Hacker News Reader

Comment by andy_threos_io | Hacker News Reader

I checked this geekbenchmark with our several different computers on hand, and I can confirm that it's total useless measurements for real world applications or performance.

View on HN

Torvalds ripped it apart nearly a decade ago.

It's a useless benchmark, what I want to see is things like time to compile a bunch of different software, things that take long enough for the processor/cooling to reach thermal equilibrium etc.

I.e. stuff that more closely matches the real world

He ripped apart a very different benchmark for what it was worth, that was GB3 at the time I believe. 5 was a rewrite to make it cross platform-equal. In real world use it actually is far more relevant than thermally limited benchmarks. It just measures max peak burst performance...which is important because 90% of all users use their computer to do only bursty tasks rather than long term processing. See exporting a 10 second clip on an iPhone or loading a heavy SPA webpage on a Mac. These are 5 second burst tasks where real world use would not be thermally limited but would see real change consistent with Geekbench.

It's really only intended to be one of many benchmarks to tell the whole story; of course Linus would attack that because it doesn't make any real sense in his use and isn't the full story for him. If Geekbench was not tested, it would not cover the majority of computing uses and it would weigh cpus that had poor turbo or burst performance unfairly high for most uses.

Geekbench is kinda like 0-60MPH times and other tests (like SPEC2006) are like top speed I guess? The whole story is between them.

I believe this is the discussion OP is talking about: https://yarchive.net/comp/linux/benchmarks.html

>There’s been a lot of criticism about more common benchmark suites such as GeekBench, but frankly I've found these concerns or arguments to be quite unfounded. The only factual differences between workloads in SPEC and workloads in GB5 is that the latter has less outlier tests which are memory-heavy, meaning it’s more of a CPU benchmark whereas SPEC has more tendency towards CPU+DRAM.

https://www.anandtech.com/show/16226/apple-silicon-m1-a14-de...

Reply on news.ycombinator.com