Benchmarking Activities

The CERN IT techlab project hosts a variety of different processor architectures (ARM, POWER8, NVIDIA and AMD GPUs, etc.). https://twiki.cern.ch/twiki/bin/viewauth/IT/TechLab For each architecture measurements of HEPSPEC06 and power consumption are done. The 2006 systems cannot compete with the standard INTEL servers CERN is buying for their compute farms. The price/performance and power/performance is factors (2-4) worse.

The usual SPEC benchmarks can be found here: https://www.spec.org/benchmarks.html

The HEP benchmark (HEPSPEC2006) is based on the integer part of the SPEC benchmark. More details about the usage can be found on this web page: http://w3.hepix.org/benchmarks/doku.php?id=homepage A HEP benchmarking group has been established with a focus on how well the 'artificial' benchmarks are representing the performance of the actual experiment event processing codes. latest status report at the fall HEPIX 2016: * HEP__benchmark_group_status-report-2016-10-19.pdf:

A large (graphical) collection of SPECint_rate measurements (AMD, INTEL, Sparc, IBM) over the last years can be found here: https://public.tableau.com/profile/karlarao#!/

-- BerndPanzerSteindel - 2017-01-18


This topic: Main > TWikiUsers > BerndPanzerSteindel > TechMarketPerf > BenchAct
Topic revision: r1 - 2017-01-18 - BerndPanzerSteindel
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback