Message boards :
Number crunching :
SETI applications for NVIDIA GPU improvement - how you can help
Message board moderation
Previous · 1 . . . 8 · 9 · 10 · 11 · 12 · 13 · 14 · Next
Author | Message |
---|---|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
8.19 (r3528) is deployed on beta. Please test. Your host already receives SoG so no abortions required. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Are there any default settings for switches (ex  -sbs N where N=128) of which we should be aware? It tested on beta currently as new stock RC. SETI apps news We're not gonna fight them. We're gonna transcend them. |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
. . I take it that you want r3528 tested, but is there a beta installer for it? I read the instructions BilG kindly posted on manual installation but decided there are too many ways I could stuff it up. . . Hi Jim, . . So now my poor little donkey will have to be changed over to S@H Beta. Another learning curve for me. I have never run multiple projects before :( Stephen . |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
. . Hi Raistmer, . . OK I will now learn how to run S@H Beta on the little donkey (C2D with GT730). Stephen . |
Jimbocous Send message Joined: 1 Apr 13 Posts: 1856 Credit: 268,616,081 RAC: 1,349 |
Hi. Just remember not to run QOpt or GUPPIRescheduler while you have any Beta tasks present on your PC or are accepting work from Beta project. That's what the whole Y prompt you'll get from Kevvy is all about. |
Jimbocous Send message Joined: 1 Apr 13 Posts: 1856 Credit: 268,616,081 RAC: 1,349 |
8.19 (r3528) is deployed on beta. Please test. OK, Thanks ... |
EdwardPF Send message Joined: 26 Jul 99 Posts: 389 Credit: 236,772,605 RAC: 374 |
As of 22:30UT I'm still getting v8.01 any ideas?? Ed F |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
How about resetting the beta project and see if that corrects the issues |
EdwardPF Send message Joined: 26 Jul 99 Posts: 389 Credit: 236,772,605 RAC: 374 |
Tried that with no luck .... I guess I'll "quit" the beta project and join it again ... (getting desperate here!) Ed F Edit: Nope ... still getting v8.01 cuda50's |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Tried that with no luck .... I guess I'll "quit" the beta project and join it again ... (getting desperate here!) As I said before you need to lower quota to 1 per day to start to receive some other plan classes immediately. Current quotas for CUDA work ~30 tasks. So, ~30 tasks for eacg CUDA plan class needed to be aborted. SETI apps news We're not gonna fight them. We're gonna transcend them. |
EdwardPF Send message Joined: 26 Jul 99 Posts: 389 Credit: 236,772,605 RAC: 374 |
on new_left I have aborted 44 v8.01 WU's and still I get more ... ??? Ed F Edit: for each plan!!! o.k. more work to do ... Edit2: finished killing the WU's and nothing would run ... I cant kill 30 WU's each for 3 classes of cuda on 4 computers ... Back to main and more SoG fun |
Jimbocous Send message Joined: 1 Apr 13 Posts: 1856 Credit: 268,616,081 RAC: 1,349 |
Seeing a few -9 restarts. As I recall, this is no big deal? |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
. . HI Guys, . . I have been trying to create an account on Beta to run r3528 but I am unable to. . . I go to the home page but when I select "Create new account" it throws up an error ... Fatal error: Call to undefined function recaptcha_get_head_extra() in /disks/carolyn/b/home/boincadm/projects/beta/html/user/create_account_form.php on line 41 . . So what am I doing wrong? . . As usual, any/all help appreciated. Stephen . [edit] I went in through "Add Projects" by modifying the URL for the Seti@Home option in the list to the Beta url. This time it worked. Stephen . |
Jimbocous Send message Joined: 1 Apr 13 Posts: 1856 Credit: 268,616,081 RAC: 1,349 |
Excellent! |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
. . It is running OK, but I have to ask, are those WUs for Beta "special" in any way? So far have had 6 - 8 SoG 8.19, about 20 CUDA50 and a dozen or more CUDA42, and they are running in about 24/25 mins, which is interesting because using r3500 in main they were taking about 35 to 38 mins. . . Is there anything you want me to look for Raistmer? . . I was a little confused to find S@H-main and S@H-beta tasks all mixed up in the one table in manager :( Stephen . |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Tasks on beta similar to tasks on main (besides 4-bit data coding that doesn't affect runtime). One should get familar with strong AR dependence of MB runtimes and never forget about it. Also, you have 3 hosts - about what you talk? SETI apps news We're not gonna fight them. We're gonna transcend them. |
Jimbocous Send message Joined: 1 Apr 13 Posts: 1856 Credit: 268,616,081 RAC: 1,349 |
The one thing I never understand is this. Here we are trying to prove out latest build of SoG. I have two hosts on Beta project (ID: 77530,77532) that are getting only 1-2 SoG, mostly use Cuda42 and Cuda50, where ID: 78348 is getting a nice run of SoG. Very similar machines. I can understand the need for baselining and comparison, but is that intentional or an issue? Doesn't make much sense. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14674 Credit: 200,643,578 RAC: 874 |
The one thing I never understand is this. It's because the BOINC infrastructure framework wasn't really designed for this sort of experimentation. It was designed for deterministic production runs: "this task, that hardware - so send this application". It doesn't cope well with choices: "maybe this app, maybe that app, we'll try them both (or all five) until we can decide which is best". We could force SoG r3528 to the foreground by deprecating the cuda apps, but then the other requirement for testing - comparison of results with other applications for similarity - would be lost. Testing - the way we do it - is a multi-stage process. Do apps actually compile and run on the multiplicity of hardware out there? Do apps produce the expected answers for a small set of known test pieces? Do apps produce the expected answers for a larger set of real-world cases, at Beta? Do apps produce the right answer for every possible case that the radio telescopes and data recorders - themselves subject to error and failure - throw at us? If you can design a testing regime that answers that last question, you've probably talked yourself into a job. |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
Tasks on beta similar to tasks on main (besides 4-bit data coding that doesn't affect runtime). . . Hi, . . I can't see a tasks or computers button to check the results for task AR so I am in the dark there, but I had noticed the tasks d/l's were 700Kb not the usual 358Kb. So I was wondering if they were all HAR except for the guppis and VLAR tagged files. . . It is on Mi_Burrito the Core(2) Duo E7600 with the GT730, host #7972413. It is the only one on S@H-Beta. . . The VLAR runtimes indicate it is not significantly faster and it is yet to run any guppis but the Arecibo tasks are all running in about 25 mins. So I thought perhaps it was another long run of HAR WUs. Stephen . |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Dr. K increased the size of the work units on Beta to 700KB to see if they had any impact on the processing of the work units. Originally MB work units were cut down to smaller sizes to allow for better computations long time ago when they were sent out. So the question was, did the work units need to be "cut down" or could the larger size work units be crunched and if so, how would it affect they times. There is no problems crunching the larger work units and the times did not change significantly. |
©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.