Tesla on Mac Pro, "no CUDA-capable device is detected"

I’ve installed a Tesla c1060 into a Mac Pro. There are green lights on both pcie power ports for the tesla. I’ve installed the driver version 4.2.5, the cuda tool kit version 4.1.28 and the GPU computing SDK 4.1.28.

I’ve followed the Mac Guide :

http://developer.download.nvidia.com/compute/DevZone/docs/html/C/doc/CUDA_Getting_Started_Mac.pdf

The card is getting recognized in the “About this mac” as a GPU device id 0x05e7, revision id 0x00a1

I’m running OS X lion 10.7.3

I’ve verified that the cud a kext is loaded correctly

I’ve reset the SMC by having the system unplugged for 15+ seconds.

I’ve built the GPU SDK successfully.

But when I run deviceQuery, the cud a card is not detected. I get the following:

[deviceQuery] starting...

./deviceQuery Starting...

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

cudaGetDeviceCount returned 38

-> no CUDA-capable device is detected

[deviceQuery] test results...

FAILED

Searching the system logs through “console” these are the entries with the word Cuda in it.

4/5/12 9:26:31.755 AM com.apple.SystemStarter: The domain/default pair of (com.nvidia.CUDAPref, IgnoreStartupCompatabilityCheck) does not exist

4/5/12 9:26:55.547 AM com.apple.SystemStarter: CUDA software is compatible with GPU driver.

4/5/12 9:27:01.434 AM com.apple.SystemStarter: Starting CUDA Kernel Extension

The “IgnoreStartupCompatabilityCheck does not exist” looks suspicious, but not fatal. Nor do I know how to resolve that issue.

Anyone experience this issue before? I looked through the forums but saw nothing conclusive about trouble with a cuda only card, most problems I read about related to the graphic GPUs like 4000.

There was one post about someone getting this problem when upgrading to version 4. I’m starting with version 4 and have never installed a prior version.

I booted through bootcamp to windows 7 64 bit. The Tesla was successfully detected and installed. So I know the card is connected correctly and has proper power.

GPUs on Mac require EFI information, this is why some cards have a special “Mac” version.
AFAIK, Tesla cards have no EFI info.

When you boot in Windows, the OS is using the BIOS information.