| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 9
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I use BOINC on ubuntu 12.04 on a laptop equipped with NVIDIA GT630M(with Optimus) and Intel HD4000 graphics.
The problem is, all the HCC GPU workunits results in "computation error". In fact, the error is only caused by unknown version of NVIDIA driver. In the event log, I saw this line, NVIDIA GPU 0: GeForce GT 630M (driver version unknown, CUDA version 4.20, compute capability 2.1, 2048MB, 2017MB available, 182 GFLOPS peak And from the result log, I see this: (notice the line with "...no version information available ...") <core_client_version>7.0.27</core_client_version> What I have tried is updating BOINC to the latest development release and also the CUDA driver as well as the bumblebee. Still, it doesn't work. Is there a way for me to tell the HCC workunits the NVIDIA version instead of automatic detection? I hope the HCC project team can help me with this... |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi,
----------------------------------------We meet again [same subject as the thread on the BOINC Dev forum]. The suspicion is that since the GT630 is blacklisted, that it also applies to the GT630M... no good work returned on a frequent basis. Note that the HCC-GPU does *not* use CUDA, but is coded in OpenCL (not OpenGL), so if your startup log does not list a driver that is OpenCL 1.1 capable, it's a no go. intel_gpu [which is in your CPU] are only starting to be coded as recognized by the latest BOINC, but no projects that I know of have been ported to this APU platform. It will come, but think that's a longer future. As for drivers in Ubuntu, stick to the official ones you find in the Additional Driver app [Think it was moved software sources in the Ubuntu Software Centre as of 12.04 or 12.10]. Doubt Optimus/Bumblebee works [it did not for me under Linux for HCC] Sorry. Edit: P.S. Can you please post the actual startup message log, first 35 lines. This gives us a chance to read the exact diagnostics. [Edit 1 times, last edit by Former Member at Mar 10, 2013 7:59:36 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi,
----------------------------------------Wow, quite surprising that you are also a member of WCG... haha... Actually the GT630M supports OpenCL 1.1, as I searched on google, so I think my GPU will be capable of running the tasks. Here's the first few lines of the event log: Without optirun (bumblebee):
With optirun:
hey, wait, I think the Nvidia driver is successfully recognized in the OpenCL entry that was not existing before! [Edit 1 times, last edit by Former Member at Mar 10, 2013 8:30:22 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
This particular computer has finished its daily quota of 1 task because of all the computation errors submitted by the HCC GPU tasks. Well, now I can't do anything but waiting. I will inform you about the progress when it grabbed any of the GPU WUs later on.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Exactly my point. First stick to the standard drivers and get things to work, then if you start fooling with bumblebee and the like, what I did, things go belly up. Now see if the server sends work and if it crunches.
TTYL |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
In the meanwhile, I would like to ask another question about the boinc daemon: How do I modify the /etc/init.d/boinc-client script so that it is run with the Bumblebee's "optirun" command? As you can see, I run the boinc with "sudo -u boinc optirun boinc" so that the boinc application runs under the user "boinc". I know this is a really ugly method but I can't think of other methods....
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
If you delete all the control files such as client_state.xml, account_wcg...xml, after detaching from WCG, the server on re-attach may be creating a new ID, so you get like 5 WU's to play with.
Modding boinc-client script...only done that to change the data_dir location. Maybe another Linux GPU crunching expert knows. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Oh my goodness, not sure what I have done but the OpenCL is gone from the event log! It was running a GPU task happily but now I am not able to continue running it anymore!
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Can anyone tell me how does the BOINC detect the OpenCL libraries, what is the environment variables that I need to set, etc...?
|
||
|
|
|