0

I'm struggling to decide whether to create a new developer PC setup based on recent AMD or Intel CPUs because i cannot find any specific performance measurements focused on compilers and linkers; i'm specifically interested in MSVC and C++ on Windows.

My major concern is that the Intel efficiency cores might not perform well in heavily parallelized build scenarios and comparable AMD CPUs would perform better, but that might be wrong.

Are there any performance comparisons available? Are there recommendations?

Boundary:

I know about distributed build systems like Incredibuild or FASTbuild but this question is purely about single systems.

Also it should be assumed that besides the CPU (and all what depends on it) as many system components as possible should be comparable or equal.

3
  • Intel E cores are pretty strong although I still think AMD is faster in multicore stuff like compiling C (Linus Torvalds also has Threadripper)
    – Irsu85
    Commented Oct 5, 2022 at 10:13
  • In a purely theoretical sense I believe the thread rippers benchmark higher than intel competitors for pure multi core scenarios, that being said I'd expect the storage device to be a greater bottleneck for compilation. I personally compile code to a RAM disk. I suspect the parallelization difference is going to heavily depend on the implementation of the compiler itself. Commented Oct 6, 2022 at 2:18
  • @BennettYeo The storage device is definitively a important component and a decent NVMe shows massive improvements compared to a SATA SSD but for the comparison it can be assumed that both systems have the same drives, comparable memory, ... Last time i compared with RAM disks there wasn't such a massive improvement, but it's been quite some years ago. The codebase i'm working with is huge though. It's not like llvm or Unreal but i have to compile the two on a regular basis as well using MSVC. Commented Oct 6, 2022 at 9:38

1 Answer 1

2

Are there any performance comparisons available? Are there recommendations?

We can estimate the performance based on the numbers from timed compilation benchmarks on openbenchmarking.org. CPUs may be compared side by side, e.g. AMD Ryzen 9 7950X vs Intel Core i9-12900K.

My major concern is that the Intel efficiency cores might not perform well in heavily parallelized build scenarios ...

In my opinion based on the results, 12900k had been one of the fastest consumer CPUs before high-end Ryzen series was released recently.

More information about a particular CPU of interest may be found in benchmarks scattered over the internet, e.g. a note about software and game development performance of 7950x.

Regarding the upgrade possibilities, I'd be looking forward to the upcoming Raptor Lake release at the end of October, expecting Intel CPUs to be at least as performant as recent AMD CPUs.

Note that Raptor Lake P-cores lack AVX-512.

1
  • Good find. To add, AVX registers probably don't matter that much for code compilation, as far as I know, I haven't heard of compilers heavily leveraging SIMD vectorization. Commented Oct 11, 2022 at 15:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.