The data shows that you compared an old beta (debug version) of macOS 10.3.1 with a final version (non-debug version) of OS X 10.11.6.
Do you really think the numbers will be different once 10.13.1 is released next week?The data shows that you compared an old beta (debug version) of macOS 10.3.1 with a final version (non-debug version) of OS X 10.11.6.
This sounds a bit vague and wishy washy on your part. Do you have any data to back up your claim?The test software does probably not support the newer, more efficient technologies and system frameworks in macOS 10.3 and newer.
No, they are totally meaningful. I mean, we're comparing a UI of a 2015 OS to a 2017 OS.Xbench has not been a valid benchmark for almost 10 years. Those numbers are meaningless.
You don't really understand the Metal API by the sounds of it. Your feelings are not measurable data. You need to offer data to back up your claims and disprove that presented.No they are not, because Xbench hasn't been updated since 2006. It could be just a bug when using an app compiled with such an old SDK on 10.13, or some weird things Xbench used to measure the performance that isn't working anymore.
But Xbench source code is not available, and reverse engineered takes too much time.
If the UI was really 4x times slower, it would be easy to see. Instead on my mac the UI feels faster, and it's not even using Metal.
No they aren’t. Xbench really hasn’t been used in 10 yearsNo, they are totally meaningful. I mean, we're comparing a UI of a 2015 OS to a 2017 OS.
Just saying so doesn’t add any validity to it.No they aren’t. Xbench really hasn’t been used in 10 years
not upgrading either .. since mine platter base.Try running GeekBench. Metal and OpenCL were approx. 10% and 15% faster in Sierra than in El Capitan on my Mac Pro 5,1 6 core with a Geforce 980. But I am afraid of installing High Sierra at this moment. Too many horror stories.![]()
Extraordinary claims require extraordinary evidence.