| 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 |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
In answer to the question about my preferences, I have "Run based on preferences" and "Use GPU based on preferences" selected, but the betas still used the GPU when the computer was in use.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
In answer to the question about my preferences, I have "Run based on preferences" and "Use GPU based on preferences" selected, but the betas still used the GPU when the computer was in use. Of course, in the processor tab of the preferences the box "Use GPU while computer is in use" is *not* ticked. If those 2 conditions are met, then there is *zero* reason why GPU tasks would be running when you are using the computer. There are sporadic reports of boinc not detecting that there is user input, to emphasize that it's not the science doing that, but the core client [traffic manager]. If it happens to the WCG-GPU task, it happens to all tasks. |
||
|
|
pvh513
Senior Cruncher Joined: Feb 26, 2011 Post Count: 260 Status: Offline Project Badges: 
|
@lawrencehardin
The fact that the GPU has its own memory is certainly a complicating factor (which is why I said that keeping tasks in memory while suspended likely has no effect on GPU tasks). But it is certainly not true that they always need to restart from scratch. They need to restart from the last checkpoint. So it looks like no or very few checkpoints are taken by this client. This is definitely different from other GPU projects I have been running and I think this is also different from the CPU HCC client. This could become an issue on slower GPUs when jobs get interrupted frequently by user activity. That is what I wanted to state here. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Actually, Lawrence did mention that half way there is a checkpoint.
Let's not confuse "client" (which is the boinc core), and the science app [outside of WCG, don't care if a spade is called a spade]. The science.app has programmed checkpoints and asks the client (boinc) "Can I do a save now or not". The client then based on preferences say yes or no, ask me later. Some science.apps are not programmed to ask, such as CEP2. They just do. There's a maximum of 16 in 12 hours, so no one is likely to object. Whether HCC-GPU asks, no idea, but suspect not, since there is processing combined of 2 discreet images, which is why there's is the logical midpoint checkpoint save. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Are there still beta-WUs in the pipe?
; ; andgridPost#721 ; |
||
|
|
[AF>Libristes]Maeda
Cruncher Joined: Sep 1, 2011 Post Count: 43 Status: Offline Project Badges: 
|
Hi !
It seems that I receive task for HCC nvidia (under Linux) too. But it seems that it's not taking care of the cc_config.xml (local) I have, and setting the first GPU not to be used (GPU #0), but the #1. HCC is always taking the #0 when it's activated. Not very 'system user-friendly'. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello [AF>Libristes]Maeda,
What instructions are you using in cc_config.xml? What GPUs are in your system? Lawrence |
||
|
|
[AF>Libristes]Maeda
Cruncher Joined: Sep 1, 2011 Post Count: 43 Status: Offline Project Badges:
|
Hi Lawrence,
I use this two options : the first for computing with both gpu (of course), and the second one, when I need the first for another graphical apps (gpu#0 not used then) : <use_all_gpus>1</use_all_gpus> Here are my gpus (here, in case of not using the #0, and HCC use it) : dim. 09 déc. 2012 23:09:33 CET | | NVIDIA GPU 0 (ignored by config): GeForce GT 430 (driver version unknown, CUDA version 4020, compute capability 2.1, 1023MB, 179 GFLOPS peak) |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi Lawrence, I use this two options : the first for computing with both gpu (of course), and the second one, when I need the first for another graphical apps (gpu#0 not used then) : <use_all_gpus>1</use_all_gpus> Here are my gpus (here, in case of not using the #0, and HCC use it) : dim. 09 déc. 2012 23:09:33 CET | | NVIDIA GPU 0 (ignored by config): GeForce GT 430 (driver version unknown, CUDA version 4020, compute capability 2.1, 1023MB, 179 GFLOPS peak) Per the manual, limited use to specific BOINC versions: <ignore_cuda_dev>N</ignore_cuda_dev> ignore (don't use) a specific NVIDIA GPU. You can ignore more than one. New in 6.10.19 and only to be used till 6.12.41. <ignore_nvidia_dev>N</ignore_nvidia_dev> ignore (don't use) a specific NVIDIA GPU. You can ignore more than one. Replaces <ignore_cuda_dev/> List-add.pngNew in 6.13.0 Reminder: Zero will disable device zero, it does not disable the line, only removal does. <ignore_nvidia_dev>1</ignore_nvidia_dev> will ignore the use of the second Nvidia GPU in the system. Before 6.13.0, <ignore_cuda_dev/> was used. <ignore_nvidia_dev> was added to get it more in line with the specific <ignore_ati_dev>. Edit: I've got a GT 220 and it serially fails HCC tasks with Kernel time exceed, so disabled the profile for that device to not send HCC-GPU. The log file should have a 3rd line reporting on the OpenCL version. Has to be 1.1, mine shows 1.0. For HCC-GPU the CUDA is irrelevant. [Edit 2 times, last edit by Former Member at Dec 9, 2012 11:04:47 PM] |
||
|
|
[AF>Libristes]Maeda
Cruncher Joined: Sep 1, 2011 Post Count: 43 Status: Offline Project Badges:
|
Hi SekeRob !
I'm using 6.12.34 version of Boinc, then I can use this option imo. |
||
|
|
|