In the future, you can change the error reporting for each iteration -- it's in undoc.txt. (I found them to be very annoying, although that may have been the OCD because all the lines weren't the same length :P (Before you ask George, I've only gotten errors on two tests right after I'd changed hardware around or messed with clock settings. After each error-test, I've turned in many successful double checks, though the tests with errors haven't been double checked yet.))

There's also a v[URL="http://www.mersenneforum.org/showthread.php?t=16618"]27.4[/URL] out now

[QUOTE=Prime95;292833]Prime95 will continue to warn you every 10K iterations about that one roundoff error - this is normal.

What isn't normal is the one roundoff error you did get. Worse yet, it wasn't reproducible. It might have been a true hardware error.

At a minimum, I'd turn on round off checking for every iteration. Report back with the maximum roundoff you are seeing.

P.S. Looking at the source, that FFT size is supposed to be good up to 352M. Looks more like a true hardware error. Torture test time.[/QUOTE]
This is what I get so far:

[code][Mar 13 11:18] Iteration: 30690000 / 332216777 [9.23%]. Round off: 0.0507812500 to 0.0703125000. Per iteration time: 155.884 ms.
[Mar 13 11:18] Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
[Mar 13 11:18] Confidence in final result is fair.
[Mar 13 11:44] Iteration: 30700000 / 332216777 [9.24%]. Round off: 0.0507812500 to 0.0781250000. Per iteration time: 155.335 ms.
[Mar 13 11:44] Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
[Mar 13 11:44] Confidence in final result is fair.
[Mar 13 12:10] Iteration: 30710000 / 332216777 [9.24%]. Round off: 0.0507812500 to 0.0781250000. Per iteration time: 155.392 ms.
[Mar 13 12:10] Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
[Mar 13 12:10] Confidence in final result is fair.
[Mar 13 12:36] Iteration: 30720000 / 332216777 [9.24%]. Round off: 0.0507812500 to 0.0781250000. Per iteration time: 155.153 ms.
[Mar 13 12:36] Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
[Mar 13 12:36] Confidence in final result is fair.
[Mar 13 13:02] Iteration: 30730000 / 332216777 [9.24%]. Round off: 0.0507812500 to 0.0781250000. Per iteration time: 155.088 ms.
[Mar 13 13:02] Possible hardware errors have occurred during the test! 1 ROUNDOFF > 0.4.
[Mar 13 13:02] Confidence in final result is fair.
[Mar 13 13:28] Iteration: 30740000 / 332216777 [9.25%]. Round off: 0.0507812500 to 0.0781250000. Per iteration time: 155.312 ms.
[/code]

[QUOTE=bcp19;292900]This is what I get so far: Round off: 0.0507812500 to 0.0781250000.[/QUOTE]

It is extraordinarily unlikely the roundoff error went from 0.078 to 0.5 for one iteration. I conclude you had a hardware problem.

[QUOTE=Prime95;292907]It is extraordinarily unlikely the roundoff error went from 0.078 to 0.5 for one iteration. I conclude you had a hardware problem.[/QUOTE]

How much does the roundoff error normally vary during a run? I was noticing on the 26M exp I just completed, that when I first turned it on, it was showing .08-.17 and during the last part of the run it was up to .11-.25.

[QUOTE=bcp19;293387]How much does the roundoff error normally vary during a run? I was noticing on the 26M exp I just completed, that when I first turned it on, it was showing .08-.17 and during the last part of the run it was up to .11-.25.[/QUOTE]

Prime95 reports the minimum and maximum roundoff errors encountered since you last (re)started computation. The fact that the max error started at 0.17 but one one or more outlier iterations it was 0.25 is not at all unusual.