| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Beta Testing
Forum: Beta Test Support Forum
Thread: HCC1 Beta Test for Linux GPU (Issues Thread) |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 49
|
|
| Author |
|
|
kateiacy
Veteran Cruncher USA Joined: Jan 23, 2010 Post Count: 1027 Status: Offline Project Badges: 
|
I had a flawless AMD/ATI driver install  . I now see GPU-use-mode controls under the Activity menu where there was once none. The event log now has a description of the AMD/ATI GPU card as follows:-- ATI GPU 0: Juniper (CAL version 1.4.1741, 1024MB, 804MB available, 2720 GFLOPS peak) -- OpenCL: ATI GPU 0: Juniper (driver version 1016.4, device version OpenCL 1.2 AMD-APP (1016.4), 1024MB, 804MB available). Oh good! Glad everything went smoothly. Interesting that you'd already had fglrx-updates installed. My guess had been that you had the open source driver. It's possible that just manually restarting BOINC would have done the trick for you since there was an AMD proprietary driver there, but as long as this worked for you, great! And now you know how to install software and change drivers in Ubuntu!  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Oh good! Glad everything went smoothly. Interesting that you'd already had fglrx-updates installed. My guess had been that you had the open source driver. It's possible that just manually restarting BOINC would have done the trick for you since there was an AMD proprietary driver there, but as long as this worked for you, great! And now you know how to install software and change drivers in Ubuntu! Yup, I had the 'fglrx-updates' and a simple restart might have done the trick there as well. I'll experiment right now: I'll install the fglrx-updates and do the restart. Stay tuned...; ; andzgridPost#717 ; edit1_2012.12.05We.1830utc: Ok, my Ubu12.10-machine is now on 'fglrx-updates', and after the BOINC-restart, BOINC did show the GPU-use-mode controls under Activity. Looking back now, it seemed that all that was required was a BOINC-restart for either the 'updates' variety or the non-updates variety -- at least for my machine. I also explored around the SC (software-center) and found that I have 4 'Catalyst Control Center for the AMD graphics accelerators' softwares: 1] fglrx-amdcccle-updates (marked as installed) 2] fglrx-amdcccle (installable) 3] fglrx-amdcccle:i386 (installable) 4] fglrx-amdcccle-updates:i386 (installable) The 'i386' there, I guess, is for my machine's motherboard AMD graphics (I guess not usable for WCG OpenCL), which apparently also comes with the 'updates' and non-updates variety. The SC does not spell out the difference between the softwares; it doesn't get any more cryptic than that! Grrrr.... ; [Edit 1 times, last edit by Former Member at Dec 5, 2012 6:30:08 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I got quite a few of the GPU Betas, but as noted previously, they do not honor the setting to not use the GPU while the computer is in use. They brought my work machine to it's knees, and I had to abort them. After aborting, I got more; until I finally had to turn off getting Beta WU's for that profile. I would like to run the Betas during offpeak hours, but I cannot afford to have then running while I am trying to get work done (sigh...)
 . |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Well, the number one to check is that activity menu to see if the GPU option is also set to "run based on preferences". The "run based on preferences" is controlled separately for CPU and GPU.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
A] For 1st 9-WUs: 9-valid
----------------------------------------PointsGranted(min)=173.0; PointsGranted(max)=199.3 B] Next 23-WUs: 23-valid PointsGranted(min)=60.7; PointsGranted(max)=77.5 C] Common: Beta_HCC_GPUv7.08; Three(3)-items uploaded per doneWU. ElapsedTime(min): 8.4minutes; ElapsedTime(max): 10.8minutes. D] Visible Issues -- video-lag at GPU-crunch phase when desktop is used. E] Setup Ubu12.10_64bit, BOINC_v7.0.27_64bit HD5770_PCIe2_16x, AMD1090T, 8MB-RAM desktop-UI not decoupled from HD5770 stock-clocked, 24/7-operation ; ; andzgridPost#718 ; edit1_2012.12.06Th.0615utc: -- typo correction -- updated data ; [Edit 1 times, last edit by Former Member at Dec 6, 2012 6:29:39 AM] |
||
|
|
ashrader330
Advanced Cruncher Joined: Jan 6, 2008 Post Count: 97 Status: Offline Project Badges:
|
I now have 21 pages of valid and pending validation and not a single error or invalid. This is with a GTX 570 and GTX 460. Looks real good to me!
---------------------------------------- Run time: 4.2y HPF2, 6.9y FAAH, 7.9y HFCC, 20.8y HCC, 26.0y CEP2, 26.0y MCM, 2.1y UGM, 2.0y OET WU: 4.8k HPF2, 12.3k FAAH, 12.4k HFCC, 135k HCC, 34.3k CEP2, 43.8k MCM, 4.2k UGM, 19.7k OET |
||
|
|
widdershins
Veteran Cruncher Scotland Joined: Apr 30, 2007 Post Count: 674 Status: Offline Project Badges: 
|
Every unit that I've processed that I've looked at that had errored out isn't actually an error in my view, they've all aborted at the very start with the line.
----------------------------------------ERROR: Kernel execution time estimate too high, exiting. I haven't found one that has errored out for any other reason. Everyone that has made it past the first couple of seconds has validated. The machine is running a mix of other WCG CPU WU's and has processed over 5 years of HCC CPU WU's so I wonder if the Kernel execution time test is needing either some refinement or a higher abort limit? [Edit 1 times, last edit by widdershins at Dec 5, 2012 10:39:21 PM] |
||
|
|
pvh513
Senior Cruncher Joined: Feb 26, 2011 Post Count: 260 Status: Offline Project Badges: 
|
What I noticed is that if a GPU WU is suspended, it starts completely from scratch on resume. This was the nVidia client with "leave applications in memory while suspended" enabled (but I think that makes no difference for GPU WUs...). This is not the behavior I see with other GPU projects. I noticed this too late to test it on AMD, so I don't know if it happens there too.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello pvh513,
GPU operations cannot be suspended or halted. If you stop to think, the board would have to duplicate the memory to hold the memory state plus some additional hardware and memory to store the state of the devices on the GPU board. If the GPU were suspended a second time, it would need all that doubled. In general, if you run N jobs and suspend N-1 jobs in the middle, you would need N-1 special GPU memory banks to store the intermediate work. (Or very slow transfer to CPU main memory, which might be slower than rerunning the job from start.) So GPUs always start at the beginning after an interruption. The work unit contains 2 jobs so it checkpoints after the first job. If it is suspended during the second job it redoes the second job but does not repeat the first. Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Two work units for me and all marked as invalid.
I use an ATI 6370m on Debian wheezy with the fgrlx driver. The desktop is unusable during gpu work. |
||
|
|
|