Poor floating point performance of Amazon EC2 c3.8xlarge machine

[ksuszka]

I've run PerformanceTest 8 benchmarks on various compute-optimized EC2 machines and it reported very poor perfomance of floating point benchmark on the most powerful EC2 machine c3.8xlarge which has 32 virtual cores of the Xeon E5-2680v2 processor.

I've also run AIDA64 floating point performance tests and three of them gave me expected results (almost linear performance boost regarding to number of cores) and fourth test gave decreased performance result, but only 16 cores were used.

So my question is - what is the reason for those results? Is PerfomanceTest 8 floating point benchmark not able to use all cores, or is there some other cause?

[David (PassMark)]

The Xeon E5-2680 v2 only has 10 cores. There is hyperthreading to get it to 20 threads, but hyperthreading doesn't give anywhere near the same performance increase as a full core.

So unless this was on a physical quad CPU system, you aren't going to be getting 32 real cores.

The floating point test will use all the CPUs and Cores available (including hyperthreading).