[Gate-users] Predicting GATE performance with SPECfp?
Porter Ian (RBV) NHS Christie Tr
Ian.Porter at christie.nhs.uk
Wed Nov 18 18:37:17 CET 2015
I am producing a specification for a new cluster to run GATE, with the aim to have four times the compute performance of our existing cluster, and to use more modern CPUs. I am assuming that there will not be a multi-threaded version of GATE before we purchase the hardware, so I am looking at buying multiple CPUs and running one instance of GATE per CPU core. Therefore, I need to find a measure of the performance of current CPUs relative to those in our existing cluster, so that I can then choose the best CPU based on price/performance. I have not been able to find any benchmark information for GATE/GEANT on different CPUs, so instead I am using SPECfp_rate2006 results downloaded from spec.org, as my hunch is that this is the most appropriate metric to use.
I am not familiar with the kind of operations performed by GATE/GEANT, so I don't know if my hunch is correct. Is anybody able to say whether SPECfp_rate is a good indicator of GATE/GEANT performance in this scenario? Our calculations typically take around an hour to complete, and I am assuming that disk read/writes are a negligible part of the overall compute time. Any guidance would be gratefully received.
Ian Porter, Dept. Medical Physics,
Christie Hospital NHS Foundation Trust,
Wilmslow Road, Manchester, M20 4BX. England.
This e-mail and any files transmitted with it are confidential and solely for the use of the intended recipient.
If you have received this e-mail in error you should not disseminate, distribute or copy it.
Please notify the sender immediately and delete this e-mail from your system.
More information about the Gate-users