Supported accelerators

Accelerated Computing HPC Compilers Legacy PGI Compilers
1

Q Which accelerators can be targeted by PGI Accelerator compilers?

A PGI Accelerator compilers target all CUDA-enabled NVIDIA GPU accelerators with compute capability 1.2 or higher. PGI is studying the feasibility of supporting other accelerators.

from the PGI Accelerator Compilers page.

In practice, doesn’t this mean that only the following are supported?

  • Quado Plex 2200 D2
  • Tesla S1070
  • Tesla C1060
  • Quadro FX 5800
  • Quadro FX 4800
  • Quadro CX
  • GeForce GTX 295
  • GeForce GTX 285
  • GeForce GTX 280
  • GeForce GTX 275
  • GeForce GTX 260

Also, there are no GPUs that support v. 1.2, just 1.0, 1.1 or 1.3.

The compute capability of Nvidia’s cards is not easily found, I had to go digging to generate the above list.

2

Hi gormanly,

Thanks for the note. I’ll ask our web master to fix this page since, as of the 9.0 release, we do support compute capability 1.0 and 1.1, as well as 1.3. However, the default is to use 1.3. For earlier versions you either need to use the compile option “-ta=nvidia,cc10” or “-ta=nvidia,cc11” or set in your “sitenvrc” file “set COMPUTECAP=10;” or “set COMPUTECAP=11;”.

As for a list of supported NVIDIA cards, the basic rule is that if the card supports CUDA, then the PGI Accelerator model should work as well. Granted, we haven’t tested every card, here’s NVIDIA’s list of CUDA enabled GPUs
http://www.nvidia.com/object/cuda_learn_products.html

Hope this helps,
Mat