Ansys does large use of graphics. What is more is the present opportunity to work with GPU to accelerate computation.

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=2ahUKEwjzhd-q9f_iAhWOKFAKHf2QA00QFjABegQIAhAC&url=https%3A%2F%2Fwww.ansys.com%2F-%2Fmedia%2Fansys%2Fcorporate%2Ffiles%2Fpdf%2Fsolutions%2Fit-professionals%2Fplatform-support%2Fgpu-accelerator-co-processor-capabilities-172.pdf%3Fla%3Den%26hash%3DA3C896788AC3DE769C9318C927569D8C9B98CFB5&usg=AOvVaw01343lvZ9OTcUrZyTwTXKT

Your question is not that easy to answer. ANSYS Fluent comes even with the capability to use a high-performance NVIDIA GPU (not just a simple graphics card) in order to run the linear solver for the solved equation system on the graphics processor (GPU). Several GPU's can be used (I think up to 2 or 4, but than it becomes already quite expensive in terms of hardware costs). The major benefit is, that the local GPU processing already comes with the standard license, while parallel HPC licenses are limited in the standard license and modern multi-core CPU's might have more cores than are included in the standard license. Than additional parallel HPC licenses need to be purchased in order to run ANSYS Fluent on all cores of the workstation. But that is a "luxury problem" with very expensive high-end workstations and probably not yours, if you think about the costs of a single graphics card, because it affects only latest high-performance workstations with Dual Xeon Gold (Haswell or Skylake) processors.

Otherwise for rather simple applications ANSYS Fluent runs even on a Core i5 laptop with Intel processor graphics, no problem. Not on blazing speed, but it runs. I guess, that for your case enough memory (>=16 Gb) might be more important than a high-performance graphics card.

Answer might depend on the question as well, whether you have somewhere in your surrounding another compute resource (e.g. a Linux cluster) to your disposal or not?

Regards,

Dr. Th. Frank.

Dear Dr. Thomas Frank , I want to hear Your opinion. What processor to choose for the calculation: Intel or AMD?

I agree with Thomas Frank . It is better to focus on the number of cores and threads in the processor and the amount of RAM.

Volodymyr Brazhenko : This is nowadays more a question of your budget. From my perspective the absolute latest, high-end processors are far too expensive. And with the many-cores processors CFD software usually has even another issue. CFD is a memory-bound application (besides very small and non-realistic examples). That means, that the parallel scaling of the CFD software on a single processor or compute node (usually 2-socket double-processor) might stagnate earlier, substantially before reaching the full number of available cores) because the memory bus/controller gets saturated with the data transfer of N*CFD-processes from RAM to CPU and back, so that the CPU cores need to wait instead of doing calculations. Therefore a balance between available memory bandwidth and the number of cores of the CPU is required. It can be a wise decision to buy a processor with a bit less cores and instead with a dynamic or generally higher clock speed. This is because of costs of software licenses as well, because if the software does not perfectly scale in parallel, than computation on a large number of cores on an e.g. Intel Skylake many-core CPU become unproportionally more expensive. But a general advice is not really possible without doing CFD software testing for a couple of concrete applications. I'm further hesitating to say anything about AMD vs. Intel, in particularly because I have not touched any AMD CPU based computer with the ANSYS CFD software for the last at least 6-7 years.

Regards,

Dr. Th. Frank.

Thomas Frank thank you for answer