SETI MB CUDA for Linux

[sunu]

If I just try with the -image and the -headers packages, the -headers fail, but the image installs and shows up in the menu.lst.  Is that all I really need?  I get the error that the -headers package can't be installed because it depends on itself. 

BTW, the modules that fail now are nvidia (180.44), vboxdrv (2.1.4), and vboxnetflt (2.1.4).

You need the headers package for the nvidia drivers. How do you install the 8.10 kernel? What I did is go to /etc/apt/sources.list and replace every "jaunty" word with "intrepid", so was just like it was in intrepid. Update to reload the packages info and then install these four packages:

linux-headers-2.6.27-14
linux-headers-2.6.27-14-generic
linux-image-2.6.27-14-generic
linux-restricted-modules-2.6.27-14-generic

After installation revert back to the previous "jaunty edition" sources.list and you should be ok.


Also don't give up on cuda 2.2 with 2 gpus. Put back the second GPU, and try more. Tell me how many nvidia devices exist under /dev. Try the xorg.conf tweaks I posted above and post here a new xorg.0.log with the tweaks.

And I need those debug files. But first let's  try standalone. Do you have any workunits?

[Tye]

I did the sources.list trick, and with a little finagling, I was able to get the 8.10 kernel to install with no broken deps or other warnings.  It even added the nvidia driver and vbox stuff correctly.  Booted fine but is crashy now - argh.  Hard to fix via SSH and VNC from work to home...

[sunu]

Ok, I hope the kernel is ok. Now it's time to attack those 2 gpus.

[Tye]

I think it was crashy because of nvclock - will not use it just yet.  Might need the 8.10 version, or my card is unrecognized...  Getting my wife to reboot right now so I can login and see what the deal is...

[sunu]

I also use nvclock with absolute no problems at all. I don't use that from the repositories, but I got it straight from http://www.linuxhardware.org/nvclock/

[Tye]

No worries.  It's not crashy now - one of the things I did must have helped... 

Anyway, got the nvidia-site 185 driver back in and got some debug files for you since I was able to get some workunits.

[attachment deleted by admin]

[sunu]

Tye, the debug files tell me that you've not followed my directions from http://lunatics.kwsn.net/linux/seti-mb-cuda-for-linux.msg19014.html#msg19014 step number 4 to be exact. In your case that list of locations has two more options: /home/tye/bin and /usr/local/cuda/bin . That's why you're getting those instant computation faults.

[Tye]

Aha!  I put it in my ~/bin directory and no more immediate fails.  And it frees up a CPU after a minute so.

I am still crashy (total freeze) however, so I'm thinking of going back to the most recent 9.04 kernel.

[sunu]

Watch your cards for overheat. Cuda stresses the cards very much. I use nvclock to run my card's fan always at 100%.

[Tye]

Thanks much sunu - looks like putting the file in the correct place also works with the ubuntu 185 driver I had earlier (now up to 0ubuntu3 though).  Tried it on another machine just out of curiosity and it's as simple as installing that driver (4 files from karmic 'release'), putting the CUDA app in your $PATH, putting the libraries in the seti work area and making the correct lings.

I dId have to go back to the 9.04 kernel though - kept getting hard freezes.

Anyway, thanks again - you saved me.  I thought for sure I had put the CUDA app in the right place.  Argh.

[sunu]

Glad we solved it. Have you tried again with two cards?

[Tye]

Haven't tried again with two cards, but don't forsee a problem as that was working just fine with 180.44 so the config files will likely work.  If not I'll try your trick.

Cheers!  - Tye

[Richard Haselgrove]

Came across an interesting error message in task 1294937260 while researching something else.

SETI@home MB CUDA 608 Linux 64bit SM 1.0 - r06 by Crunch3r :p

Error: API mismatch: the NVIDIA kernel module has version 180.29,
but this NVIDIA driver component has version 180.60.  Please make
sure that the kernel module and all NVIDIA driver components
have the same version.
setiathome_CUDA: Found 1 CUDA device(s):
Cuda error 'cudaGetDeviceProperties( &cDevProp, i )' in file './cudaAcceleration.cu' in line 138 : initialization error.

Something to watch for when fiddling about with Linux drivers and modules.

The anonymous owner of host 5011059 seems to be having a real problem getting his or her GTX 295 running under gentoo.

[Tye]

Figured out what the crashiness was - turns out that the 9600 GSOs I had that were 768 do not like being in the primary PCI-Ex slot.  If I put them in the secondarly slot and use a different card in the primary, then all is good.  Even the 9600 GSO 512 does fine in the primary - just not my two GSO 768s.  Argh.  Looks like no double-CUDA'ing with them, but might try mixed-mode with the 512 if I can pick up another card and put it where the 512 is being used now...

[sunu]

Came across an interesting error message in task 1294937260 while researching something else.

SETI@home MB CUDA 608 Linux 64bit SM 1.0 - r06 by Crunch3r :p

Error: API mismatch: the NVIDIA kernel module has version 180.29,
but this NVIDIA driver component has version 180.60.  Please make
sure that the kernel module and all NVIDIA driver components
have the same version.
setiathome_CUDA: Found 1 CUDA device(s):
Cuda error 'cudaGetDeviceProperties( &cDevProp, i )' in file './cudaAcceleration.cu' in line 138 : initialization error.

Something to watch for when fiddling about with Linux drivers and modules.

The anonymous owner of host 5011059 seems to be having a real problem getting his or her GTX 295 running under gentoo.

He hasn't installed the NVIDIA drivers properly.

Figured out what the crashiness was - turns out that the 9600 GSOs I had that were 768 do not like being in the primary PCI-Ex slot.  If I put them in the secondarly slot and use a different card in the primary, then all is good.  Even the 9600 GSO 512 does fine in the primary - just not my two GSO 768s.  Argh.  Looks like no double-CUDA'ing with them, but might try mixed-mode with the 512 if I can pick up another card and put it where the 512 is being used now...

Interesting.