[grtabor]

Hi. I'm looking to purchase a new computer for CFD and am looking at the options. I'm seriously considering a dual-processor, twin-core Opteron machine. Does anyone have experience with running OF on dual-core machines? My limited knowledge would suggest that such a beast would give me the possibility of running in parallel on 4 processors (twin cores functioning as separate processors) but things don't always work out like that.

I've come across a few interesting links which I will reproduce here... tests on dual core Opteron and comparisons between Opteron and Xeon machines. Quite a few are using Fluent as benchmarking software, which is of interest as the issues this highlights are probably appropriate to all CFD codes.

Opteron/Xeon comparison:
http://www.xbitlabs.com/articles/cpu...rkstation.html

Dual core processors:
http://enterprise.amd.com/Downloads/...en.pdf#search='fluent%20opteron%20dual%20core'

Fluent on single+dual core processors:
http://www-03.ibm.com/servers/enable...42.pdf#search='fluent%20opteron%20dual%20core'

(hope the links work)

Gavin

[aix]

Hi, did you read http://openfoamwiki.net/index.php/Benchmarks_stan dard_v1
Dual Opteron 250, 2.4 Ghz, 4GB, Linux and Dual Opteron 252, 2.6 Ghz, 4GB, Linux so far in the Benchmarks, but no dual-processor mobo twin-core Opteron

[jens_klostermann]

We have a Dual Opteron twin core system (2xOpteron 280) and a Dual Opteron (2xOpteron 252) system in our team.

I did run the Benchmark (http://openfoamwiki.net/index.php/Benchmarks_stan dard_v1) on all available cores of each system, so far the speedup is 1.3 to 2 from 2xOpteron 252 (running the benchmark with OF 1.2) to 2xOpteron 280 (running the benchmark with OF 1.3). But since there is quite a difference between OF 1.2 and OF 1.3 (see the wiki) I will publish the benchmark results for both systems running with OF 1.3 as soon as I've got a little spare time.