Did all of those trying to reduce the jitter. I think it was about the CPU clock not being stable.

Additionally I avoided core 0, because it was the noisiest and did some cgroups core pinning for the test workload.

There were zero page faults during the test runs and the CPU core was uncontested by other threads.

I think in 2008 CPUs were not so crazy about power and heat management.