Message boards :
Number crunching :
Inconclusive Work Units Running AP Ver 6
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 10 · Next
Author | Message |
---|---|
LadyL Send message Joined: 14 Sep 11 Posts: 1679 Credit: 5,230,097 RAC: 0 |
Keep in mind most of 560ti are factory overclocked. yes but those are inconclusive. I'm not the Pope. I don't speak Ex Cathedra! |
cliff Send message Joined: 16 Dec 07 Posts: 625 Credit: 3,590,440 RAC: 0 |
Add to that how many are crunching more than 1 WU at a time.. I can only crunch 1 per GPU, my psu is'nt up to handling both GPU's if I try to run more than 1 per.. This could be either hardware or software related or even a bit of both, given that nvidia drivers have had a tendancy to push such app's onto the CPU as well as the GPU. I watched some AP6 tasks running with a graph depicting GPU usage.. under 285.62 nv drivers the percentage continually varied between 6% and 43% with some peaks of 66% and very occaisionally 81%.. The lower the percentage overall it 'seemed' related to the degree of blanking reported. Unfortunately NV 301.24 stopped my monitoring software working, so I can no longer watch usage as before.. But then I no longer have any AP6 WU's to process, none are available.. [edit] seems I was wrong about AP6 I do indeed have 1 such task waiting in the queue. Regards, Cliff, Been there, Done that, Still no damm T shirt! |
musicplayer Send message Joined: 17 May 10 Posts: 2430 Credit: 926,046 RAC: 0 |
Oh, better trust the results and the numbers being returned more than the user(s) behind. |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Okay, I guess I'd better give my information. My little GTS450 is running a little overclocked at 885/1804 and is between 62 and 63*C. I'm also running 51/49 and my command line looks like this... <cmdline>-instances_per_device 1 -unroll 10 -ffa_block 8192 -ffa_block_fetch 4096</cmdline>. Anything else you need to know? PROUD MEMBER OF Team Starfire World BOINC |
X-Files 27 Send message Joined: 17 May 99 Posts: 104 Credit: 111,191,433 RAC: 0 |
Keep in mind most of 560ti are factory overclocked. I will adjust the params once I get home. unroll will be 6. Other notes: 0.51 for AP and 0.49 for MB GPUs are OC'ed thru the bios XFX PRO1250W Black Edition = kill-a-watt says 932W usage |
Jim_S Send message Joined: 23 Feb 00 Posts: 4705 Credit: 64,560,357 RAC: 31 |
Where do I find stderr_txt? I Desire Peace and Justice, Jim Scott (Mod-Ret.) |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Your accounts page /tasks/ task. Take one of the tasks you have completed and open the section marked task,click for details, on the left hand side. PROUD MEMBER OF Team Starfire World BOINC |
Jim_S Send message Joined: 23 Feb 00 Posts: 4705 Credit: 64,560,357 RAC: 31 |
Your accounts page /tasks/ task. Take one of the tasks you have completed and open the section marked task,click for details, on the left hand side. Thanks much My Friend. I Desire Peace and Justice, Jim Scott (Mod-Ret.) |
cliff Send message Joined: 16 Dec 07 Posts: 625 Credit: 3,590,440 RAC: 0 |
I've just noticed something slightly strange. I'm monitoring both my GPU's with msi afterburner and an AP6 GPU task has just been loaded and is running. Immediatley it loaded, both my GPU's started to display the same behaviour, ie graphs of usage started to oscillate wildly. This was not seen on the other GPU prior to the AP6 task being loaded. That graph was a near perfect line with only the occaisional dip at odd intervals. That GPU is running an e@h gpu task. So it seems that simply loading and running an AP6 GPU task affects both GPU's even if they are not running the same type of task. I hadnt noticed this before because I'd been running the same type of task on both GPU's ie AP6 ones. Anyone have any ideas whay this is happening? [edit] It seems that both GPU's are in some way synchronised, as the graphs rise and lower together, differing only in amplitude.Could this be a driver related process? Regards, Cliff, Been there, Done that, Still no damm T shirt! |
Horacio Send message Joined: 14 Jan 00 Posts: 536 Credit: 75,967,266 RAC: 0 |
Extra info that may be worth about my inconclusive GPU WUs (560TIs): These 3 WUs were crunched on the suspected failling GPU: ap_08jn11aa_B5_P1_00343_20120412_19722.wu (30 pulses w/o repetitives) Still Inconclusive ap_25ja12ad_B3_P1_00044_20120411_05125.wu (30 pulses w/o repetitives) Invalid (against 2 stock V6 6.01) ap_26no11ae_B2_P0_00210_20120315_31464.wu (30 pulses w/o repetitives) Invalid (against 2 stock V6 6.01) This WU was crunched on the good GPU (only 1 invalid over 2K WUs crunched): ap_11my11ab_B2_P0_00278_20120409_08996.wu (30 pulses with 30 repetitives) Still Inconclusive The command line parameters used was for sure: -instances_per_device 2 -unroll 8 -ffa_block 8192 -ffa_block_fetch 4096 (Ive recently added also -no_cpu_lock and -hp, but Im not sure if these settings were applied to those WUs.) Right now, the host with the 560Tis is not crunching SETI anymore (as there is no BOINC version doing the schedulling/workfecth in the way I want, Ive assigned only one project per host, which seems it is helping a lot to increase the SETI RAC) |
LadyL Send message Joined: 14 Sep 11 Posts: 1679 Credit: 5,230,097 RAC: 0 |
I've got something that probably explains why some cards cope with high unroll and some don't, but I can't type it up easy right now. For the time being, if you see invalids from units with 30 (rep) pulses it's most likely GPU memory corruption due to too high unroll. Dropping unroll all the way to 2 should cure it, or you can try if intermediate values work. Cards that will probably not take high unroll are GTX 470, 480, 570 and 580. Cards that should work with high unroll are GTX 460, 560, 560 Ti and 680. I'm not the Pope. I don't speak Ex Cathedra! |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
I'm posting this one http://setiathome.berkeley.edu/workunit.php?wuid=969007370 not because it's an Inconclusive but because of the high numbers 14/30 counts. When I was running Raistmers app for 5.05s a count like that would almost assuredly error out. This one and another like it earlier have validated. Okay Lady L. now you're confusing me. I'm getting some with high rep. counts but they are validating. Should I cut back on my unroll or not? So far, none of mine have errored out. PROUD MEMBER OF Team Starfire World BOINC |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
I'm posting this one http://setiathome.berkeley.edu/workunit.php?wuid=969007370 not because it's an Inconclusive but because of the high numbers 14/30 counts. When I was running Raistmers app for 5.05s a count like that would almost assuredly error out. This one and another like it earlier have validated. At least that one validated against the stock app, so we can be reasonably sure that the high counts are genuine - although the lack of signal reporting in the stock app's stderr_txt always leaves a nagging doubt in one's mind. The real dangers start when an app with a tendency to false detections becomes widespread enough to start validating against itself, as we saw during the V12/Fermi saga. |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
I'm posting this one http://setiathome.berkeley.edu/workunit.php?wuid=969007370 not because it's an Inconclusive but because of the high numbers 14/30 counts. When I was running Raistmers app for 5.05s a count like that would almost assuredly error out. This one and another like it earlier have validated. Number of signal alone tells nothing at all. I had dozens which validate in first attempt. Its only a indiction if the wingmen gets different amount of signals. With each crime and every kindness we birth our future. |
LadyL Send message Joined: 14 Sep 11 Posts: 1679 Credit: 5,230,097 RAC: 0 |
Okay Lady L. now you're confusing me. I'm getting some with high rep. counts but they are validating. Should I cut back on my unroll or not? So far, none of mine have errored out. if they are validating it's ok. if you get early exits (30/30) and they _don't_ validate, then something is off. same story as with excessive -9 on MB, really. Unless you're getting high numbers of inconclusives/invalids with those 30 signals, you're in the clear. I'm not the Pope. I don't speak Ex Cathedra! |
X-Files 27 Send message Joined: 17 May 99 Posts: 104 Credit: 111,191,433 RAC: 0 |
So far so good after lowering it down to 6 from 8. 4 valids and 0 inconclusive so far. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
I've just got my second but this time on my 2500K, http://setiathome.berkeley.edu/workunit.php?wuid=972977092. Cheers. |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
I've just got my second but this time on my 2500K, http://setiathome.berkeley.edu/workunit.php?wuid=972977092. Both you and your wingman found 30/30 pulses so it will probably get validated. With each crime and every kindness we birth our future. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
Richard Haselgrove wrote: tbret wrote:I didn't like that list over there. Here it is where it ought to be, I guess: That one should give credit when another result is received. The AP Validator has a critical level for single pulses, it only checks those with peak power at least 1% above threshold. In that WU there's a single pulse which the optimized app calculates as being 0.996% above threshold, but stock gets 1.002% above threshold. The 0.006% difference is a lot tighter than the 0.1% allowed tolerance, but the Validator ignores the one from optimized. Those cases are fairly rare, and would require revising the Validator code to eliminate. Joe |
Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489 |
OK both my inconclusives validated. http://setiathome.berkeley.edu/workunit.php?wuid=968825720 http://setiathome.berkeley.edu/workunit.php?wuid=972977092 Cheers. |
©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.