| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Beta Testing
Forum: Beta Test Support Forum
Thread: Round 3 - GPU Beta Test for Help Conquer Cancer |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 151
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Okay, as having followed these GPU threads, I have not seen the following question: when do we Linux only crunchers get to participate ??? :) I do have one proven nVidia card with 1.2 capability ready to go.
|
||
|
|
martin64
Senior Cruncher Germany Joined: May 11, 2009 Post Count: 445 Status: Offline Project Badges: 
|
Any plans for checkpointing in GPU processing? To get longer running workunits we plan on packaging multiple images in a single workunit. For those there will be checkpoints taken in between images but at this point we have no plans on checkpointing the GPU applicaiton during an image. Thanks, armstrdj Ok, thanks for the info. "no plans" doesn't necessarily mean that it won't be done - so there's still hope... Anyway, since all of my results got an "invalid", resolving that would be my first priority.  Regards, Martin  |
||
|
|
KWSN - A Shrubbery
Master Cruncher Joined: Jan 8, 2006 Post Count: 1585 Status: Offline |
Okay, as having followed these GPU threads, I have not seen the following question: when do we Linux only crunchers get to participate ??? :) I do have one proven nVidia card with 1.2 capability ready to go. Agreed. I also worry that you've eliminated so many cards already that we won't get a chance to see if they work on Linux or not.  Distributed computing volunteer since September 27, 2000 |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges: 
|
One comment on the Beta's last night
----------------------------------------When running a GPUGRID WU I get no perceptible lag in my day to day PC work (It's running one now and my PC is still responsive - no effect on Media Player) With the HCC Beta's it's a seriously drunk mouse experience worse than remote desktop or logmein .... it's that slow and worse The scheduling must be different over at GPUGRID. If WCG remains like this then you will lose lots of potential new GPU crunching fans. (Note: I have sufficient Oomph  )Dave p.s. GPUGRID uses about 85% - 95% GPU  [Edit 2 times, last edit by David Autumns at Mar 20, 2012 8:01:12 AM] |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1316 Status: Offline Project Badges: 
|
Fine: NVidia is beloved again. I got 4 on my GTX 560.
----------------------------------------I didn't free a thread, so all threads busy with cpu-tasks. 3 ran 'normal', what imo means a lot of cpu during start and finish. 1 was also consuming cpu during the run, having a favorable effect on the elapsed time. Elapsed/CPU 09:44 / 1:18 10:13 / 1:21 09:37 / 1:14 06:15 / 3:58  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
when do we Linux only crunchers get to participate ??? I am wondering that also. I really feel kinda bad crunching for the "Linux Mint " team using Win 7   I'd very much like to reboot back to Mint and stay there , so... I'm really interested how well my GPU will perform on Linux.  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Think if you trial out at other GPU supporting grids, you'll be hitting the ground running when GPU beta arrives to WCG. There's no miracle worker out in the tech room... and we do remember months of testing AND production with CEP2 on Linux only before the Windows crunchers were even allowed to sniff at a Beta. ;>)
--//-- |
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges:
|
Fine: NVidia is beloved again. I got 4 on my GTX 560. I didn't free a thread, so all threads busy with cpu-tasks. 3 ran 'normal', what imo means a lot of cpu during start and finish. 1 was also consuming cpu during the run, having a favorable effect on the elapsed time. Elapsed/CPU 09:44 / 1:18 10:13 / 1:21 09:37 / 1:14 06:15 / 3:58 FWIW GPU tasks will run much faster if you free a CPU core to help your GPU, especially with Nvidia cards. Tasks on my GTX 460 averaged 4:40 running with that configuration.
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Fine: NVidia is beloved again. I got 4 on my GTX 560. I didn't free a thread, so all threads busy with cpu-tasks. 3 ran 'normal', what imo means a lot of cpu during start and finish. 1 was also consuming cpu during the run, having a favorable effect on the elapsed time. Elapsed/CPU 09:44 / 1:18 10:13 / 1:21 09:37 / 1:14 06:15 / 3:58 Which of the 2 times do you see [copy paste WU record from Result page plz]. It would be Elapsed, not CPU [GPU processing the reason why BOINC was switched from showing CPU time to Elapsed time]. Something to document in a GPU FAQ's for those that might feel short dealt when it comes to My Grid time stats. Trying to understand if it's pure GPU kernel Elapsed time + the CPU time of the start / finish sections, or just flat the Elapsed time, exclusive of any throttling i.e. if running BOINC at 50%, the Elapsed time would then be 50% of the time on the wallclock. --//-- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The kernel time , CPU time is a little confusing to me...
----------------------------------------I think this WU ran 3 min. 47 sec. by the wallclock and the Elapsed time in Boinc manager Total kernel time: 112.660301 (1026 kernel executions) Total memory transfer time: 1.408223 Average kernel time: 0.109805 Min kernel time: 0.092764 (dx=25 dy=3 sample_dist=24 ) Max kernel time: 0.138356 dx=1 dy=1 sample_dist=0 Total time for ../../projects/www.worldcommunitygrid.org/BETA_X0900060941422200511301450_X0900060941422200511301450.jp2: 220 seconds Finished Image #0, pctComplete = 1.000000 CPU time used = 219.228205 20:43:16 (1704): called boinc_finish </stderr_txt> ]]> Dosen't really add up exactly but I guess the 220 seconds is close to how long it actually ran. To me the only important time factor is how long it takes to run by the wallclock. That will determine how many WU I'll be able to run each day. I saw on another post HD 5830 extreme 217 seconds, kernel time 138.188293 seconds, CPU time 77.407696 seconds. Does that mean he will only get credit for 77.407696 seconds even though it took 217 seconds to run ? [Edit 2 times, last edit by Former Member at Mar 20, 2012 12:04:56 PM] |
||
|
|
|