View Single Post
Old 10-21-10, 05:27 PM   #6
ltpvu
Registered User
 
Join Date: Oct 2010
Posts: 3
Default Re: Unable to install Nvidia driver in SLES 11

Hi AaronP. I have a same problem. We brought two S2050 servers from Nvidia to install in our cluster with CentOS 5.3 platform. I could not get the driver installed because of errors that is exactly same as asenjo’ errors.

Here the PCI list of one cluster node which connects to 1 PCI port of Tesla S2050

Quote:
03:00.0 PCI bridge: nVidia Corporation Tesla S870 (rev a3) (prog-if 00 [Normal decode])
04:00.0 PCI bridge: nVidia Corporation Tesla S870 (rev a3) (prog-if 00 [Normal decode])
04:01.0 PCI bridge: nVidia Corporation Tesla S870 (rev a3) (prog-if 00 [Normal decode])
04:02.0 PCI bridge: nVidia Corporation Tesla S870 (rev a3) (prog-if 00 [Normal decode])
04:03.0 PCI bridge: nVidia Corporation Tesla S870 (rev a3) (prog-if 00 [Normal decode])
07:00.0 PCI bridge: nVidia Corporation Tesla S870 (rev a3) (prog-if 00 [Normal decode])
08:00.0 PCI bridge: nVidia Corporation Tesla S870 (rev a3) (prog-if 00 [Normal decode])
08:02.0 PCI bridge: nVidia Corporation Tesla S870 (rev a3) (prog-if 00 [Normal decode])
Our cluster includes Altus 1702 servers from Penguin. Could you see if this is also the BIOS problem and how I can fix it?

A weird thing is the drivers in Nvidia CD (e.g. NVIDIA-Linux-x86_64-195.36.15-pkg2.run) may be not support S2050. It could not found Tesla S2050 in the support list of the README file in the doc directory of installation package. I want to use the NVIDIA-Linux-x86_64-260.19.12.run since the S2050 in the support list but it sounds CUDA support for Redhat 2.6.18-194 kernel only.

I have more four questions and expect your answers

1. Will CUDA 3.2 package work on CentOS 5.3 (kernel 2.6.18-128.1.1.el5.530g0000, gcc 4.1.2)?

2. If not, will the driver package (e.g. NVIDIA-Linux-x86_64-195.36.15-pkg2.run) work for Tesla S2050?

3. If I install NVIDIA-Linux-x86_64-260.19.12.run (which is said that supports Tesla S2050), can it work with older CUDA toolkit like 3.1, 3.0 …?

4. Could you show me how to install these S2050 drivers manually? I have tried to copy 'nvidia.ko' to "/lib/modules/$(uname -r)/kernel/drivers/video/” but it does not work. When I ran “modprobe nvidia”, I got the error message “FATAL: Module nvidia not found”

Thank you
ltpvu is offline   Reply With Quote