Has anyone used ArcMap with a GPU accelerator such as the Tesla K20?

[karldailey]

Has anyone used ArcMap with a GPU accelerator such as the Tesla K20?  I'm curious if there is a significant performance boost on complex tasks, and if so how much.  There was a blog post in 2010 on the subject matter, but I am unsure if he achieved such boosts without using special enhancements to the system.

example product:
http://www.nvidia.com/object/tesla-workstations.html

2010 blog post:
http://blogs.esri.com/esri/apl/2010/03/30/computations-on-vector-data-using-a-gpu/

[JohnMeza · Accepted solution]

Hello,
I'm not sure if you refer to a portion of functionality in ArcMap or ... ArcMap.
There are performance boosts for specific tasks run in GPU accelerated/GPGPU but those are very specific tasks.

Not all computational problems benefit from GPU accelerated/GPGPU development. The computational problems best suited for GPGPU are those take advantage many parallel streams of processing on individual cores within a GPU. A CPU has alot of capabilities that an individual GPU doesn't have, including interface with the OS, other devices (memory, storage) on or attached to the motherboard.

You mentioned ArcMap which is a big Windows app that does alot of tasks. Most of those tasks are not suitable for the type of processing suited to a GPU, but suited to a CPU and all of the resources it manages.

As you can tell from the blog you referenced, it took some effort to optimize a small portion of functionality. As far as ArcMap directly from a GPU accelerator, no. The user has to target specific areas of functionality that are suitable based on characteristics of the processing and data.

[karldailey]

Thanks John, thats the information I was looking for.  I thought it might be the case but wasn't sure