Please help, rounding error?
 

Ok, I decided a couple days ago to torture test my computer by running Prime95 overnight. At first I didn't know how to run multiple instances (I have a HT CPU), so I ran Prime95 and a guitar tuner program I have on my computer that I just happen to know uses 50% of my cpu. That put me at 100% but after a little over an hour I got a hardware failure error. Then later I just decided to run one instance of Prime95, it ran for a little over 3 hours before giving me a similar error. At the time I thought it might be either my PSU or memory, because I'm running a shuttle box and have had problems in the past with too little power. So yesterday I got my new, more powerful, PSU and swapped out the ram for different sticks. I then ran two instances of Prime95 and went to sleep. I woke up about 6 hours later and discovered that one instance of prime95 was still running fine, but the other had quit after about 5 hours giving me the same error. I have temp sensors on my cpu, my northbridge, and my psu, and I know that none of them are overheating. I attempted to keep my room pretty cold last night to make sure nothing else did as well. The error always has to do with rounding. I'm going to paste my full results.txt at the bottom of this post. And finally, I've been having an ongoing discussion of this problem on another forum, if anyone thinks they could help me and wants more details, you can get them [URL=http://forums.sudhian.com/messageview.cfm?catid=43&threadid=50156&STARTPAGE=2]here[/URL]. Ok, here is my results.txt of the instance that crashed. Oh yeah, in case it matters, I'm using the large fft's test.

[Tue Jan 27 19:38:49 2004]
Self-test 1024K passed!
[Tue Jan 27 19:57:13 2004]
Self-test 896K passed!
[Tue Jan 27 20:15:18 2004]
Self-test 768K passed!
[Tue Jan 27 20:32:49 2004]
FATAL ERROR: Rounding was 0.4990234375, expected less than 0.4
Hardware failure detected, consult stress.txt file.
[Wed Jan 28 00:06:45 2004]
Self-test 1024K passed!
[Wed Jan 28 00:22:28 2004]
Self-test 896K passed!
[Wed Jan 28 00:38:06 2004]
Self-test 768K passed!
[Wed Jan 28 00:53:46 2004]
Self-test 640K passed!
[Wed Jan 28 01:09:27 2004]
Self-test 512K passed!
[Wed Jan 28 01:25:20 2004]
Self-test 448K passed!
[Wed Jan 28 01:41:38 2004]
Self-test 384K passed!
[Wed Jan 28 01:57:27 2004]
Self-test 320K passed!
[Wed Jan 28 02:13:05 2004]
Self-test 256K passed!
[Wed Jan 28 02:28:36 2004]
Self-test 224K passed!
[Wed Jan 28 02:43:52 2004]
Self-test 192K passed!
[Wed Jan 28 02:59:20 2004]
Self-test 160K passed!
[Wed Jan 28 03:14:41 2004]
Self-test 128K passed!
[Wed Jan 28 03:31:20 2004]
Self-test 1024K passed!
[Wed Jan 28 03:38:17 2004]
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
[Thu Jan 29 19:47:41 2004]
Self-test 1024K passed!
[Thu Jan 29 20:05:26 2004]
Self-test 896K passed!
[Thu Jan 29 20:22:45 2004]
Self-test 768K passed!
[Thu Jan 29 20:39:54 2004]
Self-test 640K passed!
[Thu Jan 29 20:56:53 2004]
Self-test 512K passed!
[Thu Jan 29 21:14:19 2004]
Self-test 448K passed!
[Thu Jan 29 21:32:06 2004]
Self-test 384K passed!
[Thu Jan 29 22:39:35 2004]
Self-test 1024K passed!
[Thu Jan 29 22:57:12 2004]
Self-test 896K passed!
[Thu Jan 29 23:14:22 2004]
Self-test 768K passed!
[Thu Jan 29 23:31:31 2004]
Self-test 640K passed!
[Thu Jan 29 23:48:32 2004]
Self-test 512K passed!
[Fri Jan 30 00:06:03 2004]
Self-test 448K passed!
[Fri Jan 30 00:23:58 2004]
Self-test 384K passed!
[Fri Jan 30 00:40:58 2004]
Self-test 320K passed!
[Fri Jan 30 00:57:36 2004]
Self-test 256K passed!
[Fri Jan 30 01:14:08 2004]
Self-test 224K passed!
[Fri Jan 30 01:30:14 2004]
Self-test 192K passed!
[Fri Jan 30 01:46:24 2004]
Self-test 160K passed!
[Fri Jan 30 02:02:28 2004]
Self-test 128K passed!
[Fri Jan 30 02:19:52 2004]
Self-test 1024K passed!
[Fri Jan 30 02:37:07 2004]
Self-test 896K passed!
[Fri Jan 30 02:54:00 2004]
Self-test 768K passed!
[Fri Jan 30 03:10:52 2004]
Self-test 640K passed!
[Fri Jan 30 03:22:46 2004]
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.

This has been discussed a bit recently. Are you using the latest version, 23.8? Search this forum for "23.8" and you will find some threads with really useful suggestions.

[QUOTE=markhl]This has been discussed a bit recently. Are you using the latest version, 23.8? Search this forum for "23.8" and you will find some threads with really useful suggestions.[/QUOTE]

Looks like I was using 23.7. Just downloaded 23.8 from a thread here in the forums. Is there a chance that this will fix my problems?

so much for that idea... 23.8 gave me an error 28 minutes into the test.

Try reducing the overclock and/or using safer setting for your CAS/RAS/etc latencies.

[QUOTE=Prime95]Try reducing the overclock and/or using safer setting for your CAS/RAS/etc latencies.[/QUOTE]

nothing is being overclocked, and I have tried different ram. Will try different settings tonight.

You can also try underclocking! It is not unheard of for CPUs to fail the torture test at stock speeds.

how would I go about doing that? lowering the fsb?

there shouldn't be any reason why an earlier version of Prime95 might work and a newer one wouldn't is there?

underclocked my 3.0Ghz processor to 2.0Ghz and am running Prime95 now. Will let you know tomorrow morning what happens.

10 hours later, Prime95 is still running on my underclocked machine. Guess it's my processor.