CPU Multiplier Overclocking

This experiment is a follow-up to the anomaly discovered in my original BCLK overclock attempt: it appeared as though manually increasing the CPU multiplier and memory speed by 10% increased mypy performance by 4%. My memory tuning experiment didn't show such large gains from simply overclocking memory to >3250MT/s which led me to believe there were some gains from the CPU multiplier overclock.

For this experiment I used three configurations:

• Stock settings (34x CPU multipler for 3.4GHz base speed)

• 35.75x CPU multiplier for 35.75GHz base speed (+5.1%)

• 37.50x CPU multiplier for 3.75GHz base speed (+10.2%)

Other than the CPU multiplier, the configurations used default BIOS settings with 3200MT/s memory speed.

Results

With the exception of ripgrep, all benchmark results were less than 1% different from each other. Ripgrep took a massive hit of 7.2-7.7% slowdown for the overclocked configurations. I don't have any hypothesis why this occurred but it seems unlikely to be a random coincidence as both OC configurations experienced similar slowdown even though they ran at different times of the day.

This is actually a little suspicious since ripgrep is the one benchmark I would expect to see improved by a base clock OC because it actually uses all cores and isn't reliant (that I'm aware) on single-core boost speed.

Conclusion

Increasing the base CPU multiplier doesn't seem to have improved any benchmarks. The ripgrep performance decrease and possibly warrants more investigation - is it possible the CPU had insufficient voltage to maintain the all-core OC?