CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Bechmarking CFD codes (http://www.cfd-online.com/Forums/main/2259-bechmarking-cfd-codes.html)

Olesen June 19, 2000 10:07

Bechmarking CFD codes
 
Hi.

My company is interested in purchasing a piece of CFD software. We are mainly interested in the three products: Fluent, Star-CD and CFX. Is anyone aware of, if any benchmarking has been performed, between those three products with respect to solver speed. I would expect, that the result of such a benchmark depends greatly on the problem which is solved, but still it would be nice to have some numbers.

Regards

Niels

John C. Chien June 19, 2000 11:26

Re: Bechmarking CFD codes
 
(1). It does not make sense to know the solver's speed, without knowing the problem to be solved. (2). For simple problems, it shouldn't matter which code to use. For complex problems, in most cases, I don't even know whether a particular run will give me a converged solution or not. (3). This is my everyday experience. I have to monitor the solution and constantly adjust the parameters (relaxation factors or time steps) in order to guide the solution to a converged state. Sometimes the solution diverges in one step, sometimes in the middle of iterations. (4). And there will be many other issues, like the time involved in geometry and mesh generation, post processing time to look at the result, and the re-meshing time to improve the mesh and the solution, etc.... (5). I would strongly suggest that (a). visit the vendors's website first (it's free, but it will cost you time), and make your own judgement in terms of your applications. Most people don't want to tell others what he is going to do with the code. (b). ask the vendor to supply you the information you need, (c). ask the vendor to benchmark your problem. or get a demo disk to run your problem on your own. (6). I can only say that 99% of time, I am getting the wrong results, in terms of the mesh generation, and solutions. It is easier to use a faster computer than to find a faster code. A general cfd code is always relatively slow when compared with the specialized design code. (7). So, make sure that the code you are going to buy can solve your problem in the first place.(especially the solution you like. )

Clinton Lafferty June 19, 2000 12:15

Re: Bechmarking CFD codes
 
I would suggest the following paper. It details the pros and cons of several commerical codes available; however, it is 5 years old. I have been looking for any follow-ups.

Perspective: Selected Benchmarks From Commerical CFD Codes. Journal of Fluids Engineering. June 1995. Vol. 117. pp208-218.

An extended abstract can be found through the ASME website. http://www.asme.org/


All times are GMT -4. The time now is 12:51.