wat u ppl would suggest in terms of computer power and company for running fluent 6.2 for turbomachinery . currently i am using mixing plane and simulating more than 2 stages

Here is a reference point for you:

Computer: dual Opteron 246, MB Tyan K8W, 4GB Corsair

OS: Windows 2000 Professional

Software: Fluent 6.2

Case: Rotor 37, 420.000 elements structured grid, 2 Metis partitions

Solver: Coupled Implicit

Performance: - serial 22 sec/iteration - parallel 15 sec/iteration

Testing the same case with 64 bit Fluent on WhiteBox Linux 4 I get a 27% improvement for serial case.

BUT! Pay attention that this is a case where data transfer between processors is pretty large compared to the size of the grid (6.7 MB/iteration, value equal to that obtained for a 2.1 milion!!! cells segregated incompressible case, for which parallel performance was 110% vs 64% for current case) and this is mainly due to the coupled solver. For a segregated solver or a coupled explicit solver things are a little brighter (these ones do not require such intense data transfer between processors). So, I'd say that for 3-4 stages coupled through mixing plane interfaces, 1.5-2 milion cells maybe, you would need a minimum of 4 processors cluster (my suggestion: AMD64 3500+, socket 939 AT LEAST, 2GB RAM/processor, GigabitEthernet connection) running 64 bit Fluent on Linux. With a good partitioning you could get less than 25 sec/iteration solving with coupled implicit.

Best wishes, Razvan

thanks for ur suggetion.

sam