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

[makitango]

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

It is forbidden by default but it has the added info that you can post it when you get approval. This was very likely approved as it‘s still online and not taken down.
One can still expect all the good things to surface (hehe) and the bad ones to stay hidden.

 

[Alan Wynn]

Microsoft SQ1 is 7W TDP. The A12Z in the Mac Mini desktop is probably 15W TDP or higher.

Here are a set of questions I have been asking of people in another thread, so that we can have a clear set of metrics to judge people’s claims and the success of the product. I am gather the answers and will post them in a new thread.

The goal is to get everyone on the record so there will be no moving of the goal posts.

1. What set of benchmarks will you consider as the basis for comparison between the released Apple Silicon Mac systems and competitive Intel/AMD machines?
2. When doing our comparisons between Apple Silicon-based hardware and AMD/Intel based hardware, how will you pick the AMD/Intel chip to compare? What objective metric would you use to define equivalent systems for comparison? Machines at the same price point? Machines with the same max TDP? Something else? The point of this question is that since Apple will not be selling its SoCs to others, one cannot do it purely on price of the chip, one needs some other objective metric to decide what two items should be compared.
3. What objective criteria would Apple Silicon have to meet to be a successful product vs. Intel/AMD’s chips? (10% faster? 25%? 10% better battery life? 25%? Something else?) Once Apple starts to deliver high-end GPUs, what are your answers on those same metrics for those?
4. When did you purchase your most recent Mac from Apple or a third party reseller that was currently shipping at the time you purchased it?
5. What would be required for you to purchase an Apple Silicon-based system?

 

[Janichsan]

Before you all get to excited, a short reminder that these are the numbers to beat:

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

Compared to the current Intel based line-up the Arm equipped Macs are supposed to replace? Yes.

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

This is what an absolute beast looks like:

 

[Alan Wynn]

You realize AMD Zen 4 is on 5nm TSMC process, correct? Nothing Apple is doing will be new as they both are doing it.

Here are a set of questions I have been asking of people in another thread, so that we can have a clear set of metrics to judge people’s claims and the success of the product. I am gather the answers and will post them in a new thread.

The goal is to get everyone on the record so there will be no moving of the goal posts.

1. What set of benchmarks will you consider as the basis for comparison between the released Apple Silicon Mac systems and competitive Intel/AMD machines?
2. When doing our comparisons between Apple Silicon-based hardware and AMD/Intel based hardware, how will you pick the AMD/Intel chip to compare? What objective metric would you use to define equivalent systems for comparison? Machines at the same price point? Machines with the same max TDP? Something else? The point of this question is that since Apple will not be selling its SoCs to others, one cannot do it purely on price of the chip, one needs some other objective metric to decide what two items should be compared.
3. What objective criteria would Apple Silicon have to meet to be a successful product vs. Intel/AMD’s chips? (10% faster? 25%? 10% better battery life? 25%? Something else?) Once Apple starts to deliver high-end GPUs, what are your answers on those same metrics for those?
4. When did you purchase your most recent Mac from Apple or a third party reseller that was currently shipping at the time you purchased it?
5. What would be required for you to purchase an Apple Silicon-based system?

 

[ErikGrim]

1: Performance (GFLOPS) per watt.
2: Irrelevant as long as you follow 1.
3: Irrelevant as long as it’s not slower on native execution
4: 2014, 3-4 weeks for a custom build (relevant how?)
5: Anything that’s not slower than my current system on emulation

 

[MikeSmoke]

He referred for microsoft to build its own arm chip. The microsoft windows arm OS will be offer for sale starting in "fall"
You can ask the user in private

What I was brainstorming, and I’m not a programmer, is that if the Windows Arm on Qualcomm chips is not all that good, it would be interesting if Microsoft in collaboration with Apple, refined a windows version specifically for Apple Silicon. The power of the Apple Silicon, and the fact that it is much more than simply ‘Arm’ on other vendor’s chips, could make a difference. If possible, it would be a win/win for Microsoft and Apple.

 

[MandiMac]

What I was brainstorming, and I’m not a programmer, is that if the Windows Arm on Qualcomm chips is not all that good, it would be interesting if Microsoft in collaboration with Apple, refined a windows version specifically for Apple Silicon. The power of the Apple Silicon, and the fact that it is much more than simply ‘Arm’ on other vendor’s chips, could make a difference. If possible, it would be a win/win for Microsoft and Apple.

So, you're talking about Microsoft-certified drivers for Apple hardware, right?

 

[johngwheeler]

"Exist" as in 128 core chips were announced one week ago and will begin to be sampled to customers in Q4 2020 I believe? (To complete with AMD's 64 Core 128 thread x86 chips which do exist in servers.)

As nice as efficient portables are, it's nice to be able to run good, existing software too. Were it not for the real benefit of running iPad apps on your laptop or PC i wonder if the benefits of conversion would still outweight the costs.

Apple sell so many portables, and portables continune to be more and more in demand so perhaps even in spite of the iOS apps benefit, this would be worthwhile?

B

OK, I'll grant that 128-core ARM chips have only become commercially available very recently (Ampere), but the ARM capability for them has been around for over a year with the ARM Neoverse N1, and there have been implementations of this in the form of the AWS Graviton2 (https://www.nextplatform.com/2019/12/03/finally-aws-gives-servers-a-real-shot-in-the-arm/) - albeit with only 64 cores enabled so far. I think both Ampere and Marvell have 80-core ARM chips in production already.

Here are some benchmarks comparing the AMD Epyc 7742 with 64C/128T against the Graviton2 64c with the Epyc using both SMT and single-thread-per-core. https://www.phoronix.com/scan.php?page=article&item=epyc-vs-graviton2&num=1

Yes, the AMD Epyc does come out on top overall, but by no means in every test, and the ARM-based processor is cheaper to run (not sure about performance/Watt ratio, but I expect it is quite good).

In any case, my answer to jlocker who stated that no ARM chips can match Intel Xeons is valid. In many cases, the 64-80 core ARMs perform much better than 28-core Xeons, and cost less to run.

And to the underlying question of whether Apple will be able to create an ARM chip to replace the Xeon in the Mac Pro, the answer must be: almost certainly, given some time. Let's see if they do it within the 2 years mentioned.

 

[nickgovier]

.

 

[theluggage]

Here are a set of questions I have been asking of people in another thread, so that we can have a clear set of metrics to judge people’s claims and the success of the product.

Good luck with that - the RISC vs. CISC holy wars have been going since before the terms "RISC" and "CISC" were in use (...the 6502 vs. Z80/8080 arguments were much the same) and everybody conveniently forgets that Intel only "won" by - from Pentium Pro onwards - bolting a hardware x86-to-RISC translator onto a RISC core.

As for the rational test - all that really matters is that the ARM-based "MacBook Pros" outperform the Intel "MacBook Pros" when running native Photoshop/Lightroom/FCPX/Logic etc. and that the ARM-based "MacBook Air" runs non-pro tasks smoothly and slickly, and has a good college try at heavier work, while offering better battery life (does anybody really think it needs to be any smaller?) Oh, yes, and all this without slapping another 20% on the retail prices or bait-and-switching the customer with pathetic base specs and usurious upgrade prices. There's no point in switching until your primary workflow can be done natively - Rosetta only has to be good enough to facilitate the transition by running your "b-list" apps.

The current synthetic benchmarks & iPad speed tests don't prove anything about hardware that hasn't been released yet - but they do suggest very strongly that the above is technically feasible. The reality is that both the evangelists and the haters will have to wait and see what the real thing is like. My money is that Apple's CPU/GPU people will deliver, but the rest hangs on Apple building machines that do them justice while successfully persuading developers to "go native".

Meanwhile, like the kids I used to beat in school cross country races by simple virtue of not stopping for a cigarette..., "golf clap" for beating the Surface pro X. It's not really news that, while the whole Surface range rivals Apple in "look and feel", it also rivals Apple in charging exorbitant prices for mediocre specs.

 

[theluggage]

As an Windows user, I'm scared.

Why scared? No puppies will be harmed.

As a Windows user (to the extent of self-identifying as "a Windows user") you probably need a Windows PC. It's not as if the currently available Macs offer particularly compelling value for money once you take away the Unique Selling Point of MacOS.

As a Mac user who sometimes needs to use Windows (Running Windows on Mac was a compelling feature - in 2006 - I'm finding it less and less important) I'm confident that a solution will present itself in the 2-3 years before Intel Macs become hard to get. One of those solutions is buying a PC - something I'm considering anyway. given that it's 2020 and a thin, light and relatively cheap PC laptop would be more than good-enough for my must-have-Windows needs - and when it comes to Boot Camp whipping an ultrabook out of the drawer and waking it up is way more convenient than closing down everything I'm doing on Mac and re-booting (...and then realising that the files I need are still on the APFS partition...) .

Put simply, I'll use the tool that's best for the job - as long as the "critical mass" of my work runs on a Mac, I'll use a Mac. If that "critical mass" shifts to PC then so will I. If there's a compelling reason to own both, I'll own both. If "work" is Windows-centric then maybe I'll have a PC for work and a Mac for home use (there's a shedload of other reasons why that's a good idea).

Meanwhile, my current Mac may well be my last - not because ARM, but because none of the assortment of under-specified, unexpandable, thermally-throttled, ultra-portable. all-in-one and small-form-factor systems in Apple's current range is what I want hardware-wise, and Apple wants $6k minimum (and more like twice than for a spec that makes any sense) for a Mac Pro. While I don't think ARM will somehow, magically, persuade Apple to release the xMac, the ARM could mean MacBooks that can actually deliver their power without screaming fans and and throttling, or that the Mac Mini isn't hobbled by a GPU that sucks even by Intel standards. That could help keep me on Mac.

Windows compatibility was something that Apple could offer fairly simply and without compromise in the ~2005-2010 window when Intel Core/Core i was leading the CPU field. What I really don't want is a Mac that has been compromised in order to run x86 Windows. Wherever Mac is better than Windows, the fact that Mac has been able to re-invent itself every decade or so while Windows has a lot to do with it.

 

[cgsnipinva]

"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.

This is true - but we can extrapolate from where they are with their current set of processors. I would imagine the Mac versions of these chips would have more cores and larger cache's and be able to address larger amounts of memory. Given the high resolution and size of Mac screens - the GPU cores will get a boost.

Having said that -- it is still good to see these numbers as indicators of where we could be. It is obvious Apple has done its homework before announcing this shift.

 

[cmaier]

It is forbidden by default but it has the added info that you can post it when you get approval. This was very likely approved as it‘s still online and not taken down.

How could it be taken down? It’s not copyrighted by apple or anything. The person who posted it likely cannot take it down. And the website that shows it is not party to an agreement with apple not to post it.
[automerge]1593527150[/automerge]

and everybody conveniently forgets that Intel only "won" by - from Pentium Pro onwards - bolting a hardware x86-to-RISC translator onto a RISC core.

Not exactly, no.

 

[Alan Wynn]

So am I, because my next laptop has to be a PC now due to a Windows dependency that has been met by Boot Camp on my last two MBPs, and I hate to move away from Apple hardware but Apple has given me no choice.

What applications do you need to be able to run (trying to see from the compiled list which might be candidates for ports - either to Windows on Arm or native)? Why does it have to be BootCamp vs. virtualization?

4. 2017 MacBook Pro.
5. Boot Camp. So I'm out.

How close are you to replacing your laptop? If you are close to replacing, why not get one of the current Intel systems which should get you another 3-5 years and let you evaluate again then?

 

[apparatchik]

I think Windows on Apple Silicon will be made possible sooner or later, wether is native properly licensed Windows on ARM (it seems the next version will soon be able to run x86_64 applications, so, you would be getting the full windows app catalogue), Microsoft has everything to gain about licensing it, or wether is x86_64 emulated Windows via a new version of Parallels/VMFusion/similar.

Now, for those of us that need native x86_64 Windows -I currently use boot camp to play games- and are unwilling to buy a gaming rig, either because you game casually, or because you don't feel like having two systems (In my case Im basically a digital nomad, my life fits on a 25-pounds backpack), here I should add Im also dependent on MacOS as I do app development, the path forward has but one choice really: buy the latest, fastest Intel Mac, and, eventually, transition to pc cloud/game streaming. Ive been using Geforce Now to play AAA titles on MacOS, and cloud streaming service (remote desktops and gaming rigs) will only improve. I sure am excited about this transition, and will hold to a fast Intel Mac to hopefully arrive at a mature Apple Silicon world in a 3-4 years.

 

[VictorTango777]

Here are a set of questions I have been asking of people in another thread, so that we can have a clear set of metrics to judge people’s claims and the success of the product. I am gather the answers and will post them in a new thread.

The goal is to get everyone on the record so there will be no moving of the goal posts.

1. What set of benchmarks will you consider as the basis for comparison between the released Apple Silicon Mac systems and competitive Intel/AMD machines?
2. When doing our comparisons between Apple Silicon-based hardware and AMD/Intel based hardware, how will you pick the AMD/Intel chip to compare? What objective metric would you use to define equivalent systems for comparison? Machines at the same price point? Machines with the same max TDP? Something else? The point of this question is that since Apple will not be selling its SoCs to others, one cannot do it purely on price of the chip, one needs some other objective metric to decide what two items should be compared.
3. What objective criteria would Apple Silicon have to meet to be a successful product vs. Intel/AMD’s chips? (10% faster? 25%? 10% better battery life? 25%? Something else?) Once Apple starts to deliver high-end GPUs, what are your answers on those same metrics for those?
4. When did you purchase your most recent Mac from Apple or a third party reseller that was currently shipping at the time you purchased it?
5. What would be required for you to purchase an Apple Silicon-based system?

Too bad you weren't collecting statements when Mac users were saying these things:
System 7-MacOS 9 days: Nobody needs automatic memory management. Just click Get Info and change the number yourself.
Apple will not make a video iPod. Nobody wants video on iPod.
Apple should not allow third parties to create native iPhone applications. Web apps are sweet.
iPhone will not have copy/paste. Nobody needs copy/paste on iPhone.

 

[cmaier]

Too bad you weren't collecting statements when Mac users were saying these things:
System 7-MacOS 9 days: Nobody needs automatic memory management. Just click Get Info and change the number yourself.
Apple will not make a video iPod. Nobody wants video on iPod.
Apple should not allow third parties to create native iPhone applications. Web apps are sweet.
iPhone will not have copy/paste. Nobody needs copy/paste on iPhone.

"what good is an iphone without a physical keyboard?"
"iphone should have an iPod click wheel!"

 

[Alan Wynn]

Too bad you weren't collecting statements when Mac users were saying these things:

Those were people's opinions (many silly, but still opinions). What I am collating are objectively measurable metrics.

Apple will not make a video iPod. Nobody wants video on iPod.

These are like the people who would quote the headline of what Steve Jobs said and miss the rest of it. Typically, when he "XXX sucks and here is why" the headline would be XXX Sucks. Amusingly, the important part of his statements were what followed as they were the roadmap for Apple's product development. Apple would work to fix all those problems and then release a product with those problems address.

On the other had sometimes they were just wrong.

By the way, are you going to go on record with your answers to those questions?

 

[NickName99]

What I was brainstorming, and I’m not a programmer, is that if the Windows Arm on Qualcomm chips is not all that good, it would be interesting if Microsoft in collaboration with Apple, refined a windows version specifically for Apple Silicon. The power of the Apple Silicon, and the fact that it is much more than simply ‘Arm’ on other vendor’s chips, could make a difference. If possible, it would be a win/win for Microsoft and Apple.

This is a cool idea technically, but it seems like it could be a conflict of business interests to both Apple and Microsoft. Apple would likely prefer to have superior performing Macs over what is available on Windows, particularly on the laptop side. Microsoft probably wouldn’t want to see the top performing Windows laptops all be Apple products. Asus/Dell/Intel/Qualcomm probably wouldn’t be too happy either.

It will be fun to see where this goes over the next few years. If Apple powered thin/light laptops have double the performance and double the battery life of Intel ones, things will get interesting. It makes me curious about pricing - if Apple prices these even remotely competitively, they’ll have a hard time keeping up with demand. If they price them too high, it could undermine the success of the entire transition.

 

[Arm64MyMac]

Look at my username. I've had this username since 2013. I deserve a cookie.

 

[Alan Wynn]

This is a cool idea technically, but it seems like it could be a conflict of business interests to both Apple and Microsoft.

Doubt it will happen as a boot OS, but I am not sure it is against either parties interests.

Apple would likely prefer to have superior performing Macs over what is available on Windows, particularly on the laptop side.

Apple would have that, as their hardware and most much Mac software would take advantage of Apple Silicon’s other aspects (Neural Engine, hardware compression/decompression, Secure Enclave, etc.), and have a scheduler that takes advantage of the Asymmetric Cores.

Microsoft probably wouldn’t want to see the top performing Windows laptops all be Apple products. Asus/Dell/Intel/Qualcomm probably wouldn’t be too happy either.

Microsoft is still mostly a software and services company. They never have cared about their hardware partners (see the many initiatives they created then dropped: Plays for Sure, Pen Windows, Windows for Pens - cannot remember how many names it had, Windows on MIPS, Windows on Alpha, etc.) and already make hardware that competes with them. As to whether they really want the Surface to be a major profit center for them or are just building them to follow the Alan Kay model (“People who care about software should make their own hardware.”), using them as reference designs for others.

It will be fun to see where this goes over the next few years. If Apple powered thin/light laptops have double the performance and double the battery life of Intel ones, things will get interesting. It makes me curious about pricing - if Apple prices these even remotely competitively, they’ll have a hard time keeping up with demand. If they price them too high, it could undermine the success of the entire transition.

Apple has had a remarkably consistent profit margin over the years. Their problem has been they sometimes build products that are over-engineered. As an example, while it is amazing as a feat of engineering and priced with a reasonable margin, it can be argued that building a $999 stand did not make sense. I will be curious what they do on this front.

There are weird marketing issues as well. For example, if they released the Macbook Air at the same performance but $200 less, will people take that as these machines are somehow lesser than the current machine? One approach might be to release them at the same or slightly lower prices, and then drop them with every release. I would love to see some lower end machines targeted at students (like the white MacBooks were originally), a mid-tier Mac Mini Pro (fast processor, lots of RAM, good graphics and IO), and a much faster Mac Pro (maybe a small price drop on the entry level).

Finally, I would love to see an Extreme line. The fastest All-in-One iMac Pro Extreme, a very beefy Mac Mini Pro Extreme and a top end Mac Pro Extreme that is faster than anything that anyone else is shipping. Make it clear that if one wants the fastest and best machines, one has to buy Macs.

Do I think any of those things will happen? Maybe the Mac Mini Pro (although it might just be a Mac Mini with Apple GPU gets one there with nothing special), and a price reduced entry level (like the iPhone SE). As much as I love the other ideas, I am not sure Apple would.

I can dream.

 

[JosephAW]

If apple installed a keylogger to catch violators I wonder how developers will respond?

 

[cmaier]

If apple installed a keylogger to catch violators I wonder how developers will respond?

presumably with invasion of privacy lawsuits?

 

[jerwin]

Huh? Isn't Zen 3 even out until the end of 2020?

Zen 4 benchmarks are floating about.

https://www.extremetech.com/computing/312234-apple-a12z-arm-performance-vs-x86

We’re in a weird situation at the moment. Intel has always been Apple’s chief supplier, but AMD is selling more performant mobile CPUs today, making them the more obvious point of comparison. The 4900HS appears to score a 1116 single-core and a 7013 multi-threaded score. x86 MT is, at least, in no immediate danger, in absolute terms. Keep in mind that the 4900HS also draws far more power than either the Intel or Apple chips.

 

[burgerrecords]

Zen 4 benchmarks are floating about.

https://www.extremetech.com/computing/312234-apple-a12z-arm-performance-vs-x86

4900HS is Zen 2. Confusing, yes but 4XXX series APUs are Zen 2 while 3XXX CPUs are also Zen 2. Zen 2 is pretty great though.