Future ANSYS support for CUDA 11.0

I am part of a student design team in the market for a GPU accelerator for use in Fluent (with Fluent Meshing) for coupled systems with about 12-30 million elements. It has occurred to me that the Nvidia Tesla K80 has significantly reduced price that appears to tick all of our need boxes. But, there is one caveat: the Tesla K80 contains Kepler architecture which Nvidia is discontinuing CUDA support/development after CUDA 11.0.

I do see that there is current support right now for GPU Accelerator Capabilities for ANSYS 2020 R2 for the Tesla K80, but my question is: how long can I expect future versions of ANSYS to support CUDA 11? As our student design team receives access to the newest ANSYS every year and always look forward to CFD related updates (like the recent major Fluent upgrade), I would to know if we could get at least 3-4 years of use out of a Tesla K80 before we are forced to either purchase a newer graphics accelerator or withhold from using the newest ANSYS.

Thank you

Tagged:

Answers

  • mmadoremmadore Forum Coordinator
    edited September 23

    Hello,

    I will check into this with our dev team and get back to you.

    Matt

  • mmadoremmadore Forum Coordinator

    @BFR_FSAE Here are some comments:

    1. Never a good idea to invest in a technology that is on its sunset. K80 is really old and as Nvidia is planning on not supporting it, it is better not to buy one now and expect years of support and performance.

    2. In the fluent solver there is a limited set of capabilities that are GPU accelerated. The performance boost depends on the the case and of course on the GPU. K80 is so old that you probably will see better performance on your cpu anyway.

    You should really consider these things before investing in an outdated GPU. Hope this helps.

Sign In or Register to comment.