Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Performance / Benchmarks for 10.12 GM?
- Thread starter applCore
- Start date
-
- Tags
- benchmark performance sierra
- Sort by reaction score
Like, memory usage etc?
Wont see any difference in CPU speed. GPU drivers may be better, but I just use NVIDIA own not Apple's.
Wont see any difference in CPU speed. GPU drivers may be better, but I just use NVIDIA own not Apple's.
They have improved a bunch of things (e.g. layout algorithms) etc., so yes, the system might feel faster to you if you were using (usually poorly coded) apps where this was a factor. But its not something that you will see in a standard micro benchmark.
As you seem knowledgeable about these things, is Metal rendering of the UI enabled on dual-GPU machines now?
No idea
I don't have a new Mac Pro running Sierra. My knowledge is limited to what was said in the WWDC and the rare occasion when I read the Apple dev mailing lists. I certainly have no insider information.I am running 2x HD7950 in my Mac Pro. From my observation, the answer is NO. if there is any test I can do for you, please let me know the details on how to carry out the test.
Oh, my bad, I miss interpret that he was asking "will BOTH GPU use for GUI?"
His question make more sense now, I was thinking why Apple need to do that, because the GUI is not that demanding, and don't need that much GPU power.
Yeah, multi-GPU desktop composition would be horribly inefficient, both from the performance and the power standpoint. Unless we are talking about very large desktops
Yes, I've read that somewhere but can't find the source, I'd assumed it was in a post by leman
I imagine it is in a WWDC video somewhere...If it is I have no idea how we test it. IIRC in El Capitan, only some machines accelerated the UI with Metal, even if other machines (like dual_GPU laptops) that supported Metal didn't used Metal for the 10.11 UI. I don't think there are good benchmarks for the UI, and does Quartz Debug still work when using Metal for the UI?
Yes, I've read that somewhere but can't find the source, I'd assumed it was in a post by leman
Ahaha, this is how unhealthy rumours are being spread
I don't think I ever claimed something like this. But I do have some recollection of reading/hearing about this. WWDC seems not unlikely.I am fairly sure that you won't see much in a benchmark. There is no way to control the confounding factors as you can usually do in a microbenchmark and Metal is more about efficiency than anything else (e.g. it probably won't improve the performance of your UI drawing operations, but it will likely to result in less energy impact/resource utilisation overall).
As to testing it, it is possible to see whether Metal APIs are accessed when an app is running. I believe this is how netkas did his original investigation on Metal usage in OS X UI operations.
