Problems installing Tesla C2050 on Dell T7500

Accelerated Computing CUDA CUDA Programming and Performance
1

I am trying to get a Tesla C2050 to work in a Dell Precision T7500 workstation. I have installed the newest developer driver

260.24 from the Cuda Developer Site together with the Cuda Toolkit 3.2.9 and the GPU computing code samples version 3.2 under Ubuntu 10.04.

Installing the driver and compiling the examples works fine. Also, the driver itself is running. However, when I call deviceQuery I receive the following error message:

[codebox]

NVIDIA: could not open the device file /dev/nvidia1 (Input/output error).

cudaGetDeviceCount FAILED CUDA Driver and Runtime version may be mismatched.

FAILED

[/codebox]

Correct symlinks in /dev are present:

[codebox]

crw-rw-rw- 1 root root 195, 0 2010-09-21 22:29 /dev/nvidia0

crw-rw-rw- 1 root root 195, 1 2010-09-21 22:29 /dev/nvidia1

crw-rw-rw- 1 root root 195, 255 2010-09-21 22:29 /dev/nvidiactl

[/codebox]

In /var/log/messages the following error message appears:

[codebox]

Sep 21 22:44:02 smlx14 kernel: [ 924.433774] NVRM: rm_init_adapter(1) failed

Sep 21 22:55:41 smlx14 kernel: [ 1621.947432] NVRM: RmInitAdapter failed! (0x26:0x48:1028)

[/codebox]

I would be grateful for any comments on this issue.

Best wishes

Timo

2

Hi Timo,

I was experiencing the same error (with the driver 256.53 and cuda 3.2).

Tired of trying to install sdk.

I re-installed my old 3.0 and related sdk (I leaved the driver 256.53).

Now all works fine and deviceQuery gives me:

Device 0: “GeForce GTX 260M”

CUDA Driver Version: 3.10

CUDA Runtime Version: 3.0

…

So probabily before the 2nd installation, the version of the driver was smaller than

that of the cuda.

I suggest you to try to install a older version of cuda and sdk.

good luck

marco

3

Hi Timo,

I was experiencing the same error (with the driver 256.53 and cuda 3.2).

Tired of trying to install sdk.

I re-installed my old 3.0 and related sdk (I leaved the driver 256.53).

Now all works fine and deviceQuery gives me:

Device 0: “GeForce GTX 260M”

CUDA Driver Version: 3.10

CUDA Runtime Version: 3.0

…

So probabily before the 2nd installation, the version of the driver was smaller than

that of the cuda.

I suggest you to try to install a older version of cuda and sdk.

good luck

marco

4

Hi,

unfortunately going back to the 3.1 version of the Toolkit and SDK did not help. The error message still remains.
Best wishes

Timo

5

Hi,

unfortunately going back to the 3.1 version of the Toolkit and SDK did not help. The error message still remains.
Best wishes

Timo

6

Please attach the output of nvidia-bug-report.sh run as root on the affected system. In addition, please attach the output of “strace -o cuda-strace.log deviceQuery”.

7

Please attach the output of nvidia-bug-report.sh run as root on the affected system. In addition, please attach the output of “strace -o cuda-strace.log deviceQuery”.

8

The error is now fixed. It turned out that only one power cable was connected with the card. After connecting the second power cable everything is running fine.

Best wishes

Timo

9

The error is now fixed. It turned out that only one power cable was connected with the card. After connecting the second power cable everything is running fine.

Best wishes

Timo

10

The error is now fixed. It turned out that only one power cable was connected with the card. After connecting the second power cable everything is running fine.

Best wishes

Timo