About Maxwell GPUs


With the development of NVIDIA GPUs, Maxwell GPUs have became the most popular GPUs in super computing and entertainment fields. New GPUs such as Tesla K80, Geforce GTX 980, GTX 970, GTX 980m, GTX 970m, GTX 860m(Maxwell version), and GTX 850m have been proposed.

The CUDA C toolkit has support the Maxwell GPUs well for some years. Unfortunately, CUDA Fortran and OpenACC do not officially support Maxwell GPUs. This is a big disadvantage!

I want to know does PGI plan to support the Maxwell GPUs in the future and when?

Thank you!


Hi Nightwish,

PGI supports the Tesla line of NVIDIA devices. The Telsa line focuses on the computing market for which our products are designed. The GeForce lines are graphics cards designed for gaming and video. In the past, this wasn’t an issue since they shared the same underlying architecture. However no Telsa version of Maxwell has been released and why PGI does not currently support it.


Thank you Mat!

But the Tesla K80 may be released soon, this GPU is based on Maxwell architecture. I wish PGI may support the Maxwell to keep pace with the NVIDIA Tesla K80.


Hi Nightwish,

Now that K80 has been announced I can confirm that it uses a Kepler architecture so you will have no issues running PGI code on it.

Hope this helps,

OK, so with my new GTX 970, I will not get much benefit from a PGI compiler? I had hoped to be able to write Fortran applications that could use the GPU. But from this it looks like I can just as well use Visual C++ and forget PGI. Or is there a way to run Fortran code on the 970?