Accelerators not found

Dear all,

I have downloaded pgiws-105.exe and completed installization and licenscing. The PGI and CUDA directories are C:\Program Files\PGI and C:\CUDA respectively. However, I got the following messages as returned by the command “pgaccelinfo -v”.

===========================================
CUDA Driver Version: 3000

could not initialize CUDA runtime, error code=100

libamdcalcl.dll not found

No accelerators found.

Check that you have installed the CUDA or CAL libraries properly

Check that your PATH includes the CUDA or CAL runtime installation directory

Check the permissions on your device

Could anyone please kindly help me solve this problem? Thank you very much.

M-H Chung

Hi mhsuan-

Are you able to run any of the NVIDIA SDK examples?

i.e:

C:\Documents and Settings\All Users\Application Data\NVIDIA Corporation\NVIDIA GPU Computing SDK\C\bin\win64\Release\deviceQuery.exe

Dear dholt,

The result of execution is as follows.

deviceQuery.exe Starting…

CUDA Device Query (Runtime API) version (CUDART static linking)

There is no device supporting CUDA

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 0.0, CUDA Runtime Version = 0.0, NumDevs = 0


PASSED

Press to Quit…



Is something wrong? Thanks a lot.

M-H

I can see “NVIDIA Tesla C2050” under the “Display adapter” in the Control Panel. In fact, I have updated the driver from the following location, but in vain.

C:\NVIDIA\DisplayDriver\197.68\WinXP\International\NV4_DISP.INF

Thanks in advance for your reply.

M-H

Dear dholt,

The result of execution is as follows.

deviceQuery.exe Starting…
CUDA Device Query (Runtime API) version (CUDART static linking)
There is no device supporting CUDA
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 0.0, CUDA Runtime Version = 0.0, NumDevs = 0

PASSED

Press to Quit…

BTW, I can see “NVIDIA Tesla C2050” under the “Display adapter” in the Control Panel. In fact, I have updated the driver from the following location, but in vain.

C:\NVIDIA\DisplayDriver\197.68\WinXP\International\NV4_DISP.INF

Thanks in advance for your reply.

M-H

This sounds like a driver/toolkit compatibility issue. I would try uninstalling both and reinstalling the driver and toolkit (in that order) from the CUDA 3.0 developer page:

http://developer.nvidia.com/object/cuda_3_0_downloads.html#Windows

Thanks for your suggestions. Before doing that, I have found that the troubles possibily come from the Remote Desktop application. The situation is that I have connected my home computer to one in the workplace via Remote Desktop. The GPU card is installed in the workplace computer. When I run SDK executables, e.g., deviceQuery.exe, via the Remote Desktop window displayed on the monitor of the home computer, messages like “There is no device supporting CUDA” come out. However, the execution works normally when running the codes using the on-site mouse-keyboard-monitor set in the workplace. The outcome keeps unchanged after reinstalling the display driver and CUDA toolkit. This phenomenon really confuses me. What’s the possible reason and how can I fix the problem? Thank you very much.

BTW, the file “devdriver_3.0_winxp_32_197.13_general.exe” downloaded from the web site you provide “can not locate any driver matched with the current hardware”. Therefore, I manually find drivers for Tesla C2050 from the “Download Drivers” at www.nvidia.com.
The file is named “197.68_Tesla_winxp_32bit_international_whql.exe”.

Regards,

M-H Chung

The situation is that I have connected my home computer to one in the workplace via Remote Desktop.

That is the problem. Remote Desktop does not provide access to the underlying GPU driver. Your only option for working remotely is VNC or via SSH. I posted a description of how to enable SSH in PGI Workstation in this thread.

Thanks a lot. The SSH server works and the GPU can be fully driven now while it is a little inconvenient. Hope for an improved solution in the near future. Another problem is that, when I run computation-intensive codes like “nbody.exe” in the SDK, activation of another program, e.g., Microsoft Outlook, always causes the computer down. Is the capability of power supply not affordable? The only Tesla C2050 card is also used as the display card. The maximum power output of the power supply is 530W. Thank you very much.

M-H Chung