P-1 bad factor, mprime
 

I've searched the Wiki and Forum and found references in the forum, but have not found an exact match for this situation. My secondary machine, running Ubuntu 11.10 64 bit and mprime in the current version delivered the following:[INDENT][CODE][Worker #2 Dec 7 14:01] Stage 1 GCD complete. Time: 196.261 sec.
[Worker #2 Dec 7 14:01] P-1 found a factor in stage #1, B1=490000.
[Worker #2 Dec 7 14:01] ERROR: Bad factor for M50575351 found: 3
[Worker #2 Dec 7 14:01] Restarting P-1 from scratch.[/CODE][/INDENT]Worker #2 then restarted the exponent from the beginning of Stage 1.

Does this signify a hardware problem, or something else? If "something else" what does it signify?

Current version meaning current stable, i.e. 26.6 (and not 27.1)? Can't help, but I'm sure that somebody else will ask that.

Definitely 26.6. I'm not adverse to beta testing, but I wouldn't do it in a "production" setting.....only on the side. Good question, though. I should have been specific.

2nd edit: At last check it had not affected the reliability rating of that CPU.

[offtopic]Did you get snow from that last storm that went through? We only got rain :P:P[/offtopic]

[QUOTE=kladner;281460]Definitely 26.6. I'm not adverse to beta testing, but I wouldn't do it in a "production" setting.....only on the side. Good question, though. I should have been specific.

2nd edit: At last check it had not affected the reliability rating of that CPU.[/QUOTE]

Did you check your prime.log for errors?

Luigi

[QUOTE=ET_;281490]Did you check your prime.log for errors?

Luigi[/QUOTE]

I had not. Thanks for suggesting it, Luigi. Unfortunately, (or perhaps fortunately) there is no reference in prime.log. The only thing there for yesterday is an "Updating computer information on server" line.

I had looked at results.txt. It has the second and third lines which are in the box above.

The final stage of the P-1 calculation is to compute GCD(X,N) where N is the number to be factored, and X is the result of the previous stages. The only way that this could result in a non-factor would be if the GCD computation itself were corrupted, or if N was corrupt prior to being passed to the GCD. A corrupt X alone could not produce a non-factor.

More information/ errors
 

I saw more indications of trouble this morning, and went through results.txt more carefully and found additional error reports. Prime.log for this recent range of days shows only completions, results, and server communications. For now, I have put this machine on Blend Torture Test until I can resolve this.

Extracts of results.txt are below:

[CODE][Worker #2 Dec 9 09:49] M50575351 stage 2 is 61.34% complete. Time: 852.900 sec.
[Worker #1 Dec 9 09:52] M50573119 stage 1 is 91.25% complete. Time: 664.695 sec.
[Worker #2 Dec 9 10:03] M50575351 stage 2 is 62.24% complete. Time: 855.806 sec.
[Worker #1 Dec 9 10:03] M50573119 stage 1 is 91.95% complete. Time: 661.293 sec.
[Worker #1 Dec 9 10:15] M50573119 stage 1 is 92.66% complete. Time: 661.364 sec.
[Worker #2 Dec 9 10:18] M50575351 stage 2 is 63.14% complete. Time: 858.422 sec.
[Worker #1 Dec 9 10:26] M50573119 stage 1 is 93.37% complete. Time: 662.271 sec.

[MENUS snipped]
Your choice: [Comm thread Dec 9 10:29] Updating computer information on the server
[Comm thread Dec 9 10:29] Done communicating with server.
4


[MENUS snipped]
Your choice: [Worker #1 Dec 9 10:29] Worker starting
[Worker #1 Dec 9 10:29] Setting affinity to run worker on logical CPU #1
[Worker #2 Dec 9 10:29] Waiting 5 seconds to stagger worker starts.
[Worker #1 Dec 9 10:29] Optimal P-1 factoring of M50573119 using up to 2500MB of memory.
[Worker #1 Dec 9 10:29] Assuming no factors below 2^72 and 2 primality tests saved if a factor is found.
[Worker #1 Dec 9 10:29] Optimal bounds are B1=490000, B2=9677500
[Worker #1 Dec 9 10:29] Chance of finding a factor is an estimated 3.89%
[Worker #1 Dec 9 10:29] Using AMD K8 type-3 FFT length 2688K, Pass1=896, Pass2=3K
[Worker #1 Dec 9 10:29] M50573119 stage 1 is 93.56% complete.
^Y[Worker #2 Dec 9 10:29] Worker starting
[Worker #2 Dec 9 10:29] Setting affinity to run worker on logical CPU #2
[Worker #2 Dec 9 10:29] Optimal P-1 factoring of M50575351 using up to 2500MB of memory.
[Worker #2 Dec 9 10:29] Assuming no factors below 2^72 and 2 primality tests saved if a factor is found.
[Worker #2 Dec 9 10:29] Optimal bounds are B1=490000, B2=9677500
[Worker #2 Dec 9 10:29] Chance of finding a factor is an estimated 3.89%
[Worker #2 Dec 9 10:29] Using AMD K8 type-3 FFT length 2688K, Pass1=896, Pass2=3K
[COLOR=Red][Worker #2 Dec 9 10:29] Error reading P-1 save file.[/COLOR]
[COLOR=Red][Worker #2 Dec 9 10:29] Renaming intermediate file m5P75351.bu to m5P75351.[/COLOR]
[Worker #2 Dec 9 10:29] Available memory is 2468MB.
[Worker #2 Dec 9 10:29] Using 2460MB of memory. Processing 106 relative primes (212 of 480 already processed).


[Sun Dec 4 22:55:42 2011]
SUMOUT error occurred.
[Sun Dec 4 22:55:42 2011]
SUMOUT error occurred.
[Tue Dec 6 02:01:42 2011]
UID: kladner/POD-95, M51960971 completed P-1, B1=505000, B2=10100000, E=6, We8: 40F30865
[Tue Dec 6 10:35:36 2011]
UID: kladner/POD-95, M52181383 completed P-1, B1=505000, B2=10100000, E=6, We8: 455533E1
[Tue Dec 6 14:44:48 2011]
M50573119 stage 1 is 49.93% complete.
M50575351 stage 1 is 16.40% complete.
[Wed Dec 7 14:01:30 2011]
[COLOR=Red]P-1 found a factor in stage #1, B1=490000.
ERROR: Bad factor for M50575351 found: 3
[/COLOR][Thu Dec 8 09:45:15 2011]
[COLOR=Red]P-1 found a factor in stage #2, B1=490000, B2=9677500.
ERROR: Bad factor for M50573119 found: 11
SUMOUT error occurred.[/COLOR]
[Thu Dec 8 09:50:55 2011]
M50575351 stage 1 is 72.57% complete.
[COLOR=Red]SUMOUT error occurred.[/COLOR][/CODE]The Stage 1 error above is the same one reported at the beginning of this thread. Since then there has been a Bad Factor result in the Stage 2 for a different exponent, and two SUMOUT errors.

Reliability/Confidence for this CPU is still 1.00, 0.0. However, it seems that something is not right with this machine, so Torture Test is its fate for the near future.

EDIT: We did have a power failure here on Dec 7, but it happened after the first Bad Factor result.

Any further info/advice would be appreciated.

I may be mistaken, but my understanding was that the Reliability/Confidence factors were related to LL/DC work only. Having a 1.00, 0.0 indicates you have never completed a DC/LL and thus are still at a 1.00 Reliability. My laptop is at 0.98, 7.0 indicating I have completed 7 LL with no errors. From what I remember, submitting an LL with no errors = .98 while a DC = 1.00 and if there are errors it = .50, but I can't remember where I saw it and I can't seem to narrow it down through search.

QUOTE bcp19: "I may be mistaken, but my understanding was that the Reliability/Confidence factors were related to LL/DC work only."

Thanks. I had not really understood the Reliability/Confidence report. This box in its current incarnation has only ever run P-1.

EDIT: It does seem that perhaps this thread belonged in Hardware. Maybe Torture Test will give more indications of what's going on.

Edit2: Worker 1 of this dual core Opteron ran 16 Blend tests in 48 minutes. It declared a fatal hardware error and shut down that worker. Worker 2 is still running. I'll give it a chance to crap out if it's going to, and then try Small FFTs.

For once, bcp19, I think I actually have something to contribute. I asked the same question myself a week or two ago and Brian-E directed me to [url=http://www.mersenneforum.org/showthread.php?t=11569]this thread[/url].