Prime stability weirdness

When I originally o/c'd my rig, I got my 2500 up to 2.3 GHz, 1.85v and Prime95 ran fine for a couple of hours - had done this several times during the first week of owning it.

However, I decided to run Prime95 again today and it errored out in less than a minute! *Nothing* has changed.

Why would the CPU go from stable to erroring in Prime when nothing has changed? Oddly enough, everything else I run is fine. 3Dmark, games, etc. Prime95 just seems to hint that the system is on the verge of crapping out...

When you run Prime95, it doesn't always fail right away if a system is unstable. A test of a couple of hours is good enough to get an idea if your system is stable, but to ensure true stability, you have to Prime for at least 12 hours or more at priority 10. I saw a thread about testing at priority 10, but now I can't find it.

It is possible that your system wasn't stable to begin with and is now showing that.

There's more to what he has observed than a simple instability, though it is not out of the question.

I have spent a lot more time than I would have wanted to the past month or so, running Prime95 Torture Priority 10 and though I'm not sure whether it is an issue with the particular revision of P95 that I'm using or just what, I have been seeing some odd behavior myself, only noticiable from using it so much recently.

This is hard to put into anything other than anecdotal rambling, but I have, on many occasions, gotten my rig into a state of being where p95 is very unstable, often running only 2 minutes, warm booted and then 9 minutes, warm booted then 2.5 hours, warm booted then 38 minutes, in other words all over the place with the exact same setup and room temps very stable. My best luck at breaking this "funk" is to do a full shutdown, cold boot and once Win XP finishes all those little housekeeping chores only evidenced by the drive activity LED flickering, after reaching the desktop (appx 30-45 more seconds), only then do I launch Prime95 Torture Priority 10.

My memory is not so bad yet, but I don't recall this kind of erratic behavior in past revisions of Prime95, none of which I have saved anywhere. Now, whether this is some quirk in the current revision of P95, or just that the current revision has reached a state of perfection so as to be more demanding of a perfectly stable system, I have no idea. Just reporting my observations. I don't think it is the fault of this NF7-S, since I saw a similar pattern with my A7N8X. They are however, both NF2 boards. It would be interesting to see if VIA chipset board owners are experiencing this. Note that it does not occur at conservative settings, only when pushing the envelope at or near the upper limit of my particular hardware compliment.

Anyone else want to weigh in with their particular experiences since the current revision of P95 came out?

Hoot

Actually, I've been experiencing this lately too with P95 since version 23.7.1

I've tested my system using P95 priority 10, and it passed 24hour test using 1.75v at 200x11, but after I switched out my sunon 120mm fan for a pabst 120mm I tested prime again, and prime would fail after 4hours. Increase voltage to 1.775 and it'll stablize again. Is this my chip or is it P95

p.s. I had the sunon on a rheobus so temps with Pabst is nearly identical as with the old fan and a LOT quieter!

edit: oops i meant i had the Sunon on a rheo and now the pabst is running full steam so the temps are nearly identical as before the fan switch. Guess if my temps stay the same its fine pushing the xtra vcore through the chip right?

Well, having changed fans may have incresed the core temperature just enough to put it slightly "over the stability line". Axial fans of different make, rarely deliver the same effective airflow. Heck it even varies a little within the same make and model.

For any given Vcore, Multiplier and FSB setting, there will be a finite core temperature threshold, below which those settings are stable and above which they aren't. When you're right on that imaginary line, even room temperature fluctuations can make the difference between success and failure.

Hence the age-old dilemna between loud (obnoxious) fan noise tolerance and how high you want to be able to overclock using air cooled solutions. That's the main reason I went to water cooling. Quiet, yet effective.

Hoot

once i get my 8kra+2 up and running, i'll see if it's nF2-centric but i doubt it...i've had the same problem on my g-byte..but here's something for you..on my KT6 Delta-LSR, i was able to run an unlocked xp2000 @ 200x10 on 1.7V P95 (priority 1) stable for 12 hrs...nothing doing on my g-byte 7n400L1 nF2 board...200x11 is taking 1.85 on that right now (wouldn't do 200x10)...the xp2000 is a DUT3C 0337 stepping...btw, hoot..what's the password for P95?

9876

It's in the readme that comes with Prime95.

Hoot

thanks..i couldn't remember where i saw it and i kept looking thru threads...

I have observed somthing similar with my Barton 2500+ @ 2.3

It was running fine for a few months (including summer) - tested prime95 for 24 hours (unchanged priority 1) and then 4-5 months later the explorer started crashing from time to time.

I tested prime again and boom - errors. Reduced the FSB with 1 and it was ok for 2-3 weeks. Now my FSB is down to 167 from 170
13.5x and vcore 1.8

Now it is very stable again but at a lower frequency.

Another strange thing - MBM cpu usage and some other programs for CPU usage don't work - report 0%. Only task manger works which is wierd but may be OS/software related.

I am just wondering if some parts of the CPU got burned out - it was running UD 24/7/365.

It is water cooled but themps are a little high - 50C

The mb is via based - abit kd7 with NB kt-400

Prime95 likes to act up with my overclock as well - I am using the newest verson of Prime95.

It is 24 hour priority ten stable, but only on a fresh boot (turn on computer, start Prime95). It is also stable for 24 hours mixed with PCMark, HotCPUTesterPro, and N64 emulation with no FPS cap at an¥ time.

However, if, for example, I run Prime after playing Deus Ex for an hour it will give errors within an hour or two. 3Dmark can run looped at any time with no issues, same with PCMark, and Sandra, and HotCPUTesterPro.