Advanced search

Message boards : Graphics cards (GPUs) : GPU missing

Author Message
Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41868 - Posted: 22 Sep 2015 | 6:31:09 UTC

GPU is missing from the hosts_user page. How do I fix this? WUs were working until I upgraded from 17.1 to 17.2. O/S sees the GPUs, but your web page does not.

Web page ...



PC configuration ...

gandalf@MiddleEarth ~ $ sudo inxi -Fxz
System: Host: MiddleEarth Kernel: 3.13.0-37-generic x86_64 (64 bit, gcc: 4.8.2)
Desktop: Cinnamon 2.6.13 Distro: Linux Mint 17.2 Rafaela
Machine: Mobo: ECS model: X58B-A3 SLI version: 1.X Bios: American Megatrends version: 080015 date: 05/04/2010
CPU: Quad core Intel Core i7 CPU 975 (-HT-MCP-) cache: 8192 KB flags: (lm nx sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx) bmips: 26664.7
Clock Speeds: 1: 3334.00 MHz 2: 3334.00 MHz 3: 3334.00 MHz 4: 3334.00 MHz 5: 3334.00 MHz 6: 3334.00 MHz 7: 2800.00 MHz 8: 3334.00 MHz
Graphics: Card-1: NVIDIA GK104 [GeForce GTX 680] bus-ID: 04:00.0
Card-2: NVIDIA GK107 [GeForce GTX 650] bus-ID: 05:00.0
X.org: 1.15.1 drivers: nouveau (unloaded: fbdev,vesa) tty size: 153x24 Advanced Data: N/A for root
Audio: Card-1: NVIDIA GK107 HDMI Audio Controller driver: snd_hda_intel bus-ID: 05:00.1
Card-2: NVIDIA GK104 HDMI Audio Controller driver: snd_hda_intel bus-ID: 04:00.1
Card-3: Intel 82801JI (ICH10 Family) HD Audio Controller driver: snd_hda_intel bus-ID: 00:1b.0
Sound: Advanced Linux Sound Architecture ver: k3.13.0-37-generic
Network: Card: Realtek RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller
driver: r8169 ver: 2.3LK-NAPI port: c800 bus-ID: 03:00.0
IF: eth0 state: up speed: 100 Mbps duplex: full mac: <filter>
Drives: HDD Total Size: 1000.2GB (0.8% used) 1: id: /dev/sda model: WDC_WD1001FALS size: 1000.2GB temp: 57C
Partition: ID: / size: 905G used: 7.5G (1%) fs: ext4 ID: swap-1 size: 12.87GB used: 0.00GB (0%) fs: swap
RAID: No RAID devices detected - /proc/mdstat and md_mod kernel raid module present
Sensors: System Temperatures: cpu: 80.0C mobo: N/A gpu: 43.0,50.0
Fan Speeds (in rpm): cpu: N/A
Info: Processes: 228 Uptime: 18 min Memory: 1213.2/12007.8MB Runlevel: 2 Gcc sys: 4.8.4
Client: Shell (sudo) inxi: 1.9.17


____________

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2058
Credit: 15,019,398,669
RAC: 4,416,896
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41869 - Posted: 22 Sep 2015 | 8:30:37 UTC - in response to Message 41868.

The web page shows what the BOINC manager reports, so the BOINC manager can't detect your GPU. Check the first 30 lines of the BOINC manager's log at startup. I'm not a Linux expert, but perhaps you should update the BOINC manager to 7.6.9, and the NVidia drivers.

captainjack
Send message
Joined: 9 May 13
Posts: 157
Credit: 1,191,077,233
RAC: 297,364
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 41871 - Posted: 22 Sep 2015 | 13:13:14 UTC

Highlander,

Which version of BOINC are you using and how did you install it? Did you use the software installer that comes with Linux Mint or did you install it manually using the BOINC software downloaded from the Berkeley web site?

Sometimes, when you use the BOINC software that is installed from the Linux Mint repositories, BOINC will start automatically as soon as the system starts up. When this happens, BOINC starts up before the video drivers finish loading completely and BOINC thinks that there are no usable GPU's. You can test to see if this is the case by restarting BOINC and seeing if it detects the video drivers after a restart. If this is the case, you need to put a delay in the BOINC startup script so that the video drivers will have time to load before BOINC starts up.

If that is not the case, also tell us how you installed the video drivers. Did you use the video drivers that came with Mint or did you install them manually using drivers downloaded from the NVIDIA web site?

Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41876 - Posted: 22 Sep 2015 | 20:33:30 UTC
Last modified: 22 Sep 2015 | 20:36:13 UTC

Gentlemen...thanks for the replies.

I just tried...

sudo service boinc-client restart

No joy.

How do I implement the delay in the BOINC startup script.
What code and what filename?

I installed Boinc via Synaptic Package Manager.



If I need to install other NVidia drivers, please provide the URL.
Card-1: NVIDIA GK104 [GeForce GTX 680] bus-ID: 04:00.0
Card-2: NVIDIA GK107 [GeForce GTX 650] bus-ID: 05:00.0

captainjack
Send message
Joined: 9 May 13
Posts: 157
Credit: 1,191,077,233
RAC: 297,364
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 41878 - Posted: 22 Sep 2015 | 21:44:05 UTC

Highlander,

I always install the NVIDIA drivers manually using the instructions posted here:

https://www.gpugrid.net/forum_thread.php?id=3713&nowrap=true#36671

If you scroll on down from the referenced post, there are some other helpful suggestions from other contributors. These instructions are for Ubuntu so you might need some adaptations for Mint.

I get the drivers at http://www.nvidia.com/download/index.aspx

To put the startup delay in BOINC

Edit /etc/init.d/boinc-client and insert "sleep 6" at the following location

start()
{
log_begin_msg "Starting $DESC: $NAME"
if is_running; then
log_progress_msg "already running"
else
if [ -n "$DISPLAY" -a -x /usr/bin/xhost ]; then
# grant the boinc client to perform GPU computing
xhost local:boinc || echo -n "xhost error ignored, GPU computing may not be possible"
fi
sleep 6
if [ -n "$VALGRIND_OPTIONS" ]; then
start-stop-daemon --start --quiet --background --pidfile $PIDFILE \
--make-pidfile --user $BOINC_USER --chuid $BOINC_USER \
--chdir $BOINC_DIR --exec /usr/bin/valgrind -- $VALGRIND_OPTIONS $BOINC_CLIENT $BOINC_OPTS
else
start-stop-daemon --start --quiet --background --pidfile $PIDFILE \
--make-pidfile --user $BOINC_USER --chuid $BOINC_USER \
--chdir $BOINC_DIR --exec $BOINC_CLIENT -- $BOINC_OPTS
fi


The "sleep 6" command will cause the startup script to pause for 6 seconds before loading the BOINC software. Again, these instructions are for Ubuntu so they might need some adaptation for Mint.

Hope that helps, if you continue to have issues, let us know and we can try something else.

Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41879 - Posted: 22 Sep 2015 | 22:51:59 UTC
Last modified: 22 Sep 2015 | 23:00:32 UTC

I downloaded...
http://www.nvidia.com/download/driverResults.aspx/90399/en-us

Linux x64 (AMD64/EM64T) Display Driver

Version: 304.128
Release Date: 2015.8.31
Operating System: Linux 64-bit
CUDA Toolkit: 5.0
Language: English (US)
File Size: 65.89 MB

File... NVIDIA-Linux-x86_64-304.128.run

I tried to install this via the Terminal, but as I figured, I don't know how.
HELP!

Is this OK? It's the code as it came.


start()
{
sleep 10
log_begin_msg "Starting $DESC: $NAME"
if is_running; then
log_progress_msg "already running"
else
if [ -n "$DISPLAY" -a -x /usr/bin/xhost ]; then
# grant the boinc client to perform GPU computing
xhost local:boinc || echo -n "xhost error ignored, GPU computing may not be possible"
fi
if [ -n "$VALGRIND_OPTIONS" ]; then
start-stop-daemon --start --quiet --background --pidfile $PIDFILE \
--make-pidfile --user $BOINC_USER --chuid $BOINC_USER \
--chdir $BOINC_DIR --exec /usr/bin/valgrind -- $VALGRIND_OPTIONS $BOINC_CLIENT $BOINC_OPTS
else
start-stop-daemon --start --quiet --background --pidfile $PIDFILE \
--make-pidfile --user $BOINC_USER --chuid $BOINC_USER \
--chdir $BOINC_DIR --exec $BOINC_CLIENT -- $BOINC_OPTS
fi
fi
log_end_msg 0

if [ "$SCHEDULE" = "1" ]; then
schedule
fi
}

Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41880 - Posted: 23 Sep 2015 | 1:24:29 UTC
Last modified: 23 Sep 2015 | 1:57:14 UTC

I found I needed to preface the filename with "sh".
However the resulting run produced a window saying ...

ERROR: You appear to be running an X server; please exit X before
installing. For further details, please see the section INSTALLING
THE NVIDIA DRIVER in the README available on the Linux driver
download page at www.nvidia.com.

I found http://us.download.nvidia.com/XFree86/Linux-x86_64/304.128/README/installdriver.html
Here I ran into a wall. It says "exit the X server and terminate all OpenGL applications". That is as clear as mud.

As for the Boinc Manager, I just checked. The Linux version for manual downloading is the same version as I downloaded via Synaptic. Let's hope the driver produces some good results as soon as I can get directions for exiting X.

captainjack
Send message
Joined: 9 May 13
Posts: 157
Credit: 1,191,077,233
RAC: 297,364
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 41881 - Posted: 23 Sep 2015 | 1:58:22 UTC

Highlander,

When it is telling you to exit the X server, it is saying to shutdown the desktop environment and go into a command line (terminal session) mode. For Linux Mint, it has the options of the default Cinnamon desktop, MATE, KDE or Xfce. The command you would enter in a terminal session to exit the X server is something like "sudo service KDE stop", where KDE would be replaced with the desktop option you chose (1 of the 4 listed above) when you installed Linux Mint.

I don't know anything about Mint, but it does seem like your notes about configuration showed that you were using Cinnamon. If so, your command should be "sudo service Cinnamon stop" in a terminal session.

When you do get your desktop environment shut down, you should see a blank black screen. If/when that happens press and hold the 'ctrl' and 'alt' keys then press the f1 key. That should bring up a prompt where you can log in to a command line session.

After you get to the command line session, you should be able to run the NVIDIA installer.

All of my Linux experience is with Ubuntu so I'm guessing about a few things here with your Mint setup, but maybe that will help. Let us know how it turns out.

Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41883 - Posted: 23 Sep 2015 | 4:19:43 UTC

Couldn't do anything with the EXIT X.
I did go to the Driver Manager and selected NVIDIA 304.
Rebooted and the Web page now shows the GPU data. :~)

However...
I have these apps selected...
ACEMD short runs (2-3 hours on fastest card): yes
ACEMD long runs (8-12 hours on fastest GPU): yes

I removed the GPUGrid project from the Boinc Manager, then re-added the Project.
Figured it couldn't hurt. However, no WUs have been downloaded.
I don't know the URL to see if the servers are up and running.

Any ideas as to why the WUs are not being downloaded? I've tried Updating, but no joy.


Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2058
Credit: 15,019,398,669
RAC: 4,416,896
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41885 - Posted: 23 Sep 2015 | 7:44:10 UTC - in response to Message 41883.
Last modified: 23 Sep 2015 | 7:51:33 UTC

Any ideas as to why the WUs are not being downloaded? I've tried Updating, but no joy.

That's because there are 0 unsent workunits in both queues. See the server status page.

Check the event log of you BOINC manager. You should see something like this:
23/09/2015 00:35:49 | GPUGRID | Sending scheduler request: To fetch work. 23/09/2015 00:35:49 | GPUGRID | Requesting new tasks for NVIDIA GPU 23/09/2015 00:35:53 | GPUGRID | Scheduler request completed: got 0 new tasks 23/09/2015 00:35:53 | GPUGRID | No tasks sent 23/09/2015 00:35:53 | GPUGRID | No tasks are available for Long runs (8-12 hours on fastest card) 23/09/2015 00:35:53 | GPUGRID | Project has no tasks available

Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41886 - Posted: 23 Sep 2015 | 7:50:02 UTC

Thanks Zoltan

When the WUs return, I post if I don't get any.
I'm hoping we have fixed this issue.

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2058
Credit: 15,019,398,669
RAC: 4,416,896
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41887 - Posted: 23 Sep 2015 | 7:57:11 UTC - in response to Message 41886.

I'm hoping we have fixed this issue.

I think so.

When the WUs return, I post if I don't get any.

Until then, you can add another GPU project with 0 resource share (you can set it on the project's homepage). The 0 resource share make the BOINC manager to download tasks only when there are no tasks available from projects with above 0 resource share.
If it pulls work from these 'backup' projects, we can be sure that your issue is fixed.

Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41888 - Posted: 23 Sep 2015 | 8:03:04 UTC

Thanks. That never occurred to me. I got Milkyway WUs running now. Setting its resource share to 0 makes sense now that I think of it. Thanks again.

Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41890 - Posted: 23 Sep 2015 | 8:15:55 UTC

If I understand this right, when the GPUGrid WUs hit 0 again or if the server dies, then Milkyway will automatically kick in until GPUGrid resurfaces at which time Milkyway go back to sleep so the speak.

Great idea. I can use this idea with my Win7 (AMD GPU) PC. The Collatz server is always dying.

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2058
Credit: 15,019,398,669
RAC: 4,416,896
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41892 - Posted: 23 Sep 2015 | 8:26:33 UTC - in response to Message 41890.
Last modified: 23 Sep 2015 | 8:27:12 UTC

If I understand this right, when the GPUGrid WUs hit 0 again or if the server dies, then Milkyway will automatically kick in until GPUGrid resurfaces at which time Milkyway go back to sleep so the speak.

That's right. It will ask your main project's servers for work every time it finishes a workunit from your backup project, and never pull down more than one workunit (per GPU) from the backup project.

I can use this idea with my Win7 (AMD GPU) PC. The Collatz server is always dying.

It's independent of OS, and/or project. The only limitation is that you have to use a fairly recent BOINC manager (I don't remember the exact version number).

Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41916 - Posted: 29 Sep 2015 | 7:02:55 UTC

I see the ACEMD short and ACEMD long runs now have unsent WUs available, however I can not get any downloaded to my Linuxmint PC.

Help is much appreciated so solve the predicament.

Profile Highlander_6596
Send message
Joined: 11 May 14
Posts: 11
Credit: 13,979,550
RAC: 349
Level
Pro
Scientific publications
wat
Message 41937 - Posted: 4 Oct 2015 | 7:54:02 UTC
Last modified: 4 Oct 2015 | 7:54:31 UTC

I've decided to run Collatz Conjecture WUs. Their GPU WUs work.

Post to thread

Message boards : Graphics cards (GPUs) : GPU missing