There is no device supporting CUDA

Hey guys

I have Fermi on my westmere server. I meet the below problem when I run deviceQuery program. In the first time I got “There are 60 devices supporting CUDA”; but in the second time I got “There is no device supporting CUDA.” , even for the same binary. I noticed that the Major revision number and Minor revision number are strange. Could you help me to resolve it?

[wangh@gpu02 release]$ ./deviceQuery | more
There are 60 devices supporting CUDA

Device 0: “€!n<”
Major revision number: 868398229
Minor revision number: 11042
Total amount of global memory: 232736768 bytes
Total amount of constant memory: 870783264 bytes
Total amount of shared memory per block: 6385665 bytes
Total number of registers available per block: 870783264
Warp size: 11042
Maximum number of threads per block: 88
Maximum sizes of each dimension of a block: 0 x 6387232 x 0
Maximum sizes of each dimension of a grid: 232736640 x 0 x 232736672
Maximum memory pitch: 1564477805 bytes
Texture alignment: 232736640 bytes
Clock rate: 0 kilohertz

[wangh@gpu02 release]$ ./deviceQuery | more
There is no device supporting CUDA.

Device 0: “€!n<”
Major revision number: -1481747307
Minor revision number: 11021
Total amount of global memory: 129820672 bytes
Total amount of constant memory: -1479362272 bytes
Total amount of shared memory per block: 6385665 bytes
Total number of registers available per block: -1479362272
Warp size: 11021
Maximum number of threads per block: 88
Maximum sizes of each dimension of a block: 0 x 6387232 x 0
Maximum sizes of each dimension of a grid: 129820544 x 0 x 129820576
Maximum memory pitch: 1564477805 bytes
Texture alignment: 129820544 bytes
Clock rate: 0 kilohertz

[wangh@gpu02 release] uname -a Linux gpu02.cluster 2.6.18-164.el5 #1 SMP Tue Aug 18 15:51:48 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux [wangh@gpu02 release]

Thanks for any suggestion.

  • hao

Hey guys

I have Fermi on my westmere server. I meet the below problem when I run deviceQuery program. In the first time I got “There are 60 devices supporting CUDA”; but in the second time I got “There is no device supporting CUDA.” , even for the same binary. I noticed that the Major revision number and Minor revision number are strange. Could you help me to resolve it?

[wangh@gpu02 release]$ ./deviceQuery | more
There are 60 devices supporting CUDA

Device 0: “€!n<”
Major revision number: 868398229
Minor revision number: 11042
Total amount of global memory: 232736768 bytes
Total amount of constant memory: 870783264 bytes
Total amount of shared memory per block: 6385665 bytes
Total number of registers available per block: 870783264
Warp size: 11042
Maximum number of threads per block: 88
Maximum sizes of each dimension of a block: 0 x 6387232 x 0
Maximum sizes of each dimension of a grid: 232736640 x 0 x 232736672
Maximum memory pitch: 1564477805 bytes
Texture alignment: 232736640 bytes
Clock rate: 0 kilohertz

[wangh@gpu02 release]$ ./deviceQuery | more
There is no device supporting CUDA.

Device 0: “€!n<”
Major revision number: -1481747307
Minor revision number: 11021
Total amount of global memory: 129820672 bytes
Total amount of constant memory: -1479362272 bytes
Total amount of shared memory per block: 6385665 bytes
Total number of registers available per block: -1479362272
Warp size: 11021
Maximum number of threads per block: 88
Maximum sizes of each dimension of a block: 0 x 6387232 x 0
Maximum sizes of each dimension of a grid: 129820544 x 0 x 129820576
Maximum memory pitch: 1564477805 bytes
Texture alignment: 129820544 bytes
Clock rate: 0 kilohertz

[wangh@gpu02 release] uname -a Linux gpu02.cluster 2.6.18-164.el5 #1 SMP Tue Aug 18 15:51:48 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux [wangh@gpu02 release]

Thanks for any suggestion.

  • hao

Resolve it!

The method is: 1. run ./deviceQuery as root; 2. run ./deviceQuery as common user. It looks like there is cache about the device info. Just guess.

  • Hao

Resolve it!

The method is: 1. run ./deviceQuery as root; 2. run ./deviceQuery as common user. It looks like there is cache about the device info. Just guess.

  • Hao

I (on Debian) never had to run deviceQuery as root, and it always worked.

Maybe check access rights for /dev/nvidiactl and /dev/nvidia0 (and other files if you have more than 1 GPU)?

I (on Debian) never had to run deviceQuery as root, and it always worked.

Maybe check access rights for /dev/nvidiactl and /dev/nvidia0 (and other files if you have more than 1 GPU)?