I am well aware that the GPU is not useable for GPUGRID, but as it does not work at all, I would like to try it first with an old card I have, and change it to a capable card later.

skgiven may have supplied the vital clue to solving your problem. Maybe the BOINC manager is more sophisticated than we think - its' identification of usable GPUs may be project specific (who wooda taugh!). There is, of course, an easy way to find out - why not attach to a project that will accept any NVidia cc 1.1 card and see what happens. seti@home is a popular one.

---

klepel,

You might try an NVIDIA driver from a different source. If I remember correctly, several months ago I tried the driver from xorg-edgers and it did not work for me for GPU computing. I had to do a manual install of the driver that I downloaded directly from NVIDIA. It is working fine for me on Ubuntu.

captainjack,

I tried this before as well, but with the instruction for Ubuntu from Carlesa25 in Spanish I found in the forum, I managed to change the command lines that it does something in Mint, but I was not able to kill the nouveau driver, as I do not really know what I am doing:
“kl#!/bin/sh
service mdm stop
nvidia-uninstall
apt-get purge remove nvidia*
apt-get purge xserver-xorg-video-nouveau nvidia*
update-initramfs -u -k all bash /opt/NVIDIA.run
service mdm start”

So I am really looking for somebody who has experience with Linux Mint, best with newest version, which might be able to guide me. I really would like change one Computer with a GTX 570 from Windows XP to LINUX.

klepel,

Here are some instructions that show how to install the NVIDIA drivers in Ubuntu. It includes a section on how to stop the nouveau drivers.

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

If you scroll down a few posts, there are some additional comments from skgiven that help clarify.

Hope that helps.

A GTS 8400 is not capable of running GPUGrid tasks as it's a Compute Capable 1.1 card.
CC1.3 or better is required to crunch here.

Tasks will not be sent to you unless you have a CC1.3 card or better. This is determined by the project via compute capability.

---

FAQ's

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

Dear all,

I'm not receiving tasks since last week.
The message I get is:

"This computer has finished the daily quota of 1 tasks"

That' obviously not true.
My environment is pretty old in terms of software but, unfortunately, there is no easy and fast way to upgrade it.

Any other suggestion?

Thanks in advance

Nikke,

What nvidia driver version do you have?

Matt

Hi Matt,

the system is based on:
- Centos 6.5, running kernel version 2.6.32-431.5.1.el6.x86_64
- Nvidia driver version 331.38
- Cuda version 5.5.22
- BOINC version 6.10.58

Hi Matt,

the system is based on:
- Centos 6.5, running kernel version 2.6.32-431.5.1.el6.x86_64
- Nvidia driver version 331.38
- Cuda version 5.5.22
- BOINC version 6.10.58

Hi Nikke

all your WUs was failed reporting

<core_client_version>6.10.58</core_client_version>
<![CDATA[
<message>
process exited with code 199 (0xc7, -57)
</message>
<stderr_txt>
SWAN : FATAL : Cuda driver error 3 in file 'swanlibnv2.cpp' in line 446.
# SWAN swan_assert -57

</stderr_txt>
>

Hi Matt,

the system is based on:
- Centos 6.5, running kernel version 2.6.32-431.5.1.el6.x86_64
- Nvidia driver version 331.38
- Cuda version 5.5.22
- BOINC version 6.10.58

Hi Nikke

all your WUs was failed reporting

<core_client_version>6.10.58</core_client_version>
<![CDATA[
<message>
process exited with code 199 (0xc7, -57)
</message>
<stderr_txt>
SWAN : FATAL : Cuda driver error 3 in file 'swanlibnv2.cpp' in line 446.
# SWAN swan_assert -57

</stderr_txt>
>

Yes, you've had 156 failed Work units all with v8.21 (cuda60) app
Your last successfully completed WU was on Oct 6th and that was using the v8.03 (cuda42) app.

Since Matt changed the server scheduling policy the GPUGrid server needs to know which Nvidia driver version you have installed to send you the proper WUs/app version.

If you upgrade your boinc client to version 7.4.22, boinc will determine your nvidia driver version and report it to the server. You should then receive exclusively cuda 4.2 WUs.

See this thread for new scheduling system:
http://www.gpugrid.net/forum_thread.php?id=3874

If you'd like to receive cuda 6.0 or 6.5 Work then you'll need to upgrade your nvidia driver to 340.46 for cuda 6 or 343.23 for cuda 6.5 and also upgrade your boinc client to 7.4.22

EDIT: I made multiple changes to try be more clear. sorry.

Thank you all guys,

I was thinking something like that.
The only problem I see is upgrading the BOINC client, since the version 7 seems difficult to run under CentOS (at least it was last year...)
Once upgraded it, I can also try to upgrade drivers and CUDA easily.

Do you have any link suggestion for the BOINC client (possibly a fast and easy one ;-) ) ?

Thank you all guys,

I was thinking something like that.
The only problem I see is upgrading the BOINC client, since the version 7 seems difficult to run under CentOS (at least it was last year...)
Once upgraded it, I can also try to upgrade drivers and CUDA easily.

Do you have any link suggestion for the BOINC client (possibly a fast and easy one ;-) ) ?

I Nikke,
I solved a similar problem (on kUbuntu 14.04) installing manually the latest Nvidia driver.
Do do so I followed step by step the tutorial that some posted on this thread:

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

to install the latest boinc server and client I use a custom PPA, that on my system works without any problem

http://setiathome.berkeley.edu/forum_thread.php?id=70320

Hope it could help and works also on your OS
G

Thank you all guys,

I was thinking something like that.
The only problem I see is upgrading the BOINC client, since the version 7 seems difficult to run under CentOS (at least it was last year...)
Once upgraded it, I can also try to upgrade drivers and CUDA easily.

Do you have any link suggestion for the BOINC client (possibly a fast and easy one ;-) ) ?

you can get new boinc client here if you can't get it via package-manager and repositories are too old

and try to run it manually unpacking tarball and run client/manager
http://boinc.berkeley.edu/wiki/Installing_BOINC#Linux

but ...

you have a very hold OS and most packages need to be upgrade (glibc first but this is a very important library for the system . Pay attention if you upgrade this.)

you can check dinamic link library as following :

run inside boinc folder

$ ldd {boinccmd,boincmgr,boinc_client}

Unfortunately upgrading the OS is out of scope....

I've seen that the boinc client version 7.2.33 is available from the EPEL repository but I suppose it's not enough.

Last year I tried to manually compile the BOINC client from source.....as already noticed it was a nightmare of libraries dependencies.

Nikke,

The problem is that your machine has one of the few bad driver versions that incorrectly report haing CUDA 6 capability.

Get our version of the client from

http://www.gpugrid.net/forum_thread.php?id=3736&nowrap=true#36577

This reports the exact driver version to the server and lets it give you an appropriate app version.

Matt

You're right Matt.
I've upgraded CUDA to 6.5, drivers to 343.22 and, using your compiled boinc client I'm now able to compute again.

I hope to be back soon with some GPU power ;-)

Many thanks