Originally posted by rabcor
View Post
Announcement
Collapse
No announcement yet.
macOS 10.12 Sierra vs. Ubuntu 16.04 Linux Benchmarking
Collapse
X
-
I really wish he would stop including compile time benchmarks in this way; they are entirely meaningless if it's not the same version of the same compiler with the same flags. Comparing standard Clang 3.8 on Linux to "Apple LLVM version 8.0.0 (clang-800.0.38)"(the macOS Sierra system compiler) tells you more about the build configuration for these compilers than the speed of the system. A better comparison would be between the gcc-6 package on homebrew and GCC 6.0 on Linux.
Side note: I wonder a lot why Apple doesn't use Mesa as their Intel GL stack on OS X. Seems like they would get much better performance for less effort than it takes to keep their proprietary driver up to date. It also seems like some of their GL extensions are already implemented in Mesa.Last edited by microcode; 28 September 2016, 01:45 PM.
- Likes 3
Comment
-
I will admit, there were significant improvements with their metal API between 10.11 and 10.12, but that really cant be honestly directly compared. Though i guess if you can find a game that implements metal, vulkan, and DX12 you could have a relative comparison... but that still doesn't mean each API was optimized for, just that it was included...
Comment
-
The filesystem / sql transaction benchmarks are suspicious. I suspect that Apple's system doesn't sync in the same way.
Also would be nice to compare Clear Linux or some other optimized Linux against macOS. The people who are interested in performance don't use Ubuntu.
- Likes 1
Comment
-
Originally posted by caligula View PostThe filesystem / sql transaction benchmarks are suspicious. I suspect that Apple's system doesn't sync in the same way.
Also would be nice to compare Clear Linux or some other optimized Linux against macOS. The people who are interested in performance don't use Ubuntu.Michael Larabel
https://www.michaellarabel.com/
- Likes 1
Comment
-
I just installed arch on my base model mini. I had to do some manual copying to get grub to boot with efi. Performance with geekbench is slightly higher than osx (and osx to windows 10). The only thing I noticed is reading from the disk is way better on Linux, applications start instantly and have a huge delay on os x. I think once applefs and metal roll out things will be better. I am debating if I should build a powerful zareason arch/windows box or suffer a little for a nice imac.
if anyone wants some osx vs linux/windows benchmarks I can run some.
Comment
-
Originally posted by microcode View PostI really wish he would stop including compile time benchmarks in this way; they are entirely meaningless if it's not the same version of the same compiler with the same flags. Comparing standard Clang 3.8 on Linux to "Apple LLVM version 8.0.0 (clang-800.0.38)"(the macOS Sierra system compiler) tells you more about the build configuration for these compilers than the speed of the system. A better comparison would be between the gcc-6 package on homebrew and GCC 6.0 on Linux.
Side note: I wonder a lot why Apple doesn't use Mesa as their Intel GL stack on OS X. Seems like they would get much better performance for less effort than it takes to keep their proprietary driver up to date. It also seems like some of their GL extensions are already implemented in Mesa.
Comment
Comment