Tweet
Is this STR influenced by boost technologies of the chipset or is this broadly representative of the performance of all cores on a CPU. Is this figure consistent across Single-CPU/Dual-CPU server deployments? How relevant is this rating for performance of cores in Virtual Environments for medium to large vms? Which metric should be used (STR or CPU passmark score) to determine a Dual-CPU overall rating for virtual environments hosting large vms (8+ vcpus)?
-
-
The single threaded test uses (unsurprisingly) one thread.
How the hardware deals with this depends on the hardware. On machines that have a boost / turbo clocking feature, then yes, they will score higher.
Dual CPUs sometimes suffer from NUMA issues (and are more likely to use slower ECC RAM), but ignoring this they should give similar performance to single CPU systems.
See also,
https://en.wikipedia.org/wiki/Non-uniform_memory_access
I don't know if by vms you mean VMs (Virtual machines), or if you mean something else (e.g. video management software). Or maybe the old VMS system from Digital Equipment Corporation.
As to what is relevant to virtual machines, it really isn't any different from physical machines. It depends on what software you intend to run on the machine. If you are planning on running a single threaded application (and nothing else) then the single threaded benchmark is relevant. If you are planning on running software that uses all available cores (and isn't bound by the disk or RAM) then the CPUMark is relevant. If you are planning on running multiple single threaded application and max out all cores, then it isn't really single threaded anymore.
It gets more tricky if you are planning on using just 2 cores from 32 cores however (as an example). But you can simulate this from the Advanced CPU Test in PerformanceTest.
Comment