The CUDA-client doesn't work correctly on my Vista 64-bit system. According to Bernd Machenschalk this is due to a bug which has not been fixed yet. I discussed that problem in another thread some months ago.
Therefore I will use a second partition on my harddisk for running Ubuntu-Linux 10.04 LTS. Unfortunately the CUDA-environment of 10.04 needs to be upgraded to Cuda Toolkit 3.2.And this is not as trivial to install as within a Windows-OS.
Why is it not possible to download actual drivers and CUDA-toolkits which can be installed by running apt or the Upgrade-center of Ubuntu?
I use the 32-bit-version of Ubuntu, because this is the recommended version by Canonical.
Intel Q9300 Quadcore, 2500 Mhz, 4096 MB RAM, GeForce 9800 GT, Vista Ultimate 64-bit, Ubuntu 10.10 64 bit
The CUDA-client doesn't work correctly on my Vista 64-bit system. According to Bernd Machenschalk this is due to a bug which has not been fixed yet. I discussed that problem in another thread some months ago.
Therefore I will use a second partition on my harddisk for running Ubuntu-Linux 10.04 LTS. Unfortunately the CUDA-environment of 10.04 needs to be upgraded to Cuda Toolkit 3.2.And this is not as trivial to install as within a Windows-OS.
Why is it not possible to download actual drivers and CUDA-toolkits which can be installed by running apt or the Upgrade-center of Ubuntu?
I use the 32-bit-version of Ubuntu, because this is the recommended version by Canonical.
Regarding the CUDA BRP3 app, the CUDA related libraries it depends on are libcufft.so.3, libcudart.so.3, and libcuda.so.1. You could download the toolkit from nvidia.com and extract it with -x flag. Then copy those libraries to /usr/lib, run ldconfig, and be good to go.
Why are you running a 32-bit version of Linux on 64-bit hardware?
Because the pae kernel can use all my memory, 32bit systems are often faster, and I don't have to worry about compatibility problems for applications and drivers?
In fact, what reason do I have to upgrade to 64bit at this point (beside geek-cred)? Not trying to be contrary - a real question...
Daer Mike,
thanks for the complete list of BRP requirements.
Just one small addition: From the ration CPU time vs.
wall time it seems to take 50% of one of my Intel
Q6600 cores to feed the 336 cores of my slightly
overclocked Gigabyte GTX460.
Happy 2011 and may the gravitational waves be with you ;-)
Armin
I'm seeing 40-43% load for my i7-920 @ 4.1ghz, with my 260s, with 1 WU running per GPU at 40-50% GPU load. I'll probably post 2/3 WU/card results tomorrow or the next but want to get a few to validate first, and to make sure my system's stable before I start pushing it really hard (I only swapped these back in to troubleshoot some stability/artifacting issues I think my 5870 is responsible for).
I have switched to Ubuntu 10.10 64-bit. Now einstein@home for CUDA runs fine out of the box.
While running Ubuntu 10.04 LTS 32-bit I had to upgrade CUDA manually. This didn`t seem to work, because the workunits crashed each time. An output file was missing.
Intel Q9300 Quadcore, 2500 Mhz, 4096 MB RAM, GeForce 9800 GT, Vista Ultimate 64-bit, Ubuntu 10.10 64 bit
The CUDA-client doesn't work
)
The CUDA-client doesn't work correctly on my Vista 64-bit system. According to Bernd Machenschalk this is due to a bug which has not been fixed yet. I discussed that problem in another thread some months ago.
Therefore I will use a second partition on my harddisk for running Ubuntu-Linux 10.04 LTS. Unfortunately the CUDA-environment of 10.04 needs to be upgraded to Cuda Toolkit 3.2.And this is not as trivial to install as within a Windows-OS.
Why is it not possible to download actual drivers and CUDA-toolkits which can be installed by running apt or the Upgrade-center of Ubuntu?
I use the 32-bit-version of Ubuntu, because this is the recommended version by Canonical.
Intel Q9300 Quadcore, 2500 Mhz, 4096 MB RAM, GeForce 9800 GT, Vista Ultimate 64-bit, Ubuntu 10.10 64 bit
RE: Why is it not possible
)
Just a guess, but are the CUDA parts closed source binary instead of open?
RE: The CUDA-client doesn't
)
Regarding the CUDA BRP3 app, the CUDA related libraries it depends on are libcufft.so.3, libcudart.so.3, and libcuda.so.1. You could download the toolkit from nvidia.com and extract it with -x flag. Then copy those libraries to /usr/lib, run ldconfig, and be good to go.
RE: Why are you running a
)
Because the pae kernel can use all my memory, 32bit systems are often faster, and I don't have to worry about compatibility problems for applications and drivers?
In fact, what reason do I have to upgrade to 64bit at this point (beside geek-cred)? Not trying to be contrary - a real question...
RE: Daer Mike, thanks for
)
I'm seeing 40-43% load for my i7-920 @ 4.1ghz, with my 260s, with 1 WU running per GPU at 40-50% GPU load. I'll probably post 2/3 WU/card results tomorrow or the next but want to get a few to validate first, and to make sure my system's stable before I start pushing it really hard (I only swapped these back in to troubleshoot some stability/artifacting issues I think my 5870 is responsible for).
Edit: win7-64
I have switched to Ubuntu
)
I have switched to Ubuntu 10.10 64-bit. Now einstein@home for CUDA runs fine out of the box.
While running Ubuntu 10.04 LTS 32-bit I had to upgrade CUDA manually. This didn`t seem to work, because the workunits crashed each time. An output file was missing.
Intel Q9300 Quadcore, 2500 Mhz, 4096 MB RAM, GeForce 9800 GT, Vista Ultimate 64-bit, Ubuntu 10.10 64 bit