Tesla S870 installation troubles

I have been trying to get an S870 up and running for a few hours now. I installed the drivers from the CD and the display/video card work fine. When I connect the Tesla, I can’t start X. I just get a blank screen.

Also, the output from “ls /dev/nv*” is just “/dev/nvram” which doesn’t look right.

Can anyone help?

Any error messages in the Xorg output? Usually, you can find it in /var/log/Xorg.0.log
The output of nvidia-bug-report.sh might also help find where the problem is.

Please post or send me the output of nvidia-bug-report.sh.

The Xorg output and the nvidia bug report output are dense. What should I be looking for? Also, whenever I startx, the computer locks up and I have to do a hard reset. This makes me not really want to guess-and-check.
nvidia_bug_report.txt (168 KB)

If the system has locking up when starting X, then this is likely not a CUDA bug.

According to your bug report, you’re using the ‘nv’ X driver, and not the nvidia X driver, and I don’t see any indication that X was configured to use the nvidia X driver (or to use the cards you installed in the system). Is that the actual configuration that you’re using when the system hangs?

Other than installing the drivers from the CD, what have you done to configure this system to use CUDA and/or the nvidia X driver?

I changed it to use the nv driver. If I use nvidia, it locks up.

In order to get this system running, I did a fresh install of CentOS 5, and then installed the Linux driver (I think it is 172.01). At that point I can’t start X. I think this is a CUDA bug.

Is there any need to switch to Red Hat? I have RHEL 4 disks sitting here.

CUDA has no connection with the ability to start X. Starting X is controlled by the NVIDIA display driver, not the CUDA driver.

Have you verified that you’re using the latest motherboard BIOS?
Can you setup a serial console to capture any crash output?

I wasn’t able to get any crash output. I don’t know about the motherboard BIOS. I do know that this same machine was able to use a D870. That was in Windows.

X starts fine (with the new driver) if I unplug the S870. If I plug in the S870, it won’t work. I don’t know if that is connected to CUDA, but it seems connected to the S870.

I am installing RHEL 4 so I get a blank slate to start over. Is there anything I should/shouldn’t do?

You’re stating that a serial console produced no output when the system hung? Does it capture the normnal boot time output?

I don’t really have any other suggestions, as this problem appears to be speciifc to your system.

Hmm, a serial console isn’t what I thought. I haven’t tried that (don’t know how to set one up).

I boot up at run level 3. I type startx,

–with the “nvidia” driver and the S870 hooked up, the screen goes blank.

–with the “nvidia” driver and no S870, X works

–with the “nv” driver and the S870 hooked up, X works

It is odd that it is specific to my machine when:

  1. This machine had no problems running a D870.

  2. Several similar machines (other Dell Precision 390s) run CUDA with 8800 GTX just fine.

I’ll post if I find anything useful.