Can anybody recommend me a Linux distribution for 64 bit version of Fluent 6.2? Will Fluent_install-lnamd64-6.2.16.tar be good for Intel 64 bit? I have Core 2 Quad QX6700, 4GB RAM, Asus P5B mainboard. 32 bit version of Fluent 6.2.16 runs fine under Windows, but I need to solve cases exceeding 2GB.

It worked for me using Fedora Core 4, with the one qualifier that I had to choose the "install all packages" option in order to get all the necessary libraries.

I didn't try it on a Core 2 though, but it worked on AMD Opteron, Intel Xeon 3.6 (single core) and Intel 830D dual core. I will be interested to know how well the parallel scaling works out for the quad-core. From SPEC figures it seems as if the quad core doesn't scale that well beyond two processes, but it would be great to get some real CFD performance data. Please let us know how well it goes ....

1) Performance for 400 000 cells, segregated solver viscous incompressible species transport Win XP 32 bit: 2 procesors - 176% (88% per processor) 3 procesors - 248% (83% per processor) 4 procesors - 287% (72% per processor) â€" So performance drops at 4, then it is difficult to communicate with the system. In fact there is one more process supervising, and the system also absorbs some resources, which could be a reason. I will let you know if there is a difference with larger cases, as four cores share the same memory controller. 2)Could you post results of your scaling benchmarks? 3) Could ANYBODY ELSE share his experinece with linux64 bit and Core2duo/Quad? 4) Is there any special version of FlexLm for lnamd64 or lnx86 will do?

I use Suse 10.1 and works fine for me.

I use Red Hat Enterprise Linux Server v4.2 with 2 dual core amd opteron processors (lnamd64). Works great for me. Machine has 16 GB of RAM and I use every bit of it. No problems, good performance.

Dual core machines scale OK, but it gets worse for quad core machines. The problem is that all the cores on a chip use the same memory bus - ie all data that passes to and from all four cores goes through the same memory bus. Hardware vendors are adding more cores, but not speeding up the memory bus at the same rate, so for memory intensive programs, like CFD, the cores end up waiting because they can't send and receive data fast enough. You can get quad core machines to scale well, but only by using one or two cores at any one time, which kind of defeats the point!