[Beowulf] Better C2D or Quadcore
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Chris Samuel csamuel at vpac.orgTue Nov 27 22:38:24 PST 2007
- Previous message: [Beowulf] Better C2D or Quadcore
- Next message: [Beowulf] Better C2D or Quadcore
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Wed, 28 Nov 2007, amjad ali wrote: > high per-CPU memory performance. Each CPU (core in dual core > systems) needs to have its own memory bandwith of roughly 2 or more > gigabytes. If this is indeed the case for your problems then you might find that quad core systems don't cut it - a contact of mine at another institution asked me to do some Ramspeed benchmarks to compare our 1.9GHz Barcelona test system to their Clovertown cluster nodes. They were measuring about 5GB/s aggregate RAM bandwidth when running the same test on all 8 cores (around 650MB/s per core) whereas we were seeing over 12MB/s aggregate. Yesterday we upgraded that to a 2.1GHz test system and the measured bandwidth increased to just over 13GB/s. The nice thing about this test is that the benchmark code is written in assembler so the compiler doesn't matter. Running the streams benchmark in OpenMP mode with 8 threads agrees with those numbers on the 1.9GHz Barcelona (when compiled with Intel or PGI, much lower speeds with GCC). I don't have a note of the numbers for the 2.1GHz chips with me on the train. PGI 7.1-1 : pgcc -mp -fastsse -tp barcelona-64,k8-64 -Mipa=fast -o stream ./stream.c Function Rate (MB/s) Avg time Min time Max time Copy: 12261.7219 0.1044 0.1044 0.1045 Scale: 12284.4198 0.1042 0.1042 0.1043 Add: 12213.4179 0.1574 0.1572 0.1581 Triad: 11875.6082 0.1619 0.1617 0.1621 Intel v10 : icc -static-intel -openmp -O5 -o stream-icc stream.c Function Rate (MB/s) Avg time Min time Max time Copy: 12214.5757 0.1048 0.1048 0.1050 Scale: 12155.7238 0.1054 0.1053 0.1055 Add: 12243.4062 0.1569 0.1568 0.1570 Triad: 12240.2984 0.1569 0.1569 0.1570 cheers, Chris -- Christopher Samuel - (03) 9925 4751 - Systems Manager The Victorian Partnership for Advanced Computing P.O. Box 201, Carlton South, VIC 3053, Australia VPAC is a not-for-profit Registered Research Agency -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: This is a digitally signed message part. Url : http://www.scyld.com/pipermail/beowulf/attachments/20071128/2afc801b/attachment.bin
- Previous message: [Beowulf] Better C2D or Quadcore
- Next message: [Beowulf] Better C2D or Quadcore
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
