Advanced search

Message boards : Graphics cards (GPUs) : Linux - Nvidia Driver 319.17 certificate

Author Message
Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 324
Credit: 72,394,453
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30010 - Posted: 16 May 2013 | 14:27:18 UTC

Hi: Available the new Nvidia driver is from May 2nd, with many new features and updates.

Installed and works perfectly, if yields more still do not know, we'll see.

Profile skgiven
Volunteer moderator
Project tester
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 8,281
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30015 - Posted: 16 May 2013 | 17:10:44 UTC - in response to Message 30010.

What version of Ubuntu did it work on?
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 324
Credit: 72,394,453
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30022 - Posted: 16 May 2013 | 17:56:38 UTC - in response to Message 30015.

Hello: I have installed Ubuntu 13.04 - 64 bits - Kernel 3.9 - GTX 590 and 6 GB Ram with FX 8150.

fractal
Send message
Joined: 16 Aug 08
Posts: 87
Credit: 1,182,558,826
RAC: 186,737
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30096 - Posted: 19 May 2013 | 23:57:22 UTC

I just replaced at gts450 with a gtx650ti in an ubuntu 12.0.4 LTS box and the old driver did not support the new card. I upgraded to 319.17 and it is crunching. I will won't be able to tell whether it is any faster or slower than the old one as I changed both the driver and the card at the same time. But, 319.17 installed just fine on a 12.0.4 LTS server box.

fractal
Send message
Joined: 16 Aug 08
Posts: 87
Credit: 1,182,558,826
RAC: 186,737
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30181 - Posted: 22 May 2013 | 1:45:41 UTC

Things are not looking so good. Work unit started off wanting 40 hrs on a gtx650ti. 24 hrs later it wanted 50 hrs more so I dedicated a core to it. Now with 49 hrs elapsed it wants 56 hrs to complete.

I don't know if there is an issue with the card or the driver but this system used to complete units in just under 24 hrs with a gts450 with an older driver.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1104
Credit: 6,101,732,079
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30182 - Posted: 22 May 2013 | 2:52:04 UTC - in response to Message 30181.

Things are not looking so good. Work unit started off wanting 40 hrs on a gtx650ti. 24 hrs later it wanted 50 hrs more so I dedicated a core to it. Now with 49 hrs elapsed it wants 56 hrs to complete.

I don't know if there is an issue with the card or the driver but this system used to complete units in just under 24 hrs with a gts450 with an older driver.

Hi fractal, is it a new card? Any possibility of popping it in a windows box to run a few WUs and verify the card is good. Are other projects running correctly? Sorry but can't be of much help with Linux. The NOELIA WUs (like the one you're currently running) have been taking about 17-18 hours on my OCed 650 Ti cards. AFAIK Linux should be about 10-12% faster than W7 here, but harder to get working for many.

matlock
Send message
Joined: 12 Dec 11
Posts: 34
Credit: 86,423,547
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 30185 - Posted: 22 May 2013 | 5:20:04 UTC - in response to Message 30181.

fractal,

In your nvidia-settings, what does your 'performance level' say under PowerMizer? Also, what are your temps looking like at load?

Unfortunately Nvidia doesn't provide profiling in Linux anymore for the GeForce series (need Quadro, Tesla) when using the command:


nvidia-smi -a

Most of the fields output N/A.

I've had a lot of success using Nvidia cards in Linux over the years. Instead of experimenting with WUs right now, maybe try a benchmarking tool like Unigine's Heaven 4.0 (or Valley 1.0). Download their free Linux build and run the benchmark on extreme settings.

Profile skgiven
Volunteer moderator
Project tester
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 8,281
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30192 - Posted: 22 May 2013 | 9:52:31 UTC - in response to Message 30185.

If the GPU's utilization isn't high enough, it will downclock (possibly to ~550Mhz, but perhaps lower). It's often the case that this is related to overuse of the CPU by CPU apps. The GeForce 600 series GPU's need a full CPU core/thread to support the GPUGrid app.

I can see you crunch for a few CPU projects, but I don't know your settings. Tell Boinc to use less of the CPU (75% to use 6 threads for CPU work, or 76% to 99% to use 7 threads), and see how you get on.

Some GeForce 600 series GPU's will downclock slightly if they are too hot (~70°C), but this is a Boost function and tends to be done in increments of 13MHz, so it wouldn't knobble the WU's that much, and the GTX650Ti doesn't have Boost.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

fractal
Send message
Joined: 16 Aug 08
Posts: 87
Credit: 1,182,558,826
RAC: 186,737
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30210 - Posted: 22 May 2013 | 17:02:05 UTC - in response to Message 30185.

Yeah, nvidia-smi doesn't say much

==============NVSMI LOG==============

Timestamp : Wed May 22 09:54:35 2013
Driver Version : 319.17

Attached GPUs : 1
GPU 0000:01:00.0
Product Name : GeForce GTX 650 Ti
Display Mode : N/A
Display Active : N/A
Persistence Mode : Disabled
Accounting Mode : N/A
Accounting Mode Buffer Size : N/A
Driver Model
Current : N/A
Pending : N/A
Serial Number : N/A
GPU UUID : GPU-27d68e8c-fc15-1cc8-a08c-8e1a9831d643
VBIOS Version : 80.06.3C.00.50
Inforom Version
Image Version : N/A
OEM Object : N/A
ECC Object : N/A
Power Management Object : N/A
GPU Operation Mode
Current : N/A
Pending : N/A
PCI
Bus : 0x01
Device : 0x00
Domain : 0x0000
Device Id : 0x11C610DE
Bus Id : 0000:01:00.0
Sub System Id : 0x36523842
GPU Link Info
PCIe Generation
Max : N/A
Current : N/A
Link Width
Max : N/A
Current : N/A
Fan Speed : 21 %
Performance State : N/A
Clocks Throttle Reasons : N/A
Memory Usage
Total : 1023 MB
Used : 729 MB
Free : 294 MB
Compute Mode : Default
Utilization
Gpu : N/A
Memory : N/A
Ecc Mode
Current : N/A
Pending : N/A
ECC Errors
Volatile
Single Bit
Device Memory : N/A
Register File : N/A
L1 Cache : N/A
L2 Cache : N/A
Texture Memory : N/A
Total : N/A
Double Bit
Device Memory : N/A
Register File : N/A
L1 Cache : N/A
L2 Cache : N/A
Texture Memory : N/A
Total : N/A
Aggregate
Single Bit
Device Memory : N/A
Register File : N/A
L1 Cache : N/A
L2 Cache : N/A
Texture Memory : N/A
Total : N/A
Double Bit
Device Memory : N/A
Register File : N/A
L1 Cache : N/A
L2 Cache : N/A
Texture Memory : N/A
Total : N/A
Retired Pages
Single Bit ECC : N/A
Double Bit ECC : N/A
Pending : N/A
Temperature
Gpu : 32 C
Power Readings
Power Management : N/A
Power Draw : N/A
Power Limit : N/A
Default Power Limit : N/A
Min Power Limit : N/A
Max Power Limit : N/A
Clocks
Graphics : N/A
SM : N/A
Memory : N/A
Applications Clocks
Graphics : N/A
Memory : N/A
Default Applications Clocks
Graphics : N/A
Memory : N/A
Max Clocks
Graphics : N/A
SM : N/A
Memory : N/A
Compute Processes : N/A


Though, 32c suggests it isn't working too hard.

I am already running 99% cpu to free up a core (well, thread)

top - 09:57:20 up 2 days, 13:04, 1 user, load average: 7.21, 7.29, 7.30
Tasks: 117 total, 8 running, 109 sleeping, 0 stopped, 0 zombie
Cpu(s): 0.0%us, 0.2%sy, 88.1%ni, 11.8%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st
Mem: 16326392k total, 1070696k used, 15255696k free, 153076k buffers
Swap: 8292348k total, 0k used, 8292348k free, 418284k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
18589 boinc 39 19 41144 2672 1348 R 100 0.0 2:34.00 eonclient_5.00_
18591 boinc 39 19 40992 2516 1348 R 100 0.0 2:01.92 eonclient_5.00_
18593 boinc 39 19 40840 2368 1348 R 100 0.0 1:52.80 eonclient_5.00_
18595 boinc 39 19 40692 2224 1348 R 100 0.0 1:12.54 eonclient_5.00_
18597 boinc 39 19 40540 2068 1348 R 100 0.0 1:01.39 eonclient_5.00_
18599 boinc 39 19 40464 2000 1348 R 100 0.0 0:48.35 eonclient_5.00_
18585 boinc 39 19 41292 2820 1348 R 99 0.0 2:55.04 eonclient_5.00_
1485 boinc 30 10 28.3g 214m 50m S 6 1.3 223:41.23 acemd.2868
1434 boinc 20 0 56.1g 14m 5712 S 0 0.1 5:37.67 boinc
18486 boinc 20 0 81876 1744 880 S 0 0.0 0:00.00 sshd


One thing that does concern me is a message in the system log

[ 56.391646] NVRM: Your system is not currently configured to drive a VGA console
[ 56.391649] NVRM: on the primary VGA device. The NVIDIA Linux graphics driver
[ 56.391650] NVRM: requires the use of a text-mode VGA console. Use of other console
[ 56.391652] NVRM: drivers including, but not limited to, vesafb, may result in
[ 56.391653] NVRM: corruption and stability problems, and is not supported.

which makes me wonder if I should hook a display up to the card. I have been running my NVIDIA linux boxes from the onboard video to save a few cycles on the GPU's for many years but perhaps the new driver doesn't permit that.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1104
Credit: 6,101,732,079
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30212 - Posted: 22 May 2013 | 17:34:27 UTC - in response to Message 30210.

One thing that does concern me is a message in the system log

[ 56.391646] NVRM: Your system is not currently configured to drive a VGA console
[ 56.391649] NVRM: on the primary VGA device. The NVIDIA Linux graphics driver
[ 56.391650] NVRM: requires the use of a text-mode VGA console. Use of other console
[ 56.391652] NVRM: drivers including, but not limited to, vesafb, may result in
[ 56.391653] NVRM: corruption and stability problems, and is not supported.

which makes me wonder if I should hook a display up to the card. I have been running my NVIDIA linux boxes from the onboard video to save a few cycles on the GPU's for many years but perhaps the new driver doesn't permit that.

Or perhaps a dummy plug?

Profile Elmar Klausmeier
Send message
Joined: 22 Mar 11
Posts: 2
Credit: 27,938,105
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 30222 - Posted: 22 May 2013 | 18:51:23 UTC

Maybe the following post

http://eklausmeier.wordpress.com/2013/05/16/cuda-without-x/

will help you alleviate the problem
Your system is not currently configured to drive a VGA console


Basically you pass the following additional parameters
video=vesa:off vga=normal

to the Linux kernel at startup in your GRUB configuration file.

Profile Carlesa25
Avatar
Send message
Joined: 13 Nov 10
Posts: 324
Credit: 72,394,453
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30229 - Posted: 22 May 2013 | 21:16:05 UTC

Hello: No English language domain, all for Google translator, sorry.

Possibly temperature problem, which has monitored.

Nvidia drivers mismanaged fan control and also to manually adjust the fan must be activated Coolbits, as explained in this forum.

If the card does not have a display connected Coolbits can not be activated and will have temperature problems, I've also explained in this forum to create and connect a virtual screen that allows Coolbits activate and control the fan, but is a bit complex.

The Nvidia 319.17 driver is running fine me on a virtual screen GTX590 and GTS450 only control your computer screen. Ubuntu 13.04-64bit Kernel 3.9. Greetings.

matlock
Send message
Joined: 12 Dec 11
Posts: 34
Credit: 86,423,547
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 30239 - Posted: 23 May 2013 | 0:10:00 UTC - in response to Message 30210.


Fan Speed : 21 %
Temperature
Gpu : 32 C


Your GPU is idle.

fractal
Send message
Joined: 16 Aug 08
Posts: 87
Credit: 1,182,558,826
RAC: 186,737
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30245 - Posted: 23 May 2013 | 1:07:06 UTC

Adding kernel options did not help, nor did connecting a display.

Downgrading from 319.17 to 310.44 seems to have fixed it. The time to complete is dropping rapidly (10's of seconds per update), the temperature of the card is increasing (50-51c) and top now reports

3010 boinc 30 10 28.3g 218m 51m R 44 1.4 1:20.73 acemd.2868

showing that acemd is taking half a core.

So, ymmv but 319.17 did not work very well for me with a 650ti on ubuntu 12.04 LTS

matlock
Send message
Joined: 12 Dec 11
Posts: 34
Credit: 86,423,547
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwat
Message 30246 - Posted: 23 May 2013 | 5:48:01 UTC - in response to Message 30245.


So, ymmv but 319.17 did not work very well for me with a 650ti on ubuntu 12.04 LTS


319.17 will work if installed with a properly updated Xorg and kernel. There are many issues with installing it manually, so I just stay with the distro packages. For Ubuntu there is also the xorg-edgers PPA for the latest Xorg, Nvidia, AMD, etc drivers, but I don't see support for CUDA with those packages. I think if you install CUDA after using that PPA, it will try to revert the packages to resolve the dependencies.

If looking for a distro with the latest packages, Arch Linux looks to be a good choice: https://www.archlinux.org/packages/extra/x86_64/nvidia/

I will continue to use Lubuntu (with MATE desktop environment) as my desktop OS, since it's very easy to manage the OS as a whole. If I build a dedicated crunching machine, I will likely choose Arch to have the latest packages for the few packages I will need.

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwat
Message 30252 - Posted: 23 May 2013 | 10:50:46 UTC - in response to Message 30245.

Adding kernel options did not help, nor did connecting a display.

Downgrading from 319.17 to 310.44 seems to have fixed it. The time to complete is dropping rapidly (10's of seconds per update), the temperature of the card is increasing (50-51c) and top now reports
3010 boinc 30 10 28.3g 218m 51m R 44 1.4 1:20.73 acemd.2868

showing that acemd is taking half a core.

So, ymmv but 319.17 did not work very well for me with a 650ti on ubuntu 12.04 LTS


319.17 working fine on my Ubuntu 12.04 LTS Server, which is otherwise completely stock, nothing outside the repositories. Upgraded to an ASUS GTX 650Ti from a Gigabyte GT 440 with no issues (knock on wood), just swapped the cards.

fractal, something must be wrong with your system configuration. Are you sure you have blacklisted the nouveau driver? Can you see a line like
blacklist nouveau

in /etc/modprobe.d/blacklist.conf?

fractal
Send message
Joined: 16 Aug 08
Posts: 87
Credit: 1,182,558,826
RAC: 186,737
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 30276 - Posted: 23 May 2013 | 19:38:57 UTC - in response to Message 30252.
Last modified: 23 May 2013 | 19:41:14 UTC

It is not in /etc/modprobe.d/blacklist.conf but the installer created

boinc@charles:/etc/modprobe.d$ cat nvidia-installer-disable-nouveau.conf
# generated by nvidia-installer
blacklist nouveau
options nouveau modeset=0

That initial long running work unit did finish and validate and I am half way through a second one. It looks like it will take about 18 hrs. acemd is taking 100% of one thread on a stock i7-2700K feeding this card which is steady at 50c.

And, it may be worth noting that I am running 12.04 LTS x86_64 server (no gui) stock with nothing other than boinc and nvidia drivers. I have been running gpugrid on linux machines with no gui since I started this project in 08. I was using fedora back then and generally stick with ubuntu LTS versions these days.

Post to thread

Message boards : Graphics cards (GPUs) : Linux - Nvidia Driver 319.17 certificate