OK, I've got my new system (Athlon64 3000+) running stock speeds.  I've got 2x512MB Corsair 3200 LL (2-3-2-6) which is all running in Auto Mode.  

Two benchmarks (3dMark03 and PCMark04) run fine.  The highest temp I get is 52 C after running the CPU on max for a while.  Nominal CPU temp is 45 C.  NB temp is max 37 C.

When I run Prime95.  It runs for 4-8 hours and then has an Application Fault.  This has happened 4 times now.  Please note this is a GPF, NOT an error reported in Prime95 (version 23.7.1.0).  Interestingly, the fault address was the same 3/4 times.  The run times varied but the CPU temp was always 50-52 C when the failure occurred.

Any suggestions?  Is it possible Prime95 has a conflict with some other SW on my system?

Historically I only ran into 2 issues:
1) Boot problems (RAID) rectified when I re-seated SATA cables
2) Video card issues when running 3dMark03... I had a cable interfering with my Video Card fan and I would slowly get more artifacts, snow, and then a crash.  

Since I moved the offending cable, everything has been rock stable Except for this.

Yeah, that's what I was thinking .. that it might be a sw conflict. I recall that I was ruuning one version of p95 fine on an old system (cusl2-based p3-700 @ 933mhz), and then i upgraded to a new version of p95 & started having problems.

i re-ran old versions > no problems. so it was definitely something associated with the code of the newer version.

Note that p95 works a specific *part* of your cpu .. the fpu. i feel that, if it were a HEAT prob, it usually errors within 5 minutes.

i suspect something is kicking in after this amount of time .. such as disk utilities, anti-virus .. etc.

You could try yanking the cover and directing a small fan on the cpu .. and run over-night. that would give you another data-point to evaluate.

but if this is your ONLY problem, after 4 hours, i wouldn't be sweating it. 50-52 is not bad.

Maybe you should try backing off on the memoery timings .. just to see what happens. I'm not sure how aggressively you have them set.

Anyway, congrats on the new system!

Good Call!  I went searching around for a different version just to see, and while I was running v.234, I was trolling through their Forum and look what I found!


Quote:

Try ftp://mersenne.org/gimps/p95v238.zip I've discovered a possible memory allocation bug. This bug is independent of motherboard, but very dependent on amount of memory to be allocated (and FFT size). http://www.mersenneforum.org/showthread.php?s=&threadid=1588

Yikes, I'm now running .238 and we'll know tomorrow...

BTW, I am very pleased with my new system  8) especially since it's been 3 years since my last upgrade.  I used to upgrade once a year, but Kids will do that to you...

I know.  >:(

Confirmed it's fixed.  It ran 22 hours without a glitch.  Interestingly, it now keeps the CPU utilization at 100% whereas before, it would drop on certain calculation sizes (I ran the blended torture test).  There obviously was a problem with the .237 version.

Good news is the CPU temp ranged from 50-52 throughout the test.  Of course it's -25 C (-13 F) outside...  :P