GB OCZ EL "Gold" DDR500
ASUS v9980 nVidia FX5950 Ultra
dual Hitachi 7K250 in RAID 0 -- ICH5 controller
OCZ PowerStream 520 PSU

 

[Navid]

Room temperature may have been different when you ran the two tests.
Dust build-up may have been different inside the case on the CPU heat sink.
You may have updated background programs so that they were different while you ran the two tests resulting in less loading of the CPU in one case.

Did you ever run memtest to make sure that the memory is out of the picture (as an instability factor)?

 

[BonzaiDuck]

MEMTEST86 -- Certainly. I had done this for all the over-clock settings I tried, including 3.808 Ghz with external frequency 254. I wouldn't have even tried booting to XP if I'd had MEMTEST86 errors, and usually ran the program for two to four hours if the first six iterations proved error-free.

I think -- no, I'm sure -- that I DID update at least one program: it was SpeedFan. But it seemed that I was monitoring with ASUS Probe -- not SpeedFan -- when I noticed the change in stability.

Anyway, I find this to be a curiosity. How the system could be rock-solid-stable one day, and with the same settings, less-than-perfect the next.

By the way -- here's some information I pass along from Alfredo Comparetti -- the author of SpeedFan. I was having problems with SpeedFan grabbing resources and showing "100% CPU Usage" on one of the Task-Manager Usage-History graphs. That is, SpeedFan would be running, and I would load up the CPU with PRIME95. No problem would be evident until I stopped the PRIME95 iterations, and I would wonder why the CPU temperature was not declining.

I have an nVidia FX5950 graphics card, and Comparetti says that this is the source of the problem. He suggested running SpeedFan with the command-line parameter "/NONVIDIAI2C".

The problem has since gone away.

 

[xtknight]

I have experienced some suspicious behavior with that motherboard you have (P4P800, I have Deluxe). I set my AI Overclock to 30% originally and it would work. I turned the overclock off and put it back to 30% again and the thing just refused to POST. wtf? so I put it at overclock 10 then 20 then finally 30 and it POSTed again. I don't know about these motherboards..weird weird.

 

[o1die]

Overclocking is a real crapshoot. I flashed my bios last night, and the only change was a 1 point rise in the agp/pci lock settings. My system blue screened during windows load, so I reflashed the bios using the old version, and after some false starts, got the agp/pci lock back to 66/33, and the overclock worked fine again. Sometimes, I think it's better just to spend the extra money for a faster cpu. It took me 5 tries to find this northwood, and it was the cheapest (a Fry's combo deal).

 

[dug777]

I've noticed that with my old xp2000+ t-bred @ 2600+, it used 2 be rock-solid at 1.625v, then i had to hit 1.65v, and now 1.675v, & even @ that now i'm not prime stable, but since its just a games/dvd rig i cant b bothered

Running it on an A7V8X & the worlds worst PSU