Should we continue to crunch after an error occurs??
 

I noticed that after 5 days of work on a 33M exponent, that I had 2 errors within 3 hours of each other. The system had never reported errors before, on this exponent or any others.

Should I continue with the test or throw it out and start again?

Tell us what the error message was. ILLEGAL SUMOUT is benign. Some ROUND OFF > 0.40 errors are harmless if you are near an FFT limit and it is reproducible.

If not one of these cases, I'd toss the 5 days of work.

I guess it is a bit longer than 5 days. I started the exponent on Jan. 25th.

This is a P4 doing the crunching. Here is the statement from the results file.

[quote][Sun Feb 09 05:36:18 2003]
Iteration: 13894912/33322867, ERROR: ROUND OFF (0.5) >0.40
Continuing from last save file.
[Sun Feb 09 06:20:16 2003]
Iteration: 13905664/33322867, ERROR: ROUND OFF (0.5) >0.40
Continuing from last save file. [/quote]

I recently got an ILLEGAL SUMOUT error after my computer crashed. Does it make sense that I would get an ILLEGAL SUMOUT error right after my computer crashed then re-booted?

[quote="outlnder"]
Iteration: 13894912/33322867, ERROR: ROUND OFF (0.5) >0.40
Continuing from last save file.
[Sun Feb 09 06:20:16 2003]
Iteration: 13905664/33322867, ERROR: ROUND OFF (0.5) >0.40
Continuing from last save file. [/quote]

A serious error. I'd say the chances at best 50% that your result is OK. Prime95 does not catch every error. It's your call as to weather you should finish it off - you are close to 50% complete.

By the way, have you found the cause of the error? If not, you may as well finish this exponent until you can figure out what is causing the problem.

[quote="eepiccolo"]I recently got an ILLEGAL SUMOUT error after my computer crashed. Does it make sense that I would get an ILLEGAL SUMOUT error right after my computer crashed then re-booted?[/quote]

You could have a driver that isn't saving the FPU state properly during the reboot. ILLEGAL SUMOUTs are fairly innocuous, so I wouldn't worry about it.

[quote="eepiccolo"]I recently got an ILLEGAL SUMOUT error after my computer crashed. Does it make sense that I would get an ILLEGAL SUMOUT error right after my computer crashed then re-booted?[/quote]
Yeah -- when a graphics driver used to lock up W98 hard, I got those right after reboot. Never caused a bad L-L residue though.

[quote]By the way, have you found the cause of the error? If not, you may as well finish this exponent until you can figure out what is causing the problem.[/quote]

This is a little difficult to find. This machine has been running well over a month and has never given an error before. And it has been running continuously since the 2 errors and has not given another.

All I can figure is it got bad power for a little while.

If anyone has any theories, I would be glad to entertain them.

Other errors on startup
 

I've noticed that the past 2 or 3 times I rebooted my current system (AthlonXP, WindowsXP), I got a couple of SUM(INPUTS) != SUM(OUTPUTS) within the first few iterations, and then nothing thereafter. Is this likely to be a benign problem, or should I toss my work and start over?

Thanks for the help.

I just experienced a blackout and a machine went down. In rebooting I've stuffed the overclock and started receiving errors, so I restarted the exponent.

It was 93% complete :(