...and as you know from my previous posts, i have a nVidia GT 220 (CUDA enabled), which is working properly with the 6.10.18 version in Windows (it was also working fine in 6.10.17)

what can b the cause?

FYI, i use Ubuntu 9.10 and the BOINC manager is from http://www.getdeb.net/updates/Ubuntu/9.10/?q=boinc

I read somewhere recently on this site that BOINC 'detects' the presence of suitable gpus by looking for relevant dll's. (Of the drivers, perhaps?)

Are these installed, and in a place that BOINC can 'see'?

err.. well, yeah, drivers are properly installed (as noticed with the glxinfo command, and as i can set dektop-effects and run 3D games :) )

i dont know where do i have to look for finding where BOINC looks for drivers....

It's not a right way. :-)

My advice is to add a following thing in "/etc/init.d/boinc-client" script at first:

sudo nano /etc/init.d/boinc-client

***scrolling down***

case "$1" in
  start)
    sleep 4
    start
    ;;

sleep 4 means that boinc-client will make a pause for 4 seconds before it's start (because it can simply initialize before normal graphics subsystem initialisation. It's confirmed that this methoda does work for Ubuntu/Kubuntu 9.10 amd64 when no any other ways doesn't. It may be 3 or 5 seconds, but on my PC is working reliable with value 4.

In case of unsuccessful implementation of decribed above hack you should not change it back, but next your step is to add a boinc user to video group:

sudo usermod -a -G video boinc

It should then work (may be after rebooting system, (auto)ligin under your (or another necessary) account and autostartup of boinc). If not - try to add your account in video group too. :-)

My big question is how to make BOINC 6.10.17 (the system was openSUSE 11.2 (KDE, if any) x64) detect any "usable GPUs" at all. I did not found any trick or hack to make it, and BOINC itself, of course, won't.

Please somebody help me with openSUSE!

Rising this thread relating to the same but unsolved problem with Linux Mint 10 x64 KDE...
p.s. added xhost local:boinc &> /dev/null, evrthng is Ok now

I have BOINC client version 6.10.58 installed, but it reports "No usable GPUs found":

Чт. 30 июня 2011 12:18:42 Starting BOINC client version 6.10.58 for i686-pc-linux-gnu
Чт. 30 июня 2011 12:18:42 log flags: file_xfer, sched_ops, task
Чт. 30 июня 2011 12:18:42 Libraries: libcurl/7.21.3 NSS/3.12.10.0 zlib/1.2.5 libidn/1.19 libssh2/1.2.7
Чт. 30 июня 2011 12:18:42 Data directory: /var/lib/boinc
Чт. 30 июня 2011 12:18:42 Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q8400 @ 2.66GHz [Family 6 Model 23 Stepping 10]
Чт. 30 июня 2011 12:18:42 Processor: 2.00 MB cache
Чт. 30 июня 2011 12:18:42 Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe nx lm constant_tsc arch_perfmon pebs bts aperfmperf pni dtes64 monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr pdcm sse4
Чт. 30 июня 2011 12:18:42 OS: Linux: 2.6.38.8-32.fc15.i686.PAE
Чт. 30 июня 2011 12:18:42 Memory: 3.93 GB physical, 4.00 GB virtual
Чт. 30 июня 2011 12:18:42 Disk: 167.16 GB total, 149.20 GB free
Чт. 30 июня 2011 12:18:42 Local time is UTC +4 hours
Чт. 30 июня 2011 12:18:42 No usable GPUs found

Driver version is 275.09.07. Video card is Nvidia GTX460.
I tried to restart boinc-client in case it initialize before normal graphics subsystem initialisation, but there is no change in GPU detection.
Maybe its because of Fedora 15?

You would expect the package maintainers to know all this and take it into account... ;-)

---

I tried "xhost fix" but something wrong with it:

[pupkin@*** ~]$ xhost local:boinc
non-network local connections being added to access control list
[pupkin@*** ~]$ xhost
access control enabled, only authorized clients can connect
LOCAL:
SI:localuser:pupkin

It says that "non-network local connections being added to access control list" but there is no new connections in xhost list. The same with root privileges. I will try to reinstall it.

I tried "xhost +SI:localuser:boinc" and it succesfully added boinc user to list. After that i restarted boinc-client, but manager still reports "No usable GPUs found".

Something wrong with sh installer too :\ run_client reports what boinc already running but there is no processes with "boinc" in the name.

Maybe i should install boinc from repository again and work without GPU :(

Also, I think you need to use "boinc-user" or maybe "boinc_user" instead of "boinc" in the above xhost command. On Ubuntu the user's name is "boinc" but on Fedora it's "boinc-user" or maybe "boinc_user". Doing "ps aux | grep boinc" when boinc daemon is running will reveal the user's name.

"ps aux | grep boinc" reports

pupkin    3271  0.0  0.0   4432   764 pts/0    S+   09:49   0:00 grep --color=auto boinc

Also there is an entry in /etc/passwd

boinc:x:492:489:BOINC client account.:/var/lib/boinc:/sbin/nologin

The sh installer works fine here. Perhaps you left the boinc daemon running after you uninstalled BOINC? I can't see any other explanation for why BOINC would already be running, unless you ran the run_client script twice. Did you try "ps aux | grep boinc" to look for a boinc process?

I used "ps -e | grep boinc" and find nothing.
But these command have different output:

[pupkin@*** ~]$ ps -e | grep boinc
[pupkin@*** ~]$ ps aux | grep boinc
pupkin    3830  0.0  0.0   4432   760 pts/0    S+   09:57   0:00 grep --color=auto boinc

That's strange for me.

Hm... i started boinc client by putting

sudo /etc/init.d/boinc-client start

in terminal and now these commands reports

[pupkin@*** ~]$ ps -e | grep boinc
 4228 ?        00:00:00 boinc_client
[pupkin@*** ~]$ ps aux | grep boinc
boinc     4228  0.1  0.2  31760 11000 ?        SN   10:03   0:00 /usr/bin/boinc_client --dir /var/lib/boinc
pupkin    4430  0.0  0.0   4432   764 pts/0    S+   10:05   0:00 grep --color=auto boinc

Maybe you need to reboot too.

If i reboot computer this entry in xhost disappear.

Hold on... are you suggesting that *I* should approach the package maintainers and convince them to grant boinc user video access as part of the BOINC install process?

Nah, I wouldn't be so mundane. :-)

But I mean, just see where BOINC checks for the CUDA library on a Mac OSX system:

#ifdef __APPLE__
    cudalib = dlopen("/usr/local/cuda/lib/libcuda.dylib", RTLD_NOW);

It would be nice if Linux could be more general like that. If only if KDE systems would put it in location A and GNOME systems would put it in location B. Some generalization isn't that bad, is it? I mean, all Linux versions use general commands such as sudo, grep, bash and possibly yum. So what's wrong with a general place for (part of) the drivers, so all software developers out there can 'hardwire' that place into their software, so things can be found without needing to have the user go from novice to all-out-Linux-guru-in-one-day, just to make it work?

---

That proves there is not another client process running so you should be able to start the client with the run_boinc script. Maybe you could try that again...

1. uninstall BOINC with the package manager
2. reboot Linux to make sure the boinc daemon isn't running
3. install BOINC using the sh installer
4. in a terminal run the run_client script

For now i try to "repair" BOINC from repository. There was no boinc processes when i tried to run boinc from sh installer.

That proves you didn't uninstall BOINC using the package manager.

Its because i installed it again :)

With BOINC installed from repositories, try adding the boinc user to the video group...

/usr/sbin/usermod -G video -a boinc


Also, with BOINC installed from repositories, the GPU will not be recognized in BOINC if the boinc daemon starts before X server starts. Try restarting the boinc daemon after Linux boots...

/sbin/service boinc-client restart

I already tried it, still "no usable GPUs".

There is some SELinux alerts, maybe its because of them?

Traceback (most recent call last):
File "/usr/bin/sealert", line 692, in <module>
run_as_dbus_service(username)
File "/usr/bin/sealert", line 112, in run_as_dbus_service
app = SEAlert(user, dbus_service.presentation_manager, watch_setroubleshootd=True)
File "/usr/bin/sealert", line 326, in __init__
from setroubleshoot.browser import BrowserApplet
File "/usr/lib/python2.7/site-packages/setroubleshoot/browser.py", line 46, in <module>
import report.io.GTKIO
ImportError: No module named GTKIO

Sorry for the intrusion, I don't want to hi-jack this thread, but it's been inactive for a bit.

I have the same problem with getting BOINC to detect my nVidia 8400 GS GPU. I've read through this and other threads and tried many things and still no success.

I've booted XP and ran the GPU-Z utility which detects my gpu. I'm running:

boinc client 6.10.58 for i686-pc-linux-gnu
32-bit kernel 2.6.38.3-pclos1.bfs
OpenGL version is 3.3.0 NVIDIA 275.09.07
X 11.002.00

First off, mine is not an X timing issue because a daemon restart does not detect the GPU.

I added video group to user boinc

using the ls -l /dev/nvidia*

I discovered that my /dev/nvidia0 and /dev/nvidiactl devices are members of root root. So while gnashing my teeth I added group root to user boinc. If this doesn't fix permissions, I don't know what will. :(

My boinc user is in fact boinc and not anything like boinc-user.

As a last ditch effort, I added boinc to xhost with:

xhost +si:local:boinc

After each configuration change, I restarted the boinc-client daemon and still "No usable GPUs found."

What am I missing? Any ideas?

Thank you for your reply.

Yes, I just rebooted... again. Still no luck. I checked my group membership and in fact boinc is a member of both the root and video groups. Although I never specifically added boinc to the root group, the drak gui tool I used to add the group to boinc must have done it. So it looks like maybe group membership is not the issue.

My xhost setting goes away after reboot, so I did the xhost +si:local:boinc command again and got this:

# xhost +si:local:boinc
local:boinc being added to access control list
X Error of failed request: BadValue (integer parameter out of range for operation)
Major opcode of failed request: 109 (X_ChangeHosts)
Value in failed request: 0xb
Serial number of failed request: 7
Current serial number in output stream: 9

when I do the xhost command I get this:

# xhost
access control enabled, only authorized clients can connect
LOCAL:
INET:ASUS-P5KPL-CM

I really don't understand the xhost thing all that well and don't know if this could be the source of the problem. Looks like LOCAL is a member of authorized clients and I don't know if that's good enough.

Any other ideas?

I got a bit excited because I found an nvidia 275.09.07 cuda opencl library that wasn't installed. I installed that package with synaptic, rebooted, restarted the daemon and still no cigar. :(

I just found the GPU board here and am moving my problem there. Thanks