OS X Mavericks vs OS X Yosemite speed testing: Can upgrading to Yosemite slow down your Mac
(For a broader look at the differences between the two Mac OS X operating systems' feature sets, interfaces and so on, read our OS X Mavericks vs OS X Yosemite comparison review.)
Every major update - and sometimes even minor ones - to something as fundamental as a computer's operating system is likely to affect its performance. In either direction.
And by performance we're talking about the computer's ability to do the job of running applications and keeping your system ticking along, as well as the way it can streamline your navigation and movement from process to process. In fact, there are two main aspects under review here: the optimisation of code that lets applications, graphics and data IO work as fast as required in the background. And the elements of the user interface that either facilitate or hinder you getting on with what you need to do.
Through the 2000s we saw Mac OS X get steadily leaner and faster (and Windows continue its established trend of getting slower and more bloated). Until OS X 10.6 Snow Leopard at least, after which more additions in Lion and Mountain Lion were superfluous for many users and could have slowed down Macs on their age limit. The move to a slicker OS X 10.9 Mavericks helped for some, but now we have a whole new decimal update in OS X 10.10 Yosemite.
Yosemite has wrought fundamental changes to the OS subsystem as well as the user interface layer that we see. The divisive new look provides translucency to window frames; a flat, squarer look; and a substitution from Lucinda Grande to a Helvetica font throughout. Some people like the move towards an interface that more closely resembles iOS 7; others are troubled by the cosmetic changes alone which have been describes as a glary whiteness, more garish colours and blurry and hard-to-read typography.
22 Yosemite icons compared with their Mavericks counterparts
It's not scientific but as a quick straw poll it's interesting to note that a forum thread on MacRumors entitled 'Yosemite is Beautiful' runs to 329 replies. Next to it is the 'Yosemite looks terrible!' thread with 2,662 posts.
Mavericks vs Yosemite speed testing: How we tested
We tested performance of both operating systems using a 13-inch MacBook Pro, running the same raft of benchmark tests on the most up-to-date versions of Mavericks (10.9.5) and Yosemite (10.10.2) using a partitioned flash drive on the same system.
Many of the differences were very small. To ensure we had captured real trends rather than random drifts in individual run output, the tests were run multiple times and a mean average calculated.
Mavericks vs Yosemite speed testing: Processor and memory
In the Geekbench 3 test the Mavericks system scored 3127 points for single-core mode, and 6650 points in multi-core mode. Yosemite returned results slightly slower for single-core and slightly faster for multi-core operation, of 3099 and 6716 points. That translates as tiny differences of 0.90 percent lower and 0.99 percent higher for single- and multi-core modes respectively. We'll call that one a draw.
Cinebench 11.5 had similar differences, moving from 1.30 to 1.29, or a 0.77 percent drop, on single-core; and from 3.11 to 3.08 points on multi-core, or a 0.96 percent lower result.
Cinebench 15 meanwhile showed a 1.75 percent drop in single core mode (114 down to 112 points); and 2.49 percent drop in multi-core mode (281 down to 274 points).
So these two benchmark applications show Yosemite as consistently slower, but by less than 1 percent difference in v11.5 and circa 2 percent with the more recent v15.
In the OpenGL graphics rendering section of the Cinebench tests, version 11.5 showed a very small change in framerate (26.30 down to 26.07, or 0.87 percent fewer frames per second). On the other hand Cinebench 15's graphics test indicated a move from 21.77 to 24.45 fps, or a 12.3 percent increase for Yosemite.
SunSpider is a benchmark for quantifying a web browser's JavaScript performance, so here we're measuring the difference between different versions of Safari (7.1.3 and 8.0.3) as much as the underlying operating system. However, since it's the overall user experience we're trying to quantify, the test is a useful one. Here we saw a 3.17 percent slower overall response time with Yosemite, a change from 148.1 ms in Mavericks to 152.8 ms.
Peacekeeper is another JavaScript speed test, running a battery of sub-tests of rendering, HTML5 tasks, 2D game graphics, DOM operations and text parsing. The Mavericks system returned an overall score of 5172 points while Yosemite scored 4913 points. This places the Yosemite Mac as 5.0 percent slower than Mavericks.
In a nutshell: With the exception of one anomalous result (Cinebench 15's graphics test), Yosemite consistently produces slower performance from our test Mac. But the difference was small: anywhere from slightly under 1 percent to 5 percent.
Mavericks vs Yosemite speed testing: Real-world games testing
We tried some real-world testing with two Mac games - Batman: Arkham City and Tomb Raider 2013 - and also the synthetic graphics rendering engine from Unigine Heaven. For all three tests we ran graphics at the MacBook's default native HiDPI setting that renders the 2560 x 1600-pixel screen like 1280 x 800. The differences were now more substantial, and typically in Yosemite's favour.
The Batman game moved from 30.7 fps at Medium detail to 38.0 fps, or a 24 percent increase in framerate. At High detail a similar increase shifted average framerate from 30.3 to 36.0 fps, or a 19 percent increase.
For Tomb Raider we first toggled on the Legacy OpenGL option in this game to obtain playable framerates, since the game's default uses the latest OpenGL API which drastically reduces framerates on slower graphics processors such as the Intel Iris Graphics 5100.
At Normal detail the results were effectively the same between OS versions (22.1 to 22.0, or 0.45 percent drop) while a shift to High detail restored the earlier trend with a 4 percent framerate increase in Yosemite, even if that only resulted from a smaller than 1 fps difference, from 21.0 to 21.9 fps.
Unigine Heaven also benefited very slightly in our Yosemite Mac, moving from 20.5 to 21.6 fps, or a 5 percent increase.
The GFXBench test we use for iOS and Android devices is now available for OS X, and this suite of tests showed some interesting differences.
In a nutshell: When testing with real-world games, we saw noticeably higher gaming framerates on Yosemite, with the difference varying between 4 and 24 percent.
Mavericks vs Yosemite speed testing: GFXBench graphics tests
The first Manhattan test gave an average framerate of 14.62 fps in Mavericks, and 13.20 fps in Yosemite, or a 9.7 percent slower framerate. The same test ran offscreen had a smaller difference, 28.04 down to 27.73 fps, or 1.1 percent slower in Yosemite.
Bigger, much bigger differences were recorded in the next test using the T-Rex animation. Mavericks averaged 47.22 fps onscreen, while Yosemite gave us an average of just 30.42 fps. That's around a 35 percent drop in framerate for the newer OS. Offscreen rendering followed suit, from 87.98 fps in Mavericks to 55.5 fps in Yosemite; or a 37 percent drop.
The ALU test measures shader compute performance, and here the last and current OS gave the same effective result, at 59.99 and 60.00 fps onscreen; and Yosemite pulling ahead with off-screen rendering (338.5 to 349.7 fps, or 3.3 percent improvement).
The remaining results showed some odd trends. In the Alpha Blending test, on-screen renders were within 1.5 percent, 4379 MB/s to Mavericks and 4312 MB/s to Yosemite. But using an off-screen 1080p mode the drop was precipitous, from a steady 5617 MB/s in Mavericks to wildly varying numbers in Yosemite, from 3050 to 1399 MB/s, with a mean at 1899 MB/s. That's a 66 percent drop in performance.
The Fill test had a poor showing with Yosemite in both on- and off-screen modes: Maverick's 7002 MB/s down to 4685 MB/s, and 7302 MB/s down to 4171 MB/s - with high standard deviation in those Yosemite results too. That equates to 33 percent and 43 percent drops when moving from 10.9 to 10.10.
In a nutshell: Our test Mac produced noticeably weaker performance across GFXBench's graphical tests when running Yosemite. The difference was sometimes as large as 66 percent.
Mavericks vs Yosemite speed testing: Internal IO performance
We checked internal storage transfer speeds with the two operating systems, using Intech QuickBench to measure read/write speeds with different data sizes.
Tiny differences were seen, which essentially evaporated after enough iterations and averaging. So whether with Mavericks or Yosemite installed, we saw the same over-achieving results from the little PCIe-attached flash drive, nudging 790 MB/s for sequential reads and 740 MB/s sequential writes. For small 4 kB files, random reads were around 17 MB/s and random writes around 64 MB/s.
In a nutshell: No discernible/significant difference.
In summary: Does OS X Yosemite slow down a Mac
For processor- and memory-based benchmark tests, Yosemite was typically around 1 or 2 percent slower than Mavericks on our test MacBook.
In browser benchmarks based on JavaScript speed, Yosemite was around 3-5 percent slower than Mavericks.
For graphics-related activities and tests the situation was more complicated. In the two Mac games benchmarks, one game was around 20 percent faster in Yosemite while the other was essentially the same. But do remember that while double-digit increases sound impressive that may only be a few frames per second.
In other graphics tests such as the synthetic GFXBench suite Yosemite returned much less consistent results but tended to return lower figures, sometimes dramatically so.
And for data input/output as measured via the internal drive, results were realistically the same for the two operating systems.
Outside the lab
For matters surrounding the change in interface, this is harder to measure. More translucency with frosting effect in window frames and panels would suggest more graphical work required to drive the interface, for example, which might fractionally slow down the 'feel' of some older Macs.
One of the harder aspects to quantify is latency between your action and the resulting feedback from the screen.
This may take the form of input lag when buttons are clicked or windows picked up and moved. Subjectively on our MacBook, there were no perceivable differences here.
There may be other overall slowdowns due to the reworkings in the human-user interface, however. This user, for instance, has found that the reduction in window drop shadow makes it harder to easily discern the layering of windows, to quickly recognise the edges of open stacked windows. This is an issue common to the even drier and more two-dimensional Windows 8 interface, if more marked and annoying in the latter.
We also found the blurred font issue to be a productivity slowdown, since we sometimes need more time to peer at small, less legible writing, followed by more regular screen breaks to reduce eye fatigue. The blurry writing issue is less apparent on Retina-display Macs, but conversely makes our 2012 15-inch MacBook Pro's screen (1680 x 1050) too tiring to view with OS X 10.10 installed.
Other potential slowdowns can arise in Yosemite due to the increased level of user tracking in place by default, with all simple Spotlight local searches being submitted to Microsoft and Apple, generating more network traffic at the least - without getting into the thorny issue of user-privacy violation inherent when you're using OS X 10.10 Yosemite.
Useful reading
If you are concerned that Yosemite has slowed down your Mac, two articles in particular are likely to interest you: How to speed up a Mac, and How to downgrade from OS Yosemite to Mavericks. You may also like to read our OS X Mavericks vs OS X Yosemite comparison.