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.
El Cap UI is 9x slower than Snow Leopard
- Thread starter dogslobber
- Start date
- Sort by reaction score
"User interface test - elements" show a 9 times difference.
I find it very odd BTW, as my Xbench tests across various OS versions were within 10% tolerance. No Mavericks install to back it up with a screenshot, unfortunately. My Geekbench history shows an actual increase in performance throughout the years (I run ElCap on a Mac Pro 2,1, so it's almost 10 years old now and has been continuously upgraded with new systems since OS X Tiger).
I find it very odd BTW, as my Xbench tests across various OS versions were within 10% tolerance. No Mavericks install to back it up with a screenshot, unfortunately. My Geekbench history shows an actual increase in performance throughout the years (I run ElCap on a Mac Pro 2,1, so it's almost 10 years old now and has been continuously upgraded with new systems since OS X Tiger).
Whoops! my mistake, I just homed in in the total results.
(Wojtek
More Macs than I can count, really. Like 20 or so...)
That's the problem with having a total of 20 fingers and toes.
I find it utterly surprising too but the data says
FWIW, SL seems to fly on my 2010MBA compared to the pedestrian El Cap.
Now I'm really pi$$ed off, I've never payed much attention to these bench results before and to be truthful my own user experience does not reflect what the data shows.
But the data difference is really bad between my existing Sierra and an older Mavericks test.
And this is Apple's claim to better and faster?
But the data difference is really bad between my existing Sierra and an older Mavericks test.
And this is Apple's claim to better and faster?
My observations for SL to El Cap were 9x slower and El Cap to High Sierra being 4x slower. I don’t want to multiply those numbers together. A lot of people wanted to debunk the 4x on th HS forum but they didn’t offer any justification or data of their own.
I believe the actual problem is using a wildly outdated, 10-years old benchmarking application that makes use of system functions and calls deprecated since then. They need to be emulated for compatibility sake by the newer versions of the OS, hence they are slower.
Just to make a comparison - here's my ElCap Xbench result, I highly doubt it conveys ANY meaningful data at all.
The user interface in HighSierra is rendered using Metal, Xbench doesn't use Metal at all so it reverts to an emulation. This is why we want to keep our apps updated, so they can make use of new subsystems.
Just to make a comparison - here's my ElCap Xbench result, I highly doubt it conveys ANY meaningful data at all.
The user interface in HighSierra is rendered using Metal, Xbench doesn't use Metal at all so it reverts to an emulation. This is why we want to keep our apps updated, so they can make use of new subsystems.
Last edited:
My 2010 MBA which via xbench I showed High Sierra to be 4x slower than El Cap to be 9x slower than Snow Leopard doesn't support the Metal API.
I can only dream for UI results such as yours, and except for the ram an identical machine.
System Info
Xbench Version 1.3
System Version 10.11.6 (15G17023)
Physical RAM 8192 MB
Model MacPro2,1
Drive Type OWC Mercury Electra 6G SSD
User Interface Test 44.95
Elements 44.95 206.30 refresh/sec
xbench doesn't support Metal as well and this exactly the reason this benchmark (specifically in the 'user interface' section) and it's results are not relevant in modern OSes. You are not comparing apples with apples here (pun unintended).
My Radeon R9 280X may be the reason.
Do you understand what my response said? It said my test machine doesn't SUPPORT Metal API so that removes this alleged Metal API issue from the question. My non-Metal API supporting Mac shows that Sierra is THIRTY SIX TIMES SLOWER than Snow Leopard executing general purpose UI API calls.
You seem to somehow miss the answer, so I will repeat: yes, since your MBA doesn't support Metal and you run your outdated test on it, it's no wonder the OS falls back to non-accelerated mode.
No, actually it's not because this math doesn't work like this. Compare apples to apples on relevant machines and then jump to conclusions or just record the testing procedure with a camera and count the time it takes for the test to conclude, then you'll know what's faster. Not only you try to measure your computer with a wrong tool, but your tests lack the ceteris paribus principle which is crucial to any meaningful measurement.
I would also appreciate if you behaved a little bit more civilized, your aggressive attitude and constant lack of realization that you are using a wrong tool - which both have been highlighted in multiple threads you have posted about the same alleged "issue" - do not encourage any discussion nor help.
Last edited:
The results are in for Mavericks. Only a 25% slowdown over SL for the UI test. It looks like Mavericks is the end of the road for this MBA as the follow on OS versions are terrible, as the data has proven.
Do me a favour, fire up the QuickTime Player to record the screen and do exactly the same routine on two systems: from a standstill, double-click Macintosh HD and wait until the Finder window opens. Post both videos here, please, so we can count the frames between the click and rendering the full window.
Interestingly...after years of avoiding it...I finally upgraded a 2008 Core2Duo MacBookPro from 10.6.8 to El Cap.
Amazingly (for me), the UI is actually not noticeably different - perhaps just a tiny bit laggy doing Mission control.
Overall, I am surprised (and pleased!). Things like Safari actually seem a lot faster (probably due to all the new code present in modern web pages I guess?).
Amazingly (for me), the UI is actually not noticeably different - perhaps just a tiny bit laggy doing Mission control.
Overall, I am surprised (and pleased!). Things like Safari actually seem a lot faster (probably due to all the new code present in modern web pages I guess?).