CUDA 9.0 deviceQuery fails: "CUDA driver version is insufficient for CUDA runtime version"

After following this guide to the dot (http://developer.download.nvidia.com/compute/cuda/9.0/Prod/docs/sidebar/CUDA_Installation_Guide_Linux.pdf) to install CUDA 9.0, I attempted to build the samples at `NVIDIA_CUDA-9.0_Samples/, but this resulted in a make error.

I followed a suggestion to build just deviceQuery, but when I run ./deviceQeury, I get the following output:

krishna@krishna-ubuntu:~/Documents/NVIDIA_CUDA-9.0_Samples/1_Utilities/deviceQuery$ ./deviceQuery 
./deviceQuery Starting...

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

cudaGetDeviceCount returned 35
-> CUDA driver version is insufficient for CUDA runtime version
Result = FAIL

Note that I’m running a dual boot on a laptop that also has an iGPU.

Is this error normal, and if not, how can I fix it?

update the GPU driver from a proper NVIDIA source (not ppa)

Do I need to purge CUDA before updating my GPU driver?

I installed CUDA via sudo apt-get install cuda.
Does this not install the required GPU drivers as well?

it should. Your driver install may be broken. Laptops can be tricky to get working correctly. If the dGPU is disabled dynamically, you’ll have this trouble.

Thanks.
Is there any official resource I can follow to make sure I install the GPU driver the right way?

The GPU driver has a readme file. For example:

http://us.download.nvidia.com/XFree86/Linux-x86_64/390.67/README/index.html