Geekbench running on Rosetta 2 in Apple's DTK outperforms the Surface Pro X with native ARM64 Geekbench

[captain_to_fire]

Summary

 Basically, what would a developer do once they receive Apple's Developer Transition Kit (DTK)? Do they immediately recompile their apps to support Universal Binary 2? Probably not as some of them decided to run benchmarks including Geekbench. It shows that even when emulated, the A12Z inside Apple's DTK has faster single and multi-threaded performance over the SQ1 inside the Surface Pro X which has the native ARM64 version.

 

And yes, an ARM64 Geekbench exists [here]

 

Btw, "Pentium III" under the Surface Pro X is a minor Geekbench bug because Intel never had multi-core processors during the time of Pentium III in 1999 nor will it ever run on 2995 MHz.

 

Quote

Developers taking part in Apple's preview of its Arm-based Apple silicon hardware have run Geekbench benchmarks, with the results suggesting it considerably outperforms Microsoft's Arm-based Surface Pro X. 

The Geekbench results show that Apple's Mac mini-like desktop Developer Transition Kit (DTK) with the A12Z Bionic SoC from the 2020 iPad Pro has a single-core score of over 800 and a multi-core score of over 2,900. 

 

But even running through Rosetta 2, Apple's developer kit scores come in significantly higher than Geekbench results for Microsoft's Arm-based Surface Pro X, which runs on the SQ1, an Arm processor co-developed by Microsoft and Qualcomm. The Surface Pro X scores are around 600 in a single-core benchmark and over 2,600 in a multi-core test. 

 

As well-known developer Steve Troughton-Smith points out, the results don't look good for Microsoft and Qualcomm, given that the A12Z is based on the two-year-old A12X. 

It's no secret that the Apple Silicon inside their top tier iPhones and iPads are the fastest among phones and ARM tablets.

Since Rosetta translate apps compiled for x86 into ARM during installation which seems to be better than Rosetta 1 which translates a PowerPC app into x86 every run.

However, it should be noted that the difference is not that big since Rosetta 2 will always have performance penalties. In fact when it comes to single core performance, Geekbench running on Rosetta 2 is still significantly slower than the iPad Pro because even with the same chip running the same app.

 

It should be kept in mind that Apple is not going to sell a Mac with A12Z as the DTK is only meant for developers to recompile their apps to support Universal Binary 2. After that, they have to return the DTKs to Apple. I for one will be looking forward to how their own silicon will perform against something like a 16" MacBook Pro with an i9 or a mobile Ryzen 4000 in Windows PCs.

With that said, looks like Microsoft isn't going down without a fight, or do they even want to fight? Wccftech spotted a Geekbench result showing an upcoming Surface Pro X with a seemingly upgraded CPU but it looks like only little has changed. No wonder reviews on the Surface Pro X have been quite lukewarm. If Microsoft is really serious with Windows 10 on ARM being competitive against the upcoming slew of Macs with Apple Silicon, they either have to price it low or just get an ARM license and do things the Apple way. You know what they say, "you want to get things done, you gotta do it yourself because a good help is hard to find nowadays."

 

 

Sources

Zdnet, Geekbench: DTK with A12Z, Surface Pro X, iPad Pro, wccftech

Edited July 1, 2020 by captain_to_fire

[captain_to_fire]

2 minutes ago, BlueScope819 said:

If only we were able to run these benchmarks on the iPad, it would very interesting to compare the results

Geekbench and some other benchmarks are available on the App Store 

https://apps.apple.com/us/app/geekbench-5/id1435082259

[DrMacintosh]

ARM is the future. 

[Guest]

The chip inside the developer kit isn't going to be the final product though, it's just the IPad Pro's arm chip inside a Mac mini chassis running Big Sur, the released arm processors will be different from that one. And we already knew that the A12Z outperformed the SQ1 from benchmarks when it was in the IPad pro.

[captain_to_fire]

3 minutes ago, DrMacintosh said:

ARM is the future. 

I hope Intel releases a core i9 ARM chip with hyperthreading 

[DrMacintosh]

35 minutes ago, Darpyface said:

The chip inside the developer kit isn't going to be the final product though, it's just the IPad Pro's arm chip inside a Mac mini chassis running Big Sur, the released arm processors will be different from that one. And we already knew that the A12Z outperformed the SQ1 from benchmarks when it was in the IPad pro.

Someone at Apple allegedly said, "Look what we can accomplish without even trying"

 

So just think about what Apple can do if they increase the power budget. 

[Nowak]

Is anyone really surprised? The Pro X runs on a Qualcomm SoC and Apple's SoCs have historically ran circles around Qualcomm's garbage.

[DrMacintosh]

What's going to be my favorite thing about ARM Macs is the introduction of asymmetrical core design like in iPhone and iPad. 

 

High Performance cores for large tasks and hyper efficient low power cores for basic tasks. Apple also knows how to use all these cores in tandem, so both sets can be active at the same time, rather than one group or the other. 

 

It's going to be real easy and real cheap to load these future Macs up with cores. 

[Zodiark1593]

Emulated x86 performance appears to be quite similar to ~3.0 GHz Haswell territory. Yikes!

[Belgarathian]

3 minutes ago, Zodiark1593 said:

Emulated x86 performance appears to be quite similar to ~3.0 GHz Haswell territory. Yikes!

RIP my 4790k.

[captain_to_fire]

1 hour ago, Nowak said:

Is anyone really surprised? The Pro X runs on a Qualcomm SoC and Apple's SoCs have historically ran circles around Qualcomm's garbage.

That’s why I think Microsoft should just do it themselves by obtaining an ARM license just like Apple, Samsung and Qualcomm. 

[Nowak]

8 minutes ago, captain_to_fire said:

That’s why I think Microsoft should just do it themselves by obtaining an ARM license just like Apple, Samsung and Qualcomm. 

Honestly? Yeah.

 

I wish Qualcomm didn't have such a stranglehold on the mobile ARM SoC market because their SoCs are actual trash.

[hishnash]

1 hour ago, captain_to_fire said:

themselves by obtaining an ARM license just like Apple

They would need to commit to selling ARM windows devices on mass to recover the R&D costs needed to role their own. 

 

1 hour ago, Zodiark1593 said:

Emulated x86 performance appears to be quite similar to ~3.0 GHz Haswell territory. Yikes!

Also worth noting that developer have been told that the meulation on the A12z is not as good as it will be on the real chips since the A12z does not natively support 4kb page size. Apple have not saide what `not as good` means weather that means might not run some stuff or just slower as it needs to intercept the page reads/writes? that is unknown.
 

3 hours ago, captain_to_fire said:

i9 ARM chip with hyperthreading 

There is not much point in ARM having hyperthreading, your better of saving the die area needed for this and adding more `little` cores.

[suicidalfranco]

[captain_to_fire]

Just now, suicidalfranco said:

I’m waiting for Apple’s future ads once the transition is complete 

 

[Kisai]

3 hours ago, DrMacintosh said:

Someone at Apple allegedly said, "Look what we can accomplish without even trying"

 

So just think about what Apple can do if they increase the power budget. 

That's the thing. 

 

I posted in the last thread in this subject what the geekbench benchmarks were for cpu's that people actually have, and "under Rosetta2" scores, actually puts this closer to a Sandy Bridge CPU (MacMini 2012) which means it's not that unreasonable. Particuarly since it's not the current CPU model.  So if there's really a 40% penalty for running under Rosetta2 under single-threading, that's actually not that bad. Presumably the better software vendors will just update their software and it won't be an issue for most productivity software.

 

It how ever will require games and video-players (eg VLC) to be recompiled and likely re-optimized since the ARM parts perform at a higher performance level than the Apple iMac/Macbook 13/Macbook Air's without dGPU's. Like you're not going to play a game that uses Cedega/Wine/Proton to run a Windows version to work here, since the installable windows binaries needed would need to be 64-bit ARM versions (though that's very likely to be possible, it just can't benefit from Rosetta2 since it won't be able to recompile a binary it doesn't understand.)

 

It would be nice however if some games start getting native Mac ports, but somehow I think the opposite will happen and game devs will just start ignoring the Mac platform for anything that's not mobile shovelware.

[williamcll]

Considering Mac ARM chips regularly go above 3Ghz this is not a surprise.

[TechGod]

I am incredibly excited for these ARM Macs. 

[porina]

8 hours ago, DrMacintosh said:

Someone at Apple allegedly said, "Look what we can accomplish without even trying"

 

So just think about what Apple can do if they increase the power budget. 

I think that is a discredit to the people who have worked on it directly or indirectly leading up to this result. It's not like it magically appeared overnight. Apple have worked on this over the years and it is a good result.

 

As for power budget (assuming same core count), power efficiency goes down a lot faster than performance goes up. While some uplift can be expected, wouldn't expect massive differences.

 

7 hours ago, DrMacintosh said:

What's going to be my favorite thing about ARM Macs is the introduction of asymmetrical core design like in iPhone and iPad. 

Gonna be fun to see above future vs Lakefield comparisons. 

 

7 hours ago, DrMacintosh said:

High Performance cores for large tasks and hyper efficient low power cores for basic tasks. Apple also knows how to use all these cores in tandem, so both sets can be active at the same time, rather than one group or the other. 

For sure an advantage Apple has, they have more experience, and control of hardware and software stacks to make it work. 

 

6 hours ago, Zodiark1593 said:

Emulated x86 performance appears to be quite similar to ~3.0 GHz Haswell territory. Yikes!

Without checking that statement is correct, that is comparing to a 7 year old desktop CPU at this point.

 

6 hours ago, captain_to_fire said:

That’s why I think Microsoft should just do it themselves by obtaining an ARM license just like Apple, Samsung and Qualcomm. 

It's not that easy. Doing the work once you have the licence is not trivial. Apple has had years of experience. Thus why MS would outsource or partner that work off. To gain that experience in house will take years.

[leadeater]

10 hours ago, captain_to_fire said:

I hope Intel releases a core i9 ARM chip with hyperthreading 

I don't, the model name will suck so much nobody will even notice it exists.

[captain_to_fire]

45 minutes ago, leadeater said:

I don't, the model name will suck so much nobody will even notice it exists.

core i9-1100H with four big and four LITTLE cores 

[Curufinwe_wins]

40% drop before you get to AVX and similar is a pretty steep price for any one who is time-sensitive on the compute work. But I do think we all know the vast majority of computers don't actually need their power the vast majority of the time. Unless all you do is run code and/or AAA game lol.

[Zodiark1593]

9 hours ago, porina said:

 

 

Without checking that statement is correct, that is comparing to a 7 year old desktop CPU at this point.

 

.

It was only about 5 years old when the A12 chips shipped. 
 

Still kind of a bit of a wonder that my phone wallops my desktop in single threaded (non-AVX) workloads. 

[Jotoco]

Intel got what it deserved for slacking.

 

Now do Qualcomm. 

 

Hope now more R&D goes into other arm platforms so there is actually competition. 

[justpoet]

11 hours ago, Curufinwe_wins said:

40% drop before you get to AVX and similar is a pretty steep price for any one who is time-sensitive on the compute work. But I do think we all know the vast majority of computers don't actually need their power the vast majority of the time. Unless all you do is run code and/or AAA game lol.

Even most AAA games don't push the CPUs much unless you artificially drop resolution and visual quality to try and take the GPU out of the speed equation.