WARNING: This website is obsolete! Please follow this link to get to the new Albert@Home website!
[New release] BRP app v1.22 feedback thread |
Message boards :
Problems and Bug Reports :
[New release] BRP app v1.22 feedback thread
Message board moderation
Previous · 1 · 2 · 3 · Next
Author | Message |
---|---|
Nikolay Send message Joined: 13 Jan 12 Posts: 4 Credit: 6,500 RAC: 0 |
Since v1.21/v1.22 update, I have got a lot of validation errors. Please check my stats/failed units, maybe that will help you to find a bug. My system is BOINC 7.18 + Mac OS X 10.7.3 + AMD HD6750M 1GB |
[AF>Le_Pommier] McRoger Send message Joined: 9 Feb 08 Posts: 3 Credit: 216,378 RAC: 0 |
|
[AF>Le_Pommier] McRoger Send message Joined: 9 Feb 08 Posts: 3 Credit: 216,378 RAC: 0 |
Same for me, my results Besides there is this error in the log [19:14:19][1216][INFO ] Checkpoint file unavailable: status.cpt (No such file or directory). Might it explain why calculation time is so huge compared to other platforms ? |
Alex Send message Joined: 1 Mar 05 Posts: 88 Credit: 398,734 RAC: 0 |
Not on my machine. vbox (test4theory, 2 cpu's) and three albert BRP's (2 ati, one nvidia) are running fine together. Some are waiting for validation, some are validated, no errors or invalids. win7 x64 8GB 7.0.12 |
Oliver Behnke Volunteer moderator Project administrator Project developer Send message Joined: 4 Sep 07 Posts: 130 Credit: 8,545,955 RAC: 0 |
Hi Jord, So, pretty please, can the fpops estimate be adjusted enough that they don't come in thinking to take 200+ hours? I'll forward this to Bernd but he's pretty overwhelmed with more important topics right now and the BOINC devs are of little help analyzing this right now. Please bear with us. Cheers, Oliver |
Oliver Behnke Volunteer moderator Project administrator Project developer Send message Joined: 4 Sep 07 Posts: 130 Credit: 8,545,955 RAC: 0 |
Same for me, my results 1) Please read my intro post of this thread. The Mac version is known to produce invalid results. We already disabled it. 2) Your GPU is simply not that efficient that's why it takes so long. It's not about the platform but the GPU. 3) The message you quote is an "INFO" message, so no, it's not the reason. It's normal when a fresh dataset is being analyzed for the first time - there can't be any checkpoint then. HTH, Oliver |
Oliver Behnke Volunteer moderator Project administrator Project developer Send message Joined: 4 Sep 07 Posts: 130 Credit: 8,545,955 RAC: 0 |
Since v1.21/v1.22 update, I have got a lot of validation errors. Same as for "[AF>Le_Pommier] McRoger": no working OS X OpenCL app for the time being... Oliver |
pragmatic prancing periodic problem child, left Send message Joined: 26 Jan 05 Posts: 1639 Credit: 70,000 RAC: 0 |
Hi Jord, It may be quite easy. I changed <rsc_fpops_est>300000000000000.000000</rsc_fpops_est> to <rsc_fpops_est>30000000000000.000000</rsc_fpops_est> (one zero less) and restarted BOINC. Estimated time on a new task is now 15 hours, which is more in line than the original 208 hours. Jord. BOINC FAQ Service They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake. |
Oliver Behnke Volunteer moderator Project administrator Project developer Send message Joined: 4 Sep 07 Posts: 130 Credit: 8,545,955 RAC: 0 |
It's not. The reason isn't a flaw in our runtime estimation (stored in the work unit definition) but BOINC's new automatic runtime estimation system (a.k.a. new credit system) we're also testing here on albert... Oliver |
Alex Send message Joined: 1 Mar 05 Posts: 88 Credit: 398,734 RAC: 0 |
It's not. The reason isn't a flaw in our runtime estimation (stored in the work unit definition) but BOINC's new automatic runtime estimation system (a.k.a. new credit system) we're also testing here on albert... ... and it looks like its faulty ???? |
skildude Send message Joined: 15 Nov 11 Posts: 9 Credit: 103,497 RAC: 0 |
I've gotten a great deal of invalids and inconclusives. Somethings wrong and I don't think its my GPU |
spingadus[MM] Send message Joined: 15 Oct 06 Posts: 4 Credit: 250,000 RAC: 0 |
Just wanted to post that I finally got some Albert tasks! I followed Skildude and Ageless's advice. I uninstalled the ATI drivers, then ran driversweep in safe mode. I found a lot of old Nvidia stuff as well from previous cards. I also updated to 7.0.20 just for the heck of it. The Boinc startup messages did show opencl for the gpu this time. Thanks! |
Oliver Behnke Volunteer moderator Project administrator Project developer Send message Joined: 4 Sep 07 Posts: 130 Credit: 8,545,955 RAC: 0 |
Well, let's say it's non-optimal, in particular for GPU apps. The runtime estimates are determined for every application version independently. Thus after each newly released version BOINC needs some time to gather statistics to come up with a valid/reasonable runtime estimate. Don't worry, we won't be using this new system over on einstein until it proves reliable, but we need to test it here in order to improve (fix) it at all - as soon as time permits. Best, Oliver |
oz Send message Joined: 28 Feb 05 Posts: 10 Credit: 1,285,478 RAC: 0 |
Today I had a lot of atiOpenCL tasks aborted after exactly 24:14 min. 133328 43214 7 Mar 2012 | 16:58:05 UTC 8 Mar 2012 | 6:11:00 UTC Error while computing 1,454.57 580.95 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133327 39414 7 Mar 2012 | 16:59:13 UTC 8 Mar 2012 | 6:11:00 UTC Error while computing 1,453.71 578.01 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133326 39395 7 Mar 2012 | 16:59:13 UTC 8 Mar 2012 | 6:11:00 UTC Error while computing 1,454.23 582.54 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133325 39432 7 Mar 2012 | 16:59:13 UTC 8 Mar 2012 | 8:30:35 UTC Error while computing 1,453.80 582.52 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133324 39441 7 Mar 2012 | 16:59:13 UTC 8 Mar 2012 | 8:30:35 UTC Error while computing 1,453.70 586.05 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133323 43314 7 Mar 2012 | 16:58:05 UTC 8 Mar 2012 | 6:11:00 UTC Error while computing 1,454.00 584.04 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133321 39279 7 Mar 2012 | 16:55:50 UTC 8 Mar 2012 | 6:11:00 UTC Error while computing 1,453.96 582.06 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133320 37403 7 Mar 2012 | 17:00:24 UTC 8 Mar 2012 | 8:30:35 UTC Error while computing 1,454.57 614.77 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133319 36932 7 Mar 2012 | 17:00:24 UTC 8 Mar 2012 | 8:30:35 UTC Error while computing 1,453.83 607.69 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133318 44053 7 Mar 2012 | 17:00:25 UTC 8 Mar 2012 | 10:38:10 UTC Error while computing 1,453.83 662.62 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133317 38006 7 Mar 2012 | 17:01:34 UTC 8 Mar 2012 | 12:57:42 UTC Error while computing 1,454.31 665.01 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133316 43437 7 Mar 2012 | 16:58:05 UTC 8 Mar 2012 | 6:11:00 UTC Error while computing 1,454.19 582.07 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) Bikemans: 133387 44311 7 Mar 2012 | 17:42:19 UTC 8 Mar 2012 | 9:42:23 UTC Error while computing 947.54 846.04 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133348 44229 7 Mar 2012 | 17:42:19 UTC 8 Mar 2012 | 9:42:23 UTC Error while computing 946.95 840.90 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133346 44226 7 Mar 2012 | 17:43:26 UTC 8 Mar 2012 | 10:07:15 UTC Error while computing 947.58 838.51 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133314 39550 7 Mar 2012 | 17:41:09 UTC 8 Mar 2012 | 4:31:47 UTC Error while computing 946.85 838.29 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 133274 44093 7 Mar 2012 | 17:41:09 UTC 8 Mar 2012 | 4:31:47 UTC Error while computing 947.08 842.30 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 130790 44395 7 Mar 2012 | 17:43:27 UTC 8 Mar 2012 | 10:43:45 UTC Error while computing 946.86 844.07 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) 130749 44374 7 Mar 2012 | 17:42:19 UTC 8 Mar 2012 | 9:42:23 UTC Error while computing 947.30 837.06 --- Binary Radio Pulsar Search v1.22 (atiOpenCL) PS.: Bikemans end up earlier due to better hardware |
Oliver Behnke Volunteer moderator Project administrator Project developer Send message Joined: 4 Sep 07 Posts: 130 Credit: 8,545,955 RAC: 0 |
Now that's strange. Looks like BOINC's borked runtime estimation again. Thanks for reporting... Oliver |
choks Send message Joined: 24 Feb 05 Posts: 5 Credit: 1,110,845 RAC: 0 |
Hi, I also had my tasks ended after 702 seconds (tasks 130935,130925,130906 for example). I had to divide <flops> by 10 in client_state.xml to allow tasks to finish. I just upgraded to catalyst 12.12 (7/3/2012) and the good news for Linux users is that the CPU usage was significantly reducted. 1300 seconds of CPU time per work, instead of about 3600 with 12.11. Average CPU is now about 33%. Christophe |
Trog Dog Send message Joined: 25 Nov 05 Posts: 204 Credit: 64,008 RAC: 0 |
All 1.22 wu's are erroring out with max time elapsed http://albert.phys.uwm.edu/results.php?userid=128605&offset=0&show_names=0&state=5&appid= running on boinc 7.0.20 ati drivers 12.2 |
pragmatic prancing periodic problem child, left Send message Joined: 26 Jan 05 Posts: 1639 Credit: 70,000 RAC: 0 |
I don't know if this affects the OpenCL in any way, but the Catalysts 12.2 do cause Anti Aliasing problems in some games. I noticed it after upgrading to these drivers, that all fine mist like graphics in Skyrim would become lots of square pixels. This can only be fixed by disabling AA and enabling FSAA instead. ATI says it's a game problem, not their drivers, but heck if something works before and doesn't after changing the drivers, then how can that be the game's problem when that one hasn't changed literally a bit? Jord. BOINC FAQ Service They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake. |
pragmatic prancing periodic problem child, left Send message Joined: 26 Jan 05 Posts: 1639 Credit: 70,000 RAC: 0 |
And again... Normal average run time of OpenCl tasks on my ATI HD6850 is around 6200 seconds. When not interrupted. When interrupted (due to exit BOINC, suspend BOINC or suspend task (exclusive_app or switch between applications)), task run time length increases to 31,000 - 36,000 seconds (!!). (task list) Jord. BOINC FAQ Service They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake. |
Alex Send message Joined: 1 Mar 05 Posts: 88 Credit: 398,734 RAC: 0 |
As Tullio posted in an other thread, the Albert wu's are slower than the Einstein wu's. I checked it twice with BRP3cuda32 wu's, running all of them with the same setting (GPU 0.5) on the same hardware. @ Jord: I don't see this behaviour on my machine, I turn it off sometimes, put tasks on hold, start them on my HD5830 and let it finish on the APU. I have no tasks running longer than 12800 sec. |