Do you know if Ubuntu 9.04 will be upgraded soon to that driver ? Because in my case I need to follow their distribution updates.

Yeah great, first workunit without error with Nvidia drivers version 185.18.14 application 6.66, my result .
Thanks

The fix seems to be working. I used the beta driver 190 and cuda 2.3 and all is going well with Suse 11.2.
Good work on the workaround.

Once a Ubuntu release is out, they don't update the drivers therein to new major numbers. So with 9.04 you will not see 185.xx coming in one day...

What you can do is:

1. - Install new 185.xx or 190.xx drivers from the nvidia website
2. - you could add a PPA with current drivers to your /etc/apt/sources list
-> https://launchpad.net/~brandonsnider/+ppa-packages
-> https://launchpad.net/~nvidia-vdpau/+archive/ppa
3. - use apt pinning to get the drivers from the upcoming Ubuntu karmic release
4. - grab the 185.xx driver from packages.ubuntu.com and install them manually with dpkg -i


So there are many ways to reach the same target, the same should also work for 8.04 and 8.10.
I'm running acemd 6.66 already on one 9.04 box with 190.18 driver, no issues so far...

I tested the 190.32 nvidia linux driver without problems.
I tried also the gpu fan speed new feature in nvidia-settings.
It is really handy.

How do the fan settings work? nvidia-settings --help doesn't show any fan commands.

Mike D
____________

If you can't set the fan speed in your version of nvidia-settings, install nvclock-gtk which will allow fan speed adjustments from version 0.8 beta 4.
It should appear in your menu after installation but if it does not, use the command nvclock_gtk.

I upgraded drivers to 185.18.36 and now I have problems such as this:
http://www.gpugrid.net/result.php?resultid=1259622

Boinc sees cuda device, my 3d is running fine, what could be the problem?

Btw, I use Ubuntu 9.04, 64bit.

I am not sure what the problem is. Did you install the driver from the binary files from Ubuntu or from the Nvidia files on their website? I have had more luck with the Nvidia files from their website but you have to rerun the installation with every kernel upgrade.
Less drastic things to try would be
1) Don't run the beta application from Gpugrid. It does not cause problems for me but it may be an issue with your setup.
2) Reset the BOINC project to redownload the application.
I suspect that you might do better with the latest Nvidia drivers from their website. I will warn you that while this method worked very well with Ubuntu 9.04, it did not work well for me with earlier versions of Ubuntu.
Good luck.

I am not sure what the problem is. Did you install the driver from the binary files from Ubuntu or from the Nvidia files on their website? I have had more luck with the Nvidia files from their website but you have to rerun the installation with every kernel upgrade.
Less drastic things to try would be
1) Don't run the beta application from Gpugrid. It does not cause problems for me but it may be an issue with your setup.
2) Reset the BOINC project to redownload the application.
I suspect that you might do better with the latest Nvidia drivers from their website. I will warn you that while this method worked very well with Ubuntu 9.04, it did not work well for me with earlier versions of Ubuntu.
Good luck.

No problem. Gpugrid puts such heavy demands on the cards that I have to run through a whole process restarting the application after a reboot. Start the application at stock speeds. Overclock after it has run for 30 minutes. Etc. Etc.
The routine seems different for everyone.
The one bright spot at the end of the rainbow is that, as we move from Cuda 2.1 to 2.2 to 2.3, the Gpugrid applications will become more stable and less prone to unexplained errors. I found that the Windows versions became more stable and less demanding as the application moved to Cuda 2.3.
I prefer to run in Linux for many reasons separate from Gpugrid - stability being the most important.
Sometime I would love to try Gpugrid in FreeBSD or Solaris but we don't want to give the programmers too many nightmares.

No problem. Gpugrid puts such heavy demands on the cards that I have to run through a whole process restarting the application after a reboot. Start the application at stock speeds. Overclock after it has run for 30 minutes. Etc. Etc.
The routine seems different for everyone.

The one bright spot at the end of the rainbow is that, as we move from Cuda 2.1 to 2.2 to 2.3, the Gpugrid applications will become more stable and less prone to unexplained errors. I found that the Windows versions became more stable and less demanding as the application moved to Cuda 2.3.

I prefer to run in Linux for many reasons separate from Gpugrid - stability being the most important.

Sometime I would love to try Gpugrid in FreeBSD or Solaris but we don't want to give the programmers too many nightmares.