I believe that the two problems are different. Let me take them in the
order of posting:
From: Rick Pali <[EMAIL PROTECTED]>
Subject: [Prime] Round off error...but not.
Date: Thu, 07 Jul 2005 19:43:07 -0400
[Tue Jul 05 01:59:06 2005]
Iteration: 2XXXXXXX/3XXXXXXX, ERROR: ROUND OFF (0.40625) > 0.40
Continuing from last save file.
[Tue Jul 05 02:25:28 2005]
Disregard last error. Result is reproducible and thus not a hardware
problem.
For added safety, redoing iteration using a slower, more reliable method.
Continuing from last save file.
I've seen the above error three times in the last five weeks. It suggests I
shouldn't be concerned, but given it's happened three times, is complacency
the best course of (in)action?
Indeed it is Rick. This kind of behaviour is seen when testing exponents
that are near the FFT limit. You are probably testing an exponent
close to the P4 FFT limit of 34560000. As long as each "error" is
followed by a disregard last error message you should be fine.
On Thu, 7 Jul 2005, p k wrote:
I have had similar experience in the past.
I was playing UT2004 3D FPS at the time and it is memory intensive.
The May 20th entry looks like it started twice - the second round off error
was a lower iteration.
I believe it "restarts from the last save" so I dont believe any further
action is needed.
[Thu May 12 15:50:58 2005]
Iteration: 20732032/3334xxxx, ERROR: ROUND OFF (0.5) > 0.40
Continuing from last save file.
[Fri May 20 03:37:16 2005]
Iteration: 22520192/3334xxxx, ERROR: ROUND OFF (0.498046875) > 0.40
Continuing from last save file.
Iteration: 22519296/3334xxxx, ERROR: ROUND OFF (0.5) > 0.40
Continuing from last save file.
[Fri Jun 24 16:01:12 2005] <--- Result finished not prime
I am afraid not. Continuing from the last save file is not enough. This
just means that there was an error the first time round but not the
second time which likely points to a hardware failure or failure due to
drivers. It is known that some games operating with older video/audio
drivers cause the FPU state to be set incorrectly causing an error.
Alternatively, running UT is causing Prime95 to use some memory/cache
areas that are susceptible to error that are not otherwise used during
LL tests. Read the stress.txt and try to fix the system accordingly as
there 50% chance that your test result will be erroneous and thus
worthless. A quick examination of the BAD and LUCAS_V.TXT files show
that 44 of the results returning this error code 00000300 were good and
68 were bad. There are a further 240 results - including yours that
await verification.
Anurag
_______________________________________________
Prime mailing list
[email protected]
http://hogranch.com/mailman/listinfo/prime
