Apple's A12Z Under Rosetta Outperforms Microsoft's Native Arm-Based Surface Pro X

[MacRumors]

Apple's Developer Transition Kit equipped with an A12Z iPad Pro chip began arriving in the hands of developers this morning to help them get their apps ready for Macs running Apple Silicon, and though forbidden, the first thing some developers did was benchmark the machine.

Multiple Geekbench results have indicated that the Developer Transition Kit, which is a Mac mini with an iPad Pro chip, features average single-core and multi-core scores of 811 and 2,871, respectively.

As developer Steve Troughton-Smith points out, the two-year-old A12Z in the Mac mini outperforms Microsoft's Arm-based Surface Pro X in Geekbench performance, running x86_64 code in emulation faster than the Surface Pro X can run an Arm version natively.

https://twitter.com/i/web/status/1277628592458006528

Averaging seven Geekbench 5 benchmarking results, Microsoft's Surface Pro X features a single-core score of 726 and a multi-core score of 2,831, meaning the A12Z outperforms the Surface Pro X in single-core testing and is on par or slightly better in multi-core performance.

https://twitter.com/i/web/status/1277616766861152256

The Surface Pro X features a Microsoft-designed 3GHz Arm processor based on the Qualcomm SQ1 chip.

Apple's DTK provided to developers is just a test machine using an older A12Z chip (it's the same as the A12X chip in the 2018 iPad Pro but with an extra GPU core unlocked). Apple's Arm-based Macs that run Apple Silicon will have new chips designed for the Mac and based on the A14 chip created for the 2020 iPhone lineup with a 5-nanometer process.

Apple says its Apple Silicon Macs will bring major improvements in performance and power efficiency, and the first Arm-based Mac is set to be released before the end of 2020.

Article Link: Apple's A12Z Under Rosetta Outperforms Microsoft's Native Arm-Based Surface Pro X

 

[Rudy69]

But people in the last thread told me the score was bad? /s

Let's not forget this is and older chip too. This transition is super exciting! I really can't wait to see what they do in the bigger form factors like the MacPro

 

[davidjearly]

"Apple's Arm-based Macs that run Apple Silicon will have new chips designed for the Mac and based on the A14 chip created for the 2020 iPhone lineup with a 5-nanometer process."

Er, we don't know what Apple have planned for the Mac lineup in terms of Apple Silicon and whether it will share a chip design with their mobile devices.

 

[lyngo]

I would say that some of these findings are pretty good indicators about where we are heading. I, for one, am looking forward to this and seeing where it all goes!!

 

[UltimateSyn]

So just imagine what the actual Mac processors are going to do...

 

[now i see it]

though forbidden, the first thing some developers did was benchmark the machine

Classic

 

[apparatchik]

So, Geekbench has to come out with a universal binary for MacOS in order to run native on Apple silicon? not that the native score would be far from what we already know about the A12Z in the iPad, right?

 

[m4mario]

I cant wait for Mac to be truly ahead in performance as they used to be before the transition to Intel. Intel chips kind of flattened the market.

 

[gnomeisland]

The down clock is odd. I wonder if that's an artifact of running under rosetta or if the chip is really under clocked. Yes, I know it isn't indicative of the performance from consumer machines, but I am looking forward to the inevitable ARM-based benchmarks. Imagine Geekbench has a vested interest in riding this publicity wave.

 

[mnsportsgeek]

Gonna be pretty cool when Apple’s 5G modems outperform Qualcomm too.

 

[iReality85]

Apple: "Hey, we really don't want you to benchmark these machines." ?

Devs: “Sure, no problem!” ?

Everyone else: ?

 

[NickName99]

This is absolutely hilarious. It’s running a 2 year old iPad processor, underclocked, under emulation, with only half its cores running, and it beats the Surface running native code.

The A14 gen Apple processor enabled Macs are going to be absolute beasts.

 

[Falhófnir]

Interesting as the SQ1 was meant to be getting close in raw performance to the A12X! Think the real Mac silicon is going to be epoch making stuff! Still so many people here seem determined to seize on anything less than stellar as an insurmountable problem - everything I'm seeing suggests this is going to be more than worth it on the other side!

 

[gkarris]

Quick - copy the numbers down before Apple makes them pull them down!!!

ARM Macs are gonna rock!!

Now, please lower prices to REALLY compete with PC's!

 

[Hattig]

Qualcomm SoCs for Windows on ARM are not great though. 8cx is a disappointment. ARM have never cared about the performance sector until recently (A78/X1 are far better).

 

[jlocker]

But people in the last thread told me the score was bad? /s

Let's not forget this is and older chip too. This transition is super exciting! I really can't wait to see what they do in the bigger form factors like the MacPro

To get to a level of a xeon processor it is going to take Apple about 4 years, and Intel is still going to be improving the Xeon workstation and server processors. So if you need a machine now you might as well get it and in 4 years then evaluate xeon vs Arm. I loved the PowerPC and it was a RISC processor but Apple has a long way to go to get to a workstation/server level. Get a machine get Applecare and enjoy a system for the next 4 years. Yes they should be about to match a i5 soon but then the i7 and then the i9.

 

[burgerrecords]

One thing for certain is that Apple will always be faster than the competition in all relevant use cases. But seriously, it is good they are pushing if there is space to push. I hope it lives up to the hype when comparing desktop-to-desktop.

 

[deconstruct60]

...
Multiple Geekbench results have indicated that the Developer Transition Kit, which is a Mac mini with an iPad Pro chip, features average single-core and multi-core scores of 811 and 2,871, respectively.

As developer Steve Troughton-Smith points out, the two-year-old A12Z in the Mac mini outperforms Microsoft's Arm-based Surface Pro X in Geekbench performance, running x86_64 code in emulation faster than the Surface Pro X can run an Arm version natively.

After installing a x86_64 Geekbench and running it at least once. It isn't emulation anymore. Rosetta 2 pragmatically composes a "fat binary" version of the program and simply runs that. There are probably no corner case dynamically loaded libraries that the benchmark doesn't load while composing the stats. So after run once you have another native program sitting on the disk.

P.S. there are some system calls that are being 'trapped and emulated' but the bulk of the code in Geekbench is highly directed at the CPU cores themselves. Not ducking in and out of the kernel. If enough of the benchmark gets cached that then this is just already native code just running.

 

[Dhock_Holiday]

Color me impressed, can't wait to see what Apple releases in the coming months

 

[Rudy69]

This is absolutely hilarious. It’s running a 2 year old iPad processor, underclocked, under emulation, with only half its cores running, and it beats the Surface running native code.

The A14 gen Apple processor enabled Macs are going to be absolute beasts.

The performance of the A chips are always well ahead of the competition in the mobile market. Definitely going to be interesting to see what they come up for their computers.

Unfortunately the few ARM based computers we've seen just cram underpowered chips meant for phones without really doing anything else. This left us with no real idea how good Apple's future computers would be. Don't forget ARM is an architecture, not a phone chip design. The most powerful super computer runs ARM CPUs

 

[Quicksilver867]

The ARM Surface Pro X being a complete dog is not new knowledge: this just quantifies it a little more.

 

[JPack]

The down clock is odd. I wonder if that's an artifact of running under rosetta or if the chip is really under clocked. Yes, I know it isn't indicative of the performance from consumer machines, but I am looking forward to the inevitable ARM-based benchmarks. Imagine Geekbench has a vested interest in riding this publicity wave.

Probably related to power delivery.

The iPhone Xs had power delivery issues which resulted in unstable 3DMark, as reported by Anandtech. The logic board wasn't designed for desktop use.

 

[cmaier]

After installing a x86_64 Geekbench and running it at least once. It isn't emulation anymore. Rosetta 2 pragmatically composes a "fat binary" version of the program and simply runs that. There are probably no corner case dynamically loaded libraries that the benchmark doesn't load while composing the stats. So after run once you have another native program sitting on the disk.

P.S. there are some system calls that are being 'trapped and emulated' but the bulk of the code in Geekbench is highly directed at the CPU cores themselves. Not ducking in and out of the kernel. If enough of the benchmark gets cached that then this is just already native code just running.

It’s not emulation, true, but it;s still highly non-optimal code.

 

[rp2011]

So over the next few months as developers tinker with it some more, we will be getting a stream of tidbits to tide us over .

 

[ouimetnick]

"Apple's Arm-based Macs that run Apple Silicon will have new chips designed for the Mac and based on the A14 chip created for the 2020 iPhone lineup with a 5-nanometer process."

Er, we don't know what Apple have planned for the Mac lineup in terms of Apple Silicon and whether it will share a chip design with their mobile devices.

If you watched the keynote, they mentioned that they will be making a specific line of processors for the Macintosh. It will share the same architecture, hence the DTK, but it may not even be called an A14. Could be called B14 or perhaps a A14M.. They aren’t just taking the chips from the iPad and slapping them into a computer going forward. That’s only for the DTK.