Message boards :
Number crunching :
Public beta for nVidia AstroPulse, rev 521
Message board moderation
Previous · 1 . . . 8 · 9 · 10 · 11 · 12 · 13 · 14 . . . 30 · Next
Author | Message |
---|---|
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
I like this one http://setiathome.berkeley.edu/workunit.php?wuid=795242703 5052 credits for a couple of hours work is not bad at all. PROUD MEMBER OF Team Starfire World BOINC |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Adding the space did the trick, I am now running two APs and one MB on my GPU. The second AP is moving slowly but it is moving so I guess everything is now right with the world. Thanks again Geek for pointing out my mistake. (by "moving slowly" I mean by GPU standards, it's still at least five times faster than on my CPU.) :-) PROUD MEMBER OF Team Starfire World BOINC |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
I like this one http://setiathome.berkeley.edu/workunit.php?wuid=795242703 5052 credits for a couple of hours work is not bad at all. NewCredit would be fine on a project where all tasks fully ran to completion and none exited early, The problem with NewCredit and Seti is Multibeam has the 'SETI@Home Informational message -9 result_overflow' early exit, and Astropulse has the 'In ap_remove_radar.cpp: get_indices_to_randomize: num_ffts_forecast < 100. Blanking too much rfi?' and the 'Found 30 single pulses and 30 repeating pulses, exiting.' early exits, If they weren't included in the calculations for APR and NewCredit, Credit granted would likely be a bit more stable, ie no more High Claims, But there would still be host normalization: •The host normalization mechanism reduces the claimed credit of hosts that are less efficient than average, and increases the claimed credit of hosts that are more efficient than average. Claggy |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Maybe I spoke too soon. I completed 5 APs last night. One validated, one is inconclusive. The one inconclusive and three others all show 30 repetitive pulses and will probably go invalid. I'm going to let it run this way for a little while to make sure but it looks like for me running more than one AP at a time is a no go. PROUD MEMBER OF Team Starfire World BOINC |
JohnDK Send message Joined: 28 May 00 Posts: 1222 Credit: 451,243,443 RAC: 1,127 |
Also have one inconclusive http://setiathome.berkeley.edu/result.php?resultid=2007662982 Think it was also when running 2 APs. Another thing when running 2 APs is that for some reason it used more CPU, normally it use 13% when running 1 with 275+ drivers, but when running 2 it goes sometime up to 26% I think it was. I've decided to only run one AP, seems the best. |
Slavac Send message Joined: 27 Apr 11 Posts: 1932 Credit: 17,952,639 RAC: 0 |
Also have one inconclusive http://setiathome.berkeley.edu/result.php?resultid=2007662982 One thing I did was take Kittyman's recommendation and leave one core free on my CPU. Might help in your situation. Executive Director GPU Users Group Inc. - brad@gpuug.org |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Also have one inconclusive http://setiathome.berkeley.edu/result.php?resultid=2007662982 Leaving one core free only is helpful with more than one GPU running. Wastet cycles otherwise. 30/30 units doesn´t mean it will not get credits. Just found to many signals and therfore early exit like -9 with MB. Most of them get granted. With each crime and every kindness we birth our future. |
X-Files 27 Send message Joined: 17 May 99 Posts: 104 Credit: 111,191,433 RAC: 0 |
No invalids or errors so far with 50+ AP. using cmd: -ffa_block 8192 -ffa_block_fetch 2048 -unroll 12 -instances_per_device 3 -hp cuda count: .34 Seti MB cuda count: .31 http://setiathome.berkeley.edu/results.php?hostid=2340442&offset=0&show_names=0&state=0&appid=5 |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Slavic, with only an E5400 dual and a GTS 450 it kind of defeats the purpose if I only run one core but thanks for the thought. X-Files27, I was running 6144/1536. I will try your suggestion of 8192/2048 and see how that goes. I'm running an unroll of 8 and a per device of 2 so I think I'll leave those that way for now. If I set the per device to 3 it will try to run a third instance and I'm sure my little 450 can't handle that. I also have .34/.31 set. I was a little afraid of messing with the FFA stuff as I didn't quite understand what to move it to. Mike, I know 30/30 is usually no problem but that's not what I'm getting. I'm getting the right number of single pulses but 30 repetitive pulses which is not right. I have to fiddle around and try to find out what is causing that. Another cruncher was getting the same problem a couple of days ago and I pointed it out to Raistmer but I haven't seen any solution as yet. PROUD MEMBER OF Team Starfire World BOINC |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
... Actually, 30 repetitive pulses isn't too unusual. True, if it isn't confirmed by a wingmate it's a problem. What is exceedingly unusual is 30 single pulses with no repetitive pulses, but even that may validate as your Task 2013445088 illustrates. Excerpts: ... Run time 13,945.13 CPU time 1,965.64 Validate state Valid Credit 666.64 Application version Astropulse v505 Anonymous platform (NVIDIA GPU) ... single pulses: 30 repetitive pulses: 0 percent blanked: 5.52 ... The wingmate ran stock CPU, which doesn't show signal counts. And the Astropulse validation only compares single pulse signals which are 1% above threshold, so there's a possibility it simply ignored those 30. The files have been deleted, so there's no way to check further. Joe |
Frizz Send message Joined: 17 May 99 Posts: 271 Credit: 5,852,934 RAC: 0 |
All my units result in 30 repetitive pulses for that host. - all ?/30 become invalids. - all the 30/30 get validated. And this is what I find quite bizarre: I mean, the repetitive pulses are obviously wrong. Are they not checked by the validators at all in this case? |
Lint trap Send message Joined: 30 May 03 Posts: 871 Credit: 28,092,319 RAC: 0 |
|
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Got my second invalid today. Sad i couldn´t download it for further testing. Single pulses 2 repetitive pulses 2 http://setiathome.berkeley.edu/workunit.php?wuid=776557486 With each crime and every kindness we birth our future. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
I have one inconclusive where the 3rd wingman is running Linux and Astropulse V505 v5.06. What is that?? That's just the ordinary stock application for Linux: see applications. I think segmentation violations are common with that app, but we'd best wait for a Linux specialist. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
All my units result in 30 repetitive pulses for that host. Not all the 30/30 get validated, for WU 788416427 yours was invalidated even though the other two hosts also found 30/30. Excerpt: Task Computer Sent reported Status 2007953553 5790778 24 Jul 2011 | 2:18:50 UTC 25 Jul 2011 | 13:15:59 UTC Completed, marked as invalid 2008013186 5759991 24 Jul 2011 | 3:22:02 UTC 24 Jul 2011 | 14:35:21 UTC Completed and validated 2010893549 2430488 25 Jul 2011 | 16:29:54 UTC 4 Aug 2011 | 17:14:26 UTC Completed and validated As to why your host is getting credit on the majority of its 30/30 tasks, in such a result there are actually 40 single pulses (30 reportable, 10 "best"). The criteria for a "Weakly similar" judgement by validation is that at least half the signals match, so in that sense wrong repetitive pulses don't prohibit getting credit (though your host's result should never be canonical). The reason it's not always granted credit is that single pulses are not compared if they aren't at least 1% above threshold, so in effect there will usually be less than 40 actually considered. Joe |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
All my units result in 30 repetitive pulses for that host. I can't check result's stderr, try to use default FFA values if you use smth else. |
JohnDK Send message Joined: 28 May 00 Posts: 1222 Credit: 451,243,443 RAC: 1,127 |
Any idea why when a AP task gets suspended or when I exit Boinc and a AP task is running, the screen often goes black for a sec and when the screen comes back a yellow triangle appears in the taskbar? Win7 then later wants to report a video hardware fail to MS. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
Any idea why when a AP task gets suspended or when I exit Boinc and a AP task is running, the screen often goes black for a sec and when the screen comes back a yellow triangle appears in the taskbar? Win7 then later wants to report a video hardware fail to MS. Sounds like your AP app hasn't used the threadsafe exit BOINC library yet. Sorry, that's a technote to the developer. |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Any idea why when a AP task gets suspended or when I exit Boinc and a AP task is running, the screen often goes black for a sec and when the screen comes back a yellow triangle appears in the taskbar? Win7 then later wants to report a video hardware fail to MS. Reduce unroll factor. Try -unroll 6. With each crime and every kindness we birth our future. |
JohnDK Send message Joined: 28 May 00 Posts: 1222 Credit: 451,243,443 RAC: 1,127 |
Any idea why when a AP task gets suspended or when I exit Boinc and a AP task is running, the screen often goes black for a sec and when the screen comes back a yellow triangle appears in the taskbar? Win7 then later wants to report a video hardware fail to MS. Nope, didn't help. |
©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.