Meik Frischke wrote: > Am 2024-04-17 12:33, schrieb Dale: >> I found a benchmark website that compares the two. Link below. It >> claims about 80% faster. In some ways, twice as fast. Sometimes those >> bench tests don't reflect the real world to well. Most of them seem to >> test gaming speeds which isn't of much use anyway for me. I'm more >> about compiling and such. Just wondering how much speed difference this >> would make. Maybe someone reading this did a similar upgrade or has >> seen both in action. If so, post and share your thoughts. >> > > Hi Dale, > > since Moore's Law isn't quite dead yet there is a significant > performance uplift in newer processor generations, especially with the > smaller 5nm process nodes of recent, after some years of stagnation at > 14nm (your FX-8350 was manufactured at 32nm). With each process shrink > (32nm -> 28nm -> 22 nm -> 14nm -> 10nm -> 7nm -> 5nm) new CPUs can > deliver higher performance with the same power consumption or achieve > similar performance levels with lower power consumption. > Looking at the open-benchmarking default configuration kernel compile > benchmark (pts/build-linux-kernel-1.15.0), the Ryzen 5 7600 (slower > non-X) took ~101s to compile the kernel (based on 28 submitted > results) while the FX-8350 took ~422s for the same task (based on 4 > submissions) [1]. Unlike gaming, compiling tends to scale quite well > with core count and for the gentoo use-case the measured performance > difference is in most cases similar for different packages. There are > many influencing factors for benchmarking like running kernel version, > activated options and mitigations so YMMV, but you can test it > yourself: there are ebuilds for the phoronix-benchmark-suite in > various overlays [2]. You can perform the benchmark with > $(phoronix-test-suite benchmark pts/build-linux-kernel-1.15.0) with > the "defconfig" test configuration option. > > Cheers, > Meik > > [1] > https://openbenchmarking.org/test/pts/build-linux-kernel&eval=9cdcd82c9c47af9df17263e4312f634338dbf476#metrics > [2] https://gpo.zugaina.org/app-benchmarks/phoronix-test-suite > >
If one just compares the kernel compile time, about 4 times faster. I'm not expecting the accuracy one needs to build a space ship. ;-) That's a pretty good way to measure because with Gentoo, compiling a kernel is a very common thing. As you said, it scales well. Compiling gcc would be a good one to if they have default USE flags. Obviously if one system has a lot of USE flags enabled and another is the bare minimum, there will be a difference not related to CPU speed. Rich made a good point too. Speed isn't just influenced by the CPU. Memory speed and even the speed of accessing data drive, spinning rust, SSD or whatever, also affects a system. When I get this new rig built and you see me post about it, remind me and I'll install that benchmark test and send in the results. I like doing things like that because it helps others too. I just wish there was a centralized place for them all. Right now, there are likely dozens of them and each with their own method. Thanks for that info. I'm making progress and planning a way to purchase all this. It's still not cheap but cheaper than it was before I found out I could get a cheaper CPU and upgrade later. Dale :-) :-)
