Message boards :
Number crunching :
-9 Errors on C/GPU
Message board moderation
Author | Message |
---|---|
Dead4life Send message Joined: 22 Jun 04 Posts: 5 Credit: 6,960,721 RAC: 3 |
Gentlemen. I have just added a GT240 to my rig, in an effort to double my RAC: http://setiathome.berkeley.edu/show_host_detail.php?hostid=5730784 Having read a number of threads I decided to install the Lunatics apps for CPU and GPU, and then to watch the results that were being reported for any errors. Two errors have recently shown themselves, one on CPU, and one on GPU. These units both returned -9 errors, but they were both reported at the same time, which I think is odd, considering these are the only two units that have shown the error: http://setiathome.berkeley.edu/workunit.php?wuid=685558199 GPU http://setiathome.berkeley.edu/workunit.php?wuid=685432987 CPU If you look at the sent/received times, they are in line for my usual processing times.. Any thoughts or diagnosis? Euan |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
These seem to be legitimate -9s. The rest of your work is coming in okay. Wait on those two to see if your wingmen get -9s too. All that code means is that the work unit had too much noise in it mostly caused by ground interference. It's only a problem when you get all or most of your work marked as -9. That could mean your card is overheating or going bad. I'd say you have no problem now though. PROUD MEMBER OF Team Starfire World BOINC |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
A -9 isn't an an error as such, it just means the app has found enough signals: SETI@Home Informational message -9 result_overflow NOTE: The number of results detected exceeds the storage space allocated. If both your tasks validate against your wingmen when they complete their tasks, then there was no problem, Also note that GPU's are something prone to producing false -9's, you'll know about that since all your tasks on that GPU will be -9's, Claggy |
Fred J. Verster Send message Joined: 21 Apr 04 Posts: 3252 Credit: 31,903,643 RAC: 0 |
And why those -9 errors/real-overflows, are computed, is still not clear, IMO. One (CUDA) rig, Q6600+GTS250 (VISTA32Home), is still stable.(v.037;LUNATICs) And ofcoarse the X9650+GTX470 & 480, running now 2 WU's per card, is stable. Also v0.37 LUNATICs. (In both cases, still -9 results, no errors) (Never found a reproduceble error and/or reason, for faults on the 9800GTX+ & GTS250 combi, besides heat. Lowering GPU-Clock and/or undervolting, sometimes works, but are not that easy on most of the 200 series GPU's, a seperate program (TThrotle; eFMer), is needed. Just lost my oldest FlatScreen 17''; 4 years, 75% Uptime. :( Well, they are less expensive nowadays..... ADDED <message> - exit code -12 (0xfffffff4) </message> <stderr_txt> setiathome_CUDA: Found 2 CUDA device(s): Device 1: GeForce GTX 480, 1535 MiB, regsPerBlock 32768 computeCap 2.0, multiProcs 15 clockRate = 1388000 Device 2: GeForce GTX 470, 1279 MiB, regsPerBlock 32768 computeCap 2.0, multiProcs 14 clockRate = 1215000 setiathome_CUDA: CUDA Device 1 specified, checking... Device 1: GeForce GTX 480 is okay SETI@home using CUDA accelerated device GeForce GTX 480 ---[Snipped]--- Multibeam x32f Preview, Cuda 3.0 Work Unit Info: ............... WU true angle range is : 2.702858 SETI@home error -12 Unknown error cudaAcc_find_triplets doesn't support more than MAX_TRIPLETS_ABOVE_THRESHOLD numBinsAboveThreshold in find_triplets_kernel File: d:/[Projects]/Berkeley/seti_cuda/seti_boinc/client/cuda/cudaAcc_pulsefind.cu Line: 253 Result 1770622711. During this period, 3 WU's per card, were running, switched back to 2. |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Speaking of -9s, I see our buddy smithwr3 still hasn't got the word on the new lunatic's app. Just got another inconclusive from him running the old V12 app. He's got a RAC of 4664 thanks to his I7 but I guess he doesn't care that he could be so much higher if his two 480s were turning out good work. He's got over 17,000 work units on that machine and I'd bet most of them will be marked invalid. PROUD MEMBER OF Team Starfire World BOINC |
Bernie Vine Send message Joined: 26 May 99 Posts: 9954 Credit: 103,452,613 RAC: 328 |
Speaking of -9s, I see our buddy smithwr3 still hasn't got the word on the new lunatic's app. Just got another inconclusive from him running the old V12 app. He's got a RAC of 4664 thanks to his I7 but I guess he doesn't care that he could be so much higher if his two 480s were turning out good work. He's got over 17,000 work units on that machine and I'd bet most of them will be marked invalid. Yes I had one recently paired with him. I sent a pm, but got no response. |
James Sotherden Send message Joined: 16 May 99 Posts: 10436 Credit: 110,373,059 RAC: 54 |
My i7 with its little old GTS 250 has a rac of over 11,000. Glad to know I can beat a fermi card.:) [/quote] Old James |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
My i7 with its little old GTS 250 has a rac of over 11,000. Glad to know I can beat a fermi card.:) Feel even better, he has a pair of 480s. You're beating two of them! Bernie, I sent him a PM maybe two months ago and got no reply. He must have Email notification turned off. The last time he posted was over 1400 days ago so it looks like he doesn't look in here either. PROUD MEMBER OF Team Starfire World BOINC |
Geek@Work Send message Joined: 29 Oct 02 Posts: 6 Credit: 753,666 RAC: 0 |
Berkeley could block smithwr3 account if they felt there was any damage to the project. Since they have not done so............???????????? Boinc....Boinc....Boinc....Boinc |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
It means they have no time for this manual work..... |
Fred J. Verster Send message Joined: 21 Apr 04 Posts: 3252 Credit: 31,903,643 RAC: 0 |
|
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.