How to configure Intel oneAPI to use NVIDIA GPU drivers?

Accelerated Computing CUDA CUDA Programming and Performance
1

On my W10 notebook I have NVIDA GPU drivers (NVIDIA GeForce GTX 960M), Visual Studio 2022 where I can program and run the CUDA program (CUDA 11.8 Runtime - VS22 template) without any problems.

Then I installed the Intel OneAPI Basic Toolkit and run the first SYCL example program (Base Vector Add):

Running on device: Intel(R) HD Graphics 620
Vector size: 10000
LLVM ERROR: SPIRV internal error.

The oneAPI mentioned in Supported Architectures and Terminology:

Graphics Driver Installation:

Intel® oneAPI 2023.0 is validated on Windows and Linux for the following driver versions

Windows Intel® Graphics Driver
To install the driver follow the directions in the article appropriate for your device
    Intel® Iris® Xe MAX (DG1) and 10th-13th Gen Intel® Core™ Processor Graphics https://www.intel.com/content/www/us/en/download/19344/intel-graphics-windows-dch-drivers.html
    Intel® Arc™ A-Series Graphics (DG2) https://www.intel.com/content/www/us/en/download/726609/intel-arc-graphics-windows-dch-driver.html
    Intel® Data Center GPU Flex Series (ATS-M) https://registrationcenter.intel.com/en/products/subscription/956/
    Contact your OEM representative for access to the Intel Registration Center.
Linux General Purpose Intel GPUs (GPGPU) Driver
For all Intel GPUs, see this article, https://dgpu-docs.intel.com/, and follow the directions for your device.

Doe’s anybody knows how to configure oneAPI in VS22 to use NVIDIA GPU driver?

2

Further up that page it also mentions:

GPU analyses

  • All GPU analyses are supported on Intel® processors with 9th generation of Intel HD or Iris Graphics (formerly Skylake) or newer.

Intel are a direct GPU competitor with Nvidia, so it’s unlikely they support the use of Nvidia products with their tools.

3

Is there any other possibility to create SYCL program for NVIDIA GPU? What would you recommend?

4

SYCL isn’t a NVIDIA product. It’s also not CUDA or CUDA programming. However I’m fairly certain that SYCL code can be run on NVIDIA GPUs in some circumstances, so I think google is your friend, if you don’t get any advice here.

https://www.intel.com/content/www/us/en/developer/articles/technical/compiling-sycl-with-different-gpus.html

5

Hi Robert,

Many thanks for your advice. I will have a look on it.

To my surprise I am running SHA256 under wsl on my notebook with NVIDIA GeForce GTX 960M with the speed of about 2.5 GH/s.

On the other side when I run HMAC-SHA1 and PBKDF2 as C++ VS22 CUDA program fully on the device I will get just about 700 H/s.

That’s why I am interested in SYCL to try to speed up the HMAC-SHA1 a PBKDF2 calculations.

6

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.