Given the number of postings related to HPC (Hardware) I would like to propose that a new forum be setup for HPC.

The forum may be active for a few days then become dormant like most of the other forums.

look at: http://www.beowulf.org/

> Given the number of postings related to HPC (Hardware) I would like to
: propose that a new forum be setup for HPC.

There are not many posts on this topic that I have seen and very much doubt it would reach critical mass.

However, I would suggest that what does stand a much bigger chance of being successful and a lot more useful when it comes to hardware choices is to adopt/create a CFD benchmark suite. Many of the enquirers about hardware betray a lack of understanding of what is important and what is not when it comes to CFD and such a resource here could help address this.

By far the most useful benchmark for CFD that I have seen is the NASA NAS Parallel benchmark suite and yet it is rarely mentioned in these pages. To put up sets of plots for various hardware and, more importantly, supporting text explaining what one can see in the plots, why and what it means for related types of hardware would not require much effort to reach the point of being useful.

Benchmarks like this are poor for marketing purposes because there is far too much information which requires interpretation. However, for CFD they are very useful and, in some instances, even better than benchmarking your own code if the types of algorithms run on the machine is likely to evolve over the life of the hardware.

I agree with you that a dedicated HPC forum is not needed here. Before we see a lot of HPC discussions on the main forum it makes no sence to create a dedicated forum for it.

About benchmarks. I actually think that the Fluent benchmarks are quite good. They are failry up-to-date and covers most hardware platforms. They also have different types of cases (small, medium, large, incompressible/compressible, ...)

Otherwise it would ceratinly be very valiable to have a CFD benchmark in CFD-Wiki. We have already started to build a test-case and validation suite, but it could be complemented with a special benchmark also I guess. Anyone have a good suggestion on a benchmark which could be distributed freely under a GNU free documentation license?