15-Way Linux Distribution / Operating System Comparison, Including Windows 10 & WSL

[Danielsan]

Michael and andyprough

Because "data per se" doesn't say any much... At this point all the Distros but Windows are winners

[ruthan]

No Windows 7 for comparision, that WIndows 10 update half year cycle is killing them.. To be fair, some games (real games on Open Arena which are poorly written for Window) should be added, because Windows are for games and Visual Studio.. and maybe MacOS make sense for user interface friedlyness..

[gadnet]

that could have been interesting to have freeBSD in it, not linux but hey the window linux thing do not really smell like it too and it would be interesting to see where it stand.

[Michael]

that could have been interesting to have freeBSD in it, not linux but hey the window linux thing do not really smell like it too and it would be interesting to see where it stand.

As already stated, BSDs will be in their own article next week with these numbers.

[deant]

As already stated, BSDs will be in their own article next week with these numbers.

But I want to see how BSDs compares vs Windows and Linux. Maybe you should do that too!

[Michael]

But I want to see how BSDs compares vs Windows and Linux. Maybe you should do that too!

Right, the article coming shows that.

[austin754]

Observations on a few of the workloads used in this test:

1. cachebench is no longer measuring what the test was originally written, i.e. the performance of the hardware caches. It looks like the benchmark was originally written ~20 years ago when several things were different including smaller cache sizes, much less sophisticated hardware prefetching and less sophisticated compiler optimization. You can see this because the "read" test is a much lower score than the "read/modify/write" score you've included in the article. It is also indicated by such a wide variation among operating systems on a benchmark supposedly measuring a hardware feature on the same hardware. So I'd put this one in the "synthetic" category since it is small, artificial and not necessarily measuring either a hardware attribute or a real program. I suspect a combination of compilers optimizing away code and hardware prefetching to be distorting these results.

2. blake2 is a single-threaded benchmark that runs very quickly, less than 0.5 seconds on my hardware which is less powerful than the i9 used here.

3. The ffmpeg workload file used is likely not well-tuned for a processor with some many cores. The test script passes in a "-threads" option to ffmpeg On my system at least, I get flat performance and then diminishing returns once this parameter gets to "2" and then drops later. Not completely certain all things clearlinux has done (I suspect some tuned versions of software components), but I suspect a contributor of why this shows 2x improvement on clearlinux is that the ffmpeg workload chosen is to some extent "misconfigured" for taking advantage of the hardware.

4. There is a minor issue with build-linux-kernel that shouldn't affect the overall results. The first iteration of the benchmark it runs ~29,000 processes. In the subsequent iterations only ~26,000 processes are run. I suspect some aspect of a "make clean" isn't fully done and hence iterations after the first one have less work to do. This is at least consistent between what you are measuring, so should come out in the wash. What is curious however, is that the build-php was reported to have slower compile times due to compiler versions, but these didn't carry across to build-linux-kernel.

--mev

[joebonrichie]

So i did some testing oh how clearlinux got that crazy zstd compression result. Basically, they sort of cheat, they have a patch which always enables multithreading, see https://github.com/clearlinux-pkgs/zstd/blob/master/multi-thread-default.patch. I did some testing on Solus enabling various optimisations e.g. PGO, speed flags (e.g. 03), funroll-loops, etc. which only had a negligible performance increase for the system/compress-zstd benchmark.

https://openbenchmarking.org/result/1806013-FO-ZSTDNOOPT82

[Vistaus]

No Windows 7 for comparision, that WIndows 10 update half year cycle is killing them.. To be fair, some games (real games on Open Arena which are poorly written for Window) should be added, because Windows are for games and Visual Studio.. and maybe MacOS make sense for user interface friedlyness..

I've used macOS a few times in the past (last time was little over 2 years ago) and not once did I find it user-friendly. Pretty? Yes. Usable? Yes. But user-friendly? No. Maybe a bit more user-friendly than Windows, I'll give it that edge, but that's it.

[arjan_intel]

So i did some testing oh how clearlinux got that crazy zstd compression result. Basically, they sort of cheat, they have a patch which always enables multithreading, see https://github.com/clearlinux-pkgs/zstd/blob/master/multi-thread-default.patch.
https://openbenchmarking.org/result/...FO-ZSTDNOOPT82

I'd not call it a cheat; at least it wasnt the intent. Defaulting to single threaded in this era of multi-core is not a good default realistically, and for Clear Linux I changed this default (I realize that this is an opinion, but that's ok)