The first benchmarks for Apple’s Developer Transition Kit, a Mac mini powered by Apple Silicon, reveal that the company is very slightly under-clocking the A12Z Bionic SoC, with, as expected, Rosetta 2 also negatively impacting performance.
Malcolm Owen for AppleInsider:
In terms of performance, the single-core tests resulted in a range of between 736 and 844 points and an average of 811. For multi-core benchmarks, the range is from 2,582 to 2,962, with an average score of 2,781.
These results are considerably lower than the 1,118 single-core and 4,625 multi-core benchmarks of the 12.9-inch iPad Pro which also uses the A12Z Bionic, but there are a few reasons that could be the case. For a start, Apple is running the A12Z at a slower 2.4GHz in the DTK, instead of the 2.5GHz clock in the iPad Pro line, though at this time it is unclear why this is the case.
Another interfering factor is the benchmark itself. Rather than using a native ARM-based app that runs on Apple Silicon, it is likely that benchmarks are being carried out using Geekbench 5’s macOS client. As Apple introduced Rosetta 2 as a a translator to convert apps meant for Intel-based Macs to function on Apple Silicon, it is highly likely that this introduces some overhead that affects the benchmark’s score.
MacDailyNews Take: Obviously, benchmarking software running through the Rosetta 2 translation layer is going to be impacted significantly. In fact, when we run our benchmarks, for best results, we like to run them on freshly-booted Macs running the benchmark software residing on a RAM disk.*
Apple’s Developer Transition Kit is an iPad packed into a Mac mini case, running an Apple A12 variant — not even an A13 — and yet Apple knows it’ll be plenty for developers to “make it so.” They’ll get their apps running well on an A12Z, this glorified iPad stamped with Mac branding, and when the first Macs with Apple’s custom silicon ship to the public by the end of the year, they’ll be packing A14-class SOCs.
We’ll have to bolt ’em down, lest they spontaneously take flight!
I can’t wait to see the benchmarks almost as much as I can’t wait to see the faces of the remaining Wintel boat anchor holdouts when they see the benchmarks!
Yes, this is going to be FUN! And fun, dear friends, is exactly what we need after the start of this wonderful year. — SteveJack, MacDailyNews, June 23, 2020
*To create a RAM disk on your Mac, in Terminal:
diskutil erasevolume HFS+ 'RAM Disk' `hdiutil attach -nobrowse -nomount ram://XXXXX`
XXXXX above is the size of the desired RAM disk in terms of memory blocks, as such:
2097152 –> 1 GB
4194304 –> 2 GB
8388608 –> 4 GB
for comparison my 2012 i7 iMac comes in at 812 single 3048 multicore,
the i9 2019 iMac does 1243/8286.
If these benchmark scores are using the x86 emulation, then this is an extraordinary testament of Rosetta2’s efficiency. wow!
So Rosetta 2 is causing a 24% to 37% performance hit (after accounting for the diferent CPU clocks) IF their assumption is correct that the version of Geekbench is actually an Intel variant running under Rosetta 2. I’m hoping (and expecting) there will be some performance tuning between now and the time end user systems ship by the end of the year.
I’ll have to look up the performance numbers for a comperable i5 Intel chip (likely a mobile one) running at a similar clock rate to get the raw numbers for that assumed version of Geekbench. That should turn out to be an interesting comparison — Intel native versus Intel under Rosetta 2.
Yup very reasonable hit to run legacy software. Can’t wait to see how fast the new Ax Macs shipping at the end of the year will be.
One wonders how close to the eventual launch machines these Dev machines are. I suspect there will be considerable differences for the biggest upside impact upon launch.
These machines are using the A12Z. The first retail Apple silicon Macs will probably use A14 variants (i.e., the chip family powering the flagship 2020 iPhones). The speed jump just from that will be as significant as the jump from a 2018 iPhone to 2020. Add in the additional optimizations to both hardware and software developed over the next six months. The difference will be significant, particularly when running native code without emulation.
So, looking at the 4-core i5 Intel chip at a base frequency of 2.0 GHz (slower than the A12Z at 2.4 GHz) running native Geekbench in a MacBook Pro the single core score of 1244 and multi core of 4526. This leads to the realization of running Intel code Geekbench on the A12Z developer box is 46% slower for single core and 49% slower for muti core when running under Rosetta.
If we lay this all at the feet of Rosetta, this means Rosetta is going to need a LOT of tuning between now and the end of the year when the new A-series Macs ship.
If Apple can’t (or won’t) speed up Rosetta significantly, there will be a LOT of people keeping their Intel Macs for years to come.
That depends on the native performance of the consumer Apple silicon Macs. If the native speed is fast enough, it can absorb the hit from running in Rosetta 2 emulation and perform almost as well as the buyer’s three or four year old Intel Mac.
The problem is likely to be transitory in any case. If it as easy to update Intel Mac apps into Universal 2 apps as the Keynote claimed, almost every Mac program that is still being maintained will be updated within a year at the most. At that point, the extra speed that made Rosetta 2 possible will be directly apparent in native application speed.
The only real losers will be those who still require Windows programs that will never be available in Mac versions.
Do you remember when companies like Adobe had both Mac and Windows versions? Mac users had a legitimate gripe at being second tier customers when the Windows versions would run circles around the Mac version. The new fight for the desktop is coming, and it’s not only multi-core, but possibly multi-cpu.
That said, the desktop was oversold back then, and mobile won the netbook wars…
So we have an Apple that will essentially be the new Silicon Graphics (or Sun, or DEC,…) with highly vertically integrated, extremely fast, and extremely proprietary, and the “good enough for most people” ARM Mac.
x86… your move, many potential advantages.