Hacker News Reader

It’s good to know about turbo boost when measuring cpu bound code. When the cpu gets hot, the clock drops, sometimes by a large factor. When it’s cooler and the OS wants to run code, the clock goes back up. This can be disabled in BIOS to get a more stable result. You can monitor the cpu clock in Resource Monitor on windows 10, to see how big a deal the issue is. It varies by cpu model, temperature, cooling of your particular machine etc.

View on HN

This is one of the things that people are thinking about when they ask you to rearrange your benchmarks and run them again. Sometimes the first or the last set run is the fastest, depending on how the confounding factors interact.

Extraordinary claims require extraordinary proof, and people can be rightly insistent.

...but obviously only disable it if it is disabled in the environment which you want to predict. Don't try extrapolating from an environment in which it is disabled to one in which it is not, and vice versa.

Not really true. If the particular computer you’re running the code on is warm due to weather conditions, which affects clock frequency, it doesn’t tell you anything about the performance of your code.

It depends on what you are trying to do.

Most of these benchmarks are not done to predict performance in a target machine, but to find out which is faster.

Reply on news.ycombinator.com