Thread 'BOINC WU on LinuxMint with NVIDIA GPU'

Message boards : Questions and problems : BOINC WU on LinuxMint with NVIDIA GPU
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · Next

AuthorMessage
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5129
United Kingdom
Message 60917 - Posted: 13 Mar 2015, 9:21:25 UTC

I doubt you'll ever get any GPU tasks from Milkyway. Their applications page indicates that their applications require support for the OpenCL programming language, and your message log implies that BOINC hasn't detected OpenCL support in your NVidia drivers. Roland might have some ideas on that, but if you are satisfied by running GPU projects with CUDA applications, I'd suggest you rest on your laurels before embarking on another round of driver updates.
ID: 60917 · Report as offensive
ProfileGandalf_the_Grey
Avatar

Send message
Joined: 28 Apr 13
Posts: 42
United States
Message 60918 - Posted: 13 Mar 2015, 10:16:33 UTC - in response to Message 60917.  
Last modified: 13 Mar 2015, 10:17:09 UTC

Richard - Sorry to bust your bubble, but Milkyway and Collatz are both downloading WUs now. I have them both set to 0.25C + 0.5NV.
The GPUs can't handle .25NV, but that is ok. They are working fine at .5NV.
With Collatz crashing perodically, I've set both Projects to download.

I won't need to mess with this PC again unless it stops crunching.
ID: 60918 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5129
United Kingdom
Message 60920 - Posted: 13 Mar 2015, 10:40:26 UTC - in response to Message 60918.  

Richard - Sorry to bust your bubble, but Milkyway and Collatz are both downloading WUs now. I have them both set to 0.25C + 0.5NV.
The GPUs can't handle .25NV, but that is ok. They are working fine at .5NV.
With Collatz crashing perodically, I've set both Projects to download.

I won't need to mess with this PC again unless it stops crunching.

Don't worry about bursting my bubble - I have grave doubts about MilkyWay's ability to comprehend and manage the plan_class facility, so it doesn't really surprise me that their 'opencl_nvidia' or 'opencl_nvidia_101' plan classes don't do what they say on the tin. I had a long discussion with them while they were trying to develop their multi-threaded CPU application (nbody), culminating in the revelation

When I started releasing the format was to use gpu classes and I kept the convention. Since the idea is to get a gpu version of nbody out I thought it was a place holder for turning on that option.

My comment was more of a warning not to be too concerned if your machine continued to suck on empty from that project - but if it works, that's great. Given the level of project support, it might have been a long slog tracking down any remaining problems.
ID: 60920 · Report as offensive
ProfileJord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15567
Netherlands
Message 60921 - Posted: 13 Mar 2015, 11:05:45 UTC - in response to Message 60920.  

I think that the lucky deduction that projects can make is: when the GPU is Nvidia and it has CUDA detected, it can automatically do OpenCL 1.0 at the minimum. All CUDA capable GPUs can do OpenCL 1.0, a few can do OpenCL 1.1, none can do any better because Nvidia stopped developing OpenCL.

As opposed to AMD and Intel GPUs, where you're required to have had some detection of capabilities.
ID: 60921 · Report as offensive
Roland Hughes

Send message
Joined: 7 Mar 12
Posts: 67
United States
Message 60923 - Posted: 13 Mar 2015, 13:44:06 UTC

Glad it is all working for you. Sometimes they just have to chew on things a while before they decide to let you have work.
ID: 60923 · Report as offensive
ProfileGandalf_the_Grey
Avatar

Send message
Joined: 28 Apr 13
Posts: 42
United States
Message 60937 - Posted: 14 Mar 2015, 0:36:48 UTC - in response to Message 60923.  

Roland - on my other Linux PC that still is running LM 64-bit, I tried your

sudo apt-get install nvidia-modprobe

The result was...
gandalf@MiddleEarth ~ $ sudo apt-get install nvidia-modprobe
[sudo] password for gandalf: 
Reading package lists... Done
Building dependency tree       
Reading state information... Done
E: Unable to locate package nvidia-modprobe


What am I missing?
ID: 60937 · Report as offensive
Roland Hughes

Send message
Joined: 7 Mar 12
Posts: 67
United States
Message 60944 - Posted: 14 Mar 2015, 11:51:56 UTC - in response to Message 60937.  

You CANNOT SKIP STEPS.

Every step from the beginning.

++++
Want to know the only steps different to make your current 64-bit system work?

After rebooting so nvidia-331 takes full effect.

Click on the little blue start button in the lower left corner.

In the little search bar at the top type "sour" and you should see an option for "Software Sources" appear. Click it.

Click the "Additional repositories" button on left.

Click the "Add a new repository" button at bottom.

paste in the following:

deb http://archive.ubuntu.com/ubuntu trusty-backports main restricted universe multiverse

Click "Ok"

Click the "update" or "refresh" button on upper right side of screen. (This will be greyed out with "No action required" until you actually add something and I don't remember what the text chagnes to.)

Exit out of all windows.

Click on the little blue start button in the lower left corner.

Applications->System->Synaptic Package Manager

Enter password when prompted.

After the index finishes updating type "modpro" in the "Quick filter" field

Mark nvidia-modprobe for installation, then click "Apply changes"

Exit everything and reboot

Open your BOINC manager.

Click Advanced->Event Log

Scroll to top of log (it always shows last entry when it opens) You should see something like:

Tue 10 Mar 2015 06:35:09 AM CDT | | CUDA: NVIDIA GPU 0: GeForce GT 630 (driver version unknown, CUDA version 6.0, compute capability 3.5, 2047MB, 1973MB available, 692 GFLOPS peak)

near the beginning.

-------------------

You skipped past the very beginning of this and did not add the repository.
ID: 60944 · Report as offensive
Roland Hughes

Send message
Joined: 7 Mar 12
Posts: 67
United States
Message 60947 - Posted: 14 Mar 2015, 12:00:31 UTC

First NVidia crash some time during wee hours of morning on 64-bit machine.

In all fairness though, I had streamed 2 episodes of "Person of Interest" AND watched 2 DVDs without shutting down BOINC. Came out this morning and saw the tell-tale signs with a flashing caps lock and scroll lock on keyboard.

Reset button time.

With regular use it appears the 331 driver with the modprobe manages to run about a week before it gives up the ghost.
ID: 60947 · Report as offensive
ProfileGandalf_the_Grey
Avatar

Send message
Joined: 28 Apr 13
Posts: 42
United States
Message 60951 - Posted: 14 Mar 2015, 15:08:58 UTC - in response to Message 60947.  

Roland - Could not find a ...

Click on the little blue start button in the lower left corner.

Was able to work around this instruction. I suspect this guide was written for other than LinuxMint. Knock on wood, it will still work. Will keep you advised.
ID: 60951 · Report as offensive
ProfileGandalf_the_Grey
Avatar

Send message
Joined: 28 Apr 13
Posts: 42
United States
Message 60952 - Posted: 14 Mar 2015, 15:29:03 UTC - in response to Message 60951.  

    Sat 14 Mar 2015 08:15:41 AM PDT | | Starting BOINC client version 7.2.42 for x86_64-pc-linux-gnu
    Sat 14 Mar 2015 08:15:41 AM PDT | | log flags: file_xfer, sched_ops, task, sched_op_debug
    Sat 14 Mar 2015 08:15:41 AM PDT | | Libraries: libcurl/7.35.0 OpenSSL/1.0.1f zlib/1.2.8 libidn/1.28 librtmp/2.3
    Sat 14 Mar 2015 08:15:41 AM PDT | | Data directory: /var/lib/boinc-client
    Sat 14 Mar 2015 08:15:41 AM PDT | | CUDA: NVIDIA GPU 0: GeForce GTX 650 (driver version unknown, CUDA version 6.0, compute capability 3.0, 1023MB, 887MB available, 813 GFLOPS peak)
    Sat 14 Mar 2015 08:15:41 AM PDT | | CUDA: NVIDIA GPU 1: GeForce GTX 650 (driver version unknown, CUDA version 6.0, compute capability 3.0, 1024MB, 1004MB available, 813 GFLOPS peak)


So far so good. Wish me luck. Will keep you advised.

ID: 60952 · Report as offensive
ProfileGandalf_the_Grey
Avatar

Send message
Joined: 28 Apr 13
Posts: 42
United States
Message 60954 - Posted: 14 Mar 2015, 19:16:08 UTC - in response to Message 60952.  

FYI - Upon running Boinc Manager, Collatz is running. Milkyway downloaded WUs, then aborted them. I have since deactivated the Milkyway Project.
ID: 60954 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 60955 - Posted: 14 Mar 2015, 20:27:46 UTC - in response to Message 60954.  

Milkyway downloaded WUs, then aborted them.


Exit status 201 (0xc9) EXIT_MISSING_COPROC


You need to get BOINC to detect OpenCL support. Run the command that matches your OS:

64-bit:

cd /usr/lib/x86_64-linux-gnu/ && sudo ln -s libOpenCL.so.1 libOpenCL.so


32-bit:

cd /usr/lib/i386-linux-gnu/ && sudo ln -s libOpenCL.so.1 libOpenCL.so


And then restart BOINC.
ID: 60955 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5129
United Kingdom
Message 60956 - Posted: 14 Mar 2015, 20:43:16 UTC - in response to Message 60955.  

The mystery, then, is why the tasks were allocated and downloaded in the first place. I've always thought that there's something strange about Milkyway's scheduler.
ID: 60956 · Report as offensive
ProfileGandalf_the_Grey
Avatar

Send message
Joined: 28 Apr 13
Posts: 42
United States
Message 60958 - Posted: 14 Mar 2015, 21:05:17 UTC - in response to Message 60955.  

This one is a 64-bit...

gandalf@MiddleEarth ~ $ cd /usr/lib/x86_64-linux-gnu/ && sudo ln -s libOpenCL.so.1 libOpenCL.so
[sudo] password for gandalf: 
ln: failed to create symbolic link ‘libOpenCL.so’: File exists
ID: 60958 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 60959 - Posted: 14 Mar 2015, 21:16:03 UTC - in response to Message 60956.  

Your previous post did have a very plausible sounding explanation. Milkyway's opencl_nvidia* plan classes could be set to only check for CUDA but the app_version may still require OpenCL coproc. Too bad we can't see how they, or any other project, have set up the plan classes.
ID: 60959 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 60960 - Posted: 14 Mar 2015, 21:22:19 UTC - in response to Message 60958.  

This one is a 64-bit...

gandalf@MiddleEarth ~ $ cd /usr/lib/x86_64-linux-gnu/ && sudo ln -s libOpenCL.so.1 libOpenCL.so
[sudo] password for gandalf: 
ln: failed to create symbolic link ‘libOpenCL.so’: File exists

And BOINC still doesn't detect OpenCL? That's new.

Enable <coproc_debug> log flag in cc_config.xml (in /var/lib/boinc-client), restart BOINC and then post the start-up messages.
ID: 60960 · Report as offensive
Roland Hughes

Send message
Joined: 7 Mar 12
Posts: 67
United States
Message 60964 - Posted: 14 Mar 2015, 21:34:58 UTC - in response to Message 60960.  

You may actually have to install that Boinc_app_milkyway package you had before. I don't subscribe to this project so I don't know.

I do humbly suggest you open a question on the forum/support board for the milkyway project asking them about getting things working with NVidia under 64-bit Mint. You have other projects pulling down WUs and completing so not the issue is specific to that project.

Btw, my instructions were all for Mint 17 KDE. I make it a point not to run any of the Gicky-Gnasty-Gnome desktops unless I'm forced at gunpoint <Grin>
ID: 60964 · Report as offensive
Roland Hughes

Send message
Joined: 7 Mar 12
Posts: 67
United States
Message 61010 - Posted: 17 Mar 2015, 10:34:58 UTC

Gandalf,

If you want to be able to go away for a long weekend and believe your systems are happily processing WU, you probably need to drop them all back to 32-bit. I just had yet another NVidia crash this morning on my 64-bit machine.
ID: 61010 · Report as offensive
Roland Hughes

Send message
Joined: 7 Mar 12
Posts: 67
United States
Message 61102 - Posted: 20 Mar 2015, 13:53:15 UTC

@Gandalf,

Have you solved the problem? I have been having a rash of Nvidia crashes. In a few days when the disks show up I'm going to experiment with different distros. The YAU stuff doesn't seem to be testing 64-bit again.

Will be looking at:

Chakra http://chakraos.org/forum/

RoboLinux http://www.robolinux.org/about/

Linux Lite https://www.linuxliteos.com/

I really don't want to give up my RAM and drop back to 32-bit on this machine.
ID: 61102 · Report as offensive
ProfileGandalf_the_Grey
Avatar

Send message
Joined: 28 Apr 13
Posts: 42
United States
Message 61107 - Posted: 20 Mar 2015, 15:29:29 UTC - in response to Message 61102.  

I've decided to stay with Collatz and increase the Minimum work buffer to 10.0 days and the Max additional work buffer to 4.00 days.
I'm hoping that when the Collatz crashes again that this setting will provide enough work until the Callatz comes back on line.
Time will tell.

My 32-bit PC froze and T/S showed a fan on the heat sink had dies. Have left the PC powered down until I can replace the fan.
ID: 61107 · Report as offensive
Previous · 1 · 2 · 3 · 4 · Next

Message boards : Questions and problems : BOINC WU on LinuxMint with NVIDIA GPU

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.