Quadro K1100M Cuda support

I have a Quadro K1100M in a Dell laptop. According to:

https://developer.nvidia.com/cuda-gpus

the Quadro K1100M has a Cuda Compute Capability of 3.0.

I’m using a Cuda-based toolchain that requires me to use Cude 7.5. I installed the latest Linux NVIDIA driver and Cuda 7.5 on a CentOS system. I followed the pre-installation directions found at:

http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#axzz4QgCtosZM

No indication was given that my GPU was not supported.

During the post-installation steps (http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#post-installation-actions) I successfully compiled the SAMPLES but both the “deviceQuery” and the “bandwidthTest” sample executables complained indicating that no CUDA enabled device was present.

So, my question is: is the Quadro K1100M in fact CUDA enabled as indicated on
https://developer.nvidia.com/cuda-gpus?

If so, what problems could cause CUDA to not recognize the K1100M as a supported device?

Thanks.

Quadro K1100M is supported, but most laptops these days have a dual GPU architecture, using both an NVIDIA GPU (dGPU) and an Intel graphics adaptor (iGPU) that are jointly performing the display tasks, in what NVIDIA refers to as an “Optimus” configuration.

Optimus configurations frequently have the dGPU (the NVIDIA GPU) powered off. In this case, it won’t be visible to CUDA applications.

You’ll need to figure out how to power it on. If you search for “optimus CUDA” you’ll find a lot of relevant information.

Thanks for the response! I’m a Newb to the GPU computing community. Didn’t even know about switchable graphics until your post. Have been reading a bunch and have learned some interesting things.

Turns out the Dell botched up Optimus support on the Precision 4800 that I have. Looks like the only way to get this to work is to completely disable the integrated Intel Graphics in the BIOS. I’ll be trying that out shortly.

Thanks for educating me on what appears to be basic common knowledge in the hardware graphics community…

Cheers.

Actually, having the ability to simply switch to the dGPU and use it all the time is a useful feature in modern laptops, and drastically simplifies what you are trying to do (cuda programming on an optimus laptop in linux).

In windows, the mechanism to enable the dGPU “on-the-fly” is pretty well worked out. In fact, NVIDIA considers Optimus a “windows-only” feature.

In linux, without the BIOS control to be able to turn the dGPU on and the iGPU off, you are left with solutions like bumblebee, which to me are a significantly more complicated alternative. If I were doing what you are trying to do, I would thank my lucky stars that I had a laptop with that kind of BIOS-switchable option.