Report: Super-Lightweight 12-inch MacBook Powered By Apple Silicon to Launch This Year

[EugW]

A score that pisses all over all the latest Mac Mini's, with the Fastest i7 Mini only beating it in multicore by a marginal amount.

Hell even the old A9 iPad benches almost twice as fast as the newest Celerons in el-cheopo notebooks.

The A12 might not be fast enough for a Mac Pro, but I'd warrant that even that is more than adequate for a Notebook, MacMini or iMac.

Cant wait to see what an A14 system is capable of.

I’d been saying for a long time now that A14’s performance would be fine for a MacBook. Hell, even my wife’s A10 is fine for a basic machine. Not awesome but fine. I actually bought that iPad 7th gen in 2019 knowing it only had the 2016 A10 in it. It scores roughly as fast as my 2017 MacBook and it feels similar. A14 will be about three times as fast.

However, most have been claiming that Apple would not put an A14 in a MacBook. It would be an A14X or something like that. If a Apple really did that, that would be very impressive performance for such a small package, and actually probably overkill for most MacBook users.

 

[Stridr69]

This is a good starting point for Apple, the most basic and simple Mac should get the ASIC processor and then work your way up. This unit will be great for simple word processing, web browsing, spreadsheets. The basics of computing for people. This will fill the need of the entry level computer user needing a real computer with a keyboard.

You mean an iPad with a built-in keyboard, yes?

 

[Stridr69]

If it runs Chrome, MS Office and my workplace’s VPN software (F5 VPN), then I’d very significantly consider an A14X MacBook.

Suggest you purchase/find someone with the latest iPad pro and see if those apps run at all. You'll have your answer...

 

[Stridr69]

Wonder if this will have a fan or not. I'm assuming it will need a fan, but this will be really interesting.

My bigger concern will be how locked down this is.

Does Fort Knox come to mind?

 

[Stridr69]

Inside Apple's iPad: the A4 processor | Appleinsider

Apple raised eyebrows during the original iPad unveiling when it casually noted that the new device would be using a new, Apple-designed System on a Chip named the A4. The company's future custom silicon plans are likely to expand even faster following its acquisition of Intrinsity, a privately...

appleinsider.com

.
Custom Chip not off the shelf.

And note how locked down iPads/iPhones are...

 

[Stridr69]

Any talk about a iPad Pro being able to dock and then run MacOS?

Yes...just before Hell freeze's over..

 

[Arctic Moose]

Does Fort Knox come to mind?

No.

Macs already have the T2 chip, I don’t see how running on AS instead of Intel would make any difference.

They will be running Big Sur, which is in public beta, so we already know how locked down it will be.

 

[chucker23n1]

You and I have had this discussion before, but I'll reiterate, with respect, that these are (as you know) synthetic benchmarks, and thus may not correlate well with real-world performance. I recall a quote from Linus Torvalds specifically warning against using Geekbench for cross-platform comparisons.

You might wish to invoke comparisons done with SPEC06, but this is also a synthetic benchmark. CERN, for instance, found that SPEC06 showed good correlation to its actual workloads in 2009, but that was no longer the case in 2019. Further, SPEC CPU 2017 also didn't correlate well with its current workloads. [https://indico.cern.ch/event/773049...931/3200035/20191104-CHEP2019-BMK-AV-v005.pdf]

So we won't actually know if AS "trounces" Intel in per-core performance until we see a wide range of real-world benchmarks from independent sources. As an industry insider you may have access to that info., but the rest of us can only make judgements based on the data to which we have access; and no reliable (i.e., independent, real-world) comparative info. has yet been publicly released.

I don’t understand the point here. A comparison of CPUs is going to be synthetic no matter what, but it’s close as we can get right now. When people ask how the two compare, this is the data we have to work with.

The main argument against Geekbench I would accept is it isn’t representative of lengthy workloads, which is increasingly a problem as both Apple and Intel heavily throttle their CPUs as soon as they can either afford to (low load) or are required (high thermals). So, for “how many hours will it take to transcode a podcast”, it’s not a great choice.

 

[deevey]

I don’t understand the point here. A comparison of CPUs is going to be synthetic no matter what, but it’s close as we can get right now. When people ask how the two compare, this is the data we have to work with.

The main argument against Geekbench I would accept is it isn’t representative of lengthy workloads, which is increasingly a problem as both Apple and Intel heavily throttle their CPUs as soon as they can either afford to (low load) or are required (high thermals). So, for “how many hours will it take to transcode a podcast”, it’s not a great choice.

We already have reasonable ideas on how an A12x will take to export video on an iPad vs 2019 i9 intel MacBook, at around about 3/4 the speed of the MacBook.

An A14 (whatever) Air/Macbook IMHO is going to be at least as fast or faster compared to any other similar form factor laptop.

 

[theorist9]

I don’t understand the point here. A comparison of CPUs is going to be synthetic no matter what, but it’s close as we can get right now. When people ask how the two compare, this is the data we have to work with.

Happy to explain.

My point isn't that we shouldn't be looking at Geekbench or SPEC06. I've done this myself. It's fun to see what those numbers are, and to speculate about what they *might* mean.

Rather, my point is that such examinations should be tempered with a healthy dose of informed skepticism. In particular, they should be accompanied by the awareness that these results (comparing a synthetic benchmark in iOS with one in MacOS) might not correlate well with how the platforms will actually compare when assessed with real-world workloads, when used within the same OS.

By contrast, too often (IMO) on these forums, people attach a lot more certainty to what these numbers represent than they actually have.

To put it more colloquially, they should be taken with a grain of salt. When someone uses language like "their per-core performance trounces Intel..." [emphasis mine], there's no grain of salt there.

It may be the case that AS does trounce Intel in per-core performance; and we of course all hope it does because, hey, progress. And the fact that Apple has marketed AS as offering increased performance suggests it will, since of course they have done that real-world testing internally (then again, sometimes their marketing makes wild claims that are completely wrong). But, based on the data we have, we don't yet *know* it will "trounce" Intel, when it comes to real-world performance.

The main argument against Geekbench I would accept is it isn’t representative of lengthy workloads, which is increasingly a problem as both Apple and Intel heavily throttle their CPUs as soon as they can either afford to (low load) or are required (high thermals). So, for “how many hours will it take to transcode a podcast”, it’s not a great choice.

FWIW, here is Torvalds' specific comment warning against using Geekbench ("GB") results in iOS to predict performance in a desktop OS, from May 31, 2019 (link:https://www.realworldtech.com/forum/?threadid=185109&curpostid=185132)

By: Linus Torvalds (torvalds.delete@this.linux-foundation.org), May 31, 2019 8:28 am
Room: Moderated Discussions
rwt (rwt.delete@this.noemail.com) on May 31, 2019 6:32 am wrote:
>
> But it's better to compare it with Windows system results only.
> clang results in another systems are faster than Visual C++ results.
> Just don't mix Visual C++ results and clang results in one comparison.

I'm not sure how much of it is the compiler per se, and how much of it is "system issues". But yes, I very much agree that you shouldn't treat GB as a "cross system" benchmark.

We've seen this before: cellphones tend to have simpler libraries that are statically linked, and at least iOS uses a page size that would not be relevant or realistic on a general purpose desktop setup.

At the other spectrum of issues, cellphones obviously tend to have thermal limits that may not be relevant in other form factors, although some of the "benchmark mode" tweaks might hide some of that.

 

[theorist9]

We already have reasonable ideas on how an A12x will take to export video on an iPad vs 2019 i9 intel MacBook, at around about 3/4 the speed of the MacBook.

An A14 (whatever) Air/Macbook IMHO is going to be at least as fast or faster compared to any other similar form factor laptop.

In these comparisons, do you know that the app used to export the video in iOS is identical to that used to export the video in MacOS? And do you know that the OS itself has no effect on performance?

Wolfram, for instance, has decades of experience optimizing Mathematica on both Windows and Linux, yet I've read reports that the app runs faster on Linux.

So how do you know what you are seeing with the video export isn't due to a difference between the performance of that app under iOS vs MacOS, rather than due to the difference in the processors? How do you disentangle the two?

I'm not saying AS won't be faster than Intel under MacOS. I'm saying we don't yet have proof of this.

 

[xraydoc]

Suggest you purchase/find someone with the latest iPad pro and see if those apps run at all. You'll have your answer...

Not sure what you mean by that. Of course those apps exist for iOS -- I have them on my 1st gen 11" iPad Pro. But do we know if Chrome & Office run on macOS for ARM? And how about VPN software? Will that run on macOS for ARM without modification?
Nope -- won't know until the machine is released or Apple publishes more info on macOS for ARM.

 

[chucker23n1]

Happy to explain.

My point isn't that we shouldn't be looking at Geekbench or SPEC06. I've done this myself. It's fun to see what those numbers are, and to speculate about what they *might* mean.

Rather, my point is that such examinations should be tempered with a healthy dose of informed skepticism. In particular, they should be accompanied by the awareness that these results (comparing a synthetic benchmark in iOS with one in MacOS) might not correlate well with how the platforms will actually compare when assessed with real-world workloads, when used within the same OS.

By contrast, too often (IMO) on these forums, people attach a lot more certainty to what these numbers represent than they actually have.

I don't think the skepticism has borne out thus far. Apple's CPUs simply happen to be surprisingly fast. (Add to that that Intel, especially in the 14nm era which still hasn't concluded, has done a poor job competing.)

To put it more colloquially, they should be taken with a grain of salt. When someone uses language like "their per-core performance trounces Intel..." [emphasis mine], there's no grain of salt there.

But it does.

Don't want Geekbench? Fine, use Google Octane.

• The 2019 16-inch MacBook Pro (14nm, 45W) scores 51,235
• The 2020 MacBook Air (10nm, 10W) scores 37,229
• The iPhone 11 (7nm*, ~5W) scores 48,216

*) roughly like Intel's 10nm

And if that's still too synthetic to that, just try Safari on a Mac and an iPhone.

It may be the case that AS does trounce Intel in per-core performance; and we of course all hope it does because, hey, progress. And the fact that Apple has marketed AS as offering increased performance suggests it will, since of course they have done that real-world testing internally (then again, sometimes their marketing makes wild claims that are completely wrong). But, based on the data we have, we don't yet *know* it will "trounce" Intel, when it comes to real-world performance.

But we know that 2019 Apple chips already beat 2019 Intel chips (Ice Lake) despite half the TDP. And when you compare against Comet Lake instead, which you have to, because most Intel chips aren't 10nm yet, it's an utter bloodbath, which Intel won't be able to resolve until at least late 2021.

We don't technically know what an Apple chip in a Mac will be like: the thermals will differ, and odds are the core count and clock will differ. Heck, they might change some features. Less Neural Engine, perhaps. Beefier GPU instead. We also don't know how the A14 compares to the A13.

If your issue here is the polemic use of "trounce", I don't think your priorities are right. We're not trying to get published in a scientific journal.

I mean, yeah, people here get overly excited sometimes. I don't think we'll see a "10x boost" some are expecting.

FWIW, here is Torvalds' specific comment warning against using Geekbench ("GB") results in iOS to predict performance in a desktop OS, from May 31, 2019 (link:https://www.realworldtech.com/forum/?threadid=185109&curpostid=185132)

By: Linus Torvalds (torvalds.delete@this.linux-foundation.org), May 31, 2019 8:28 am
Room: Moderated Discussions
rwt (rwt.delete@this.noemail.com) on May 31, 2019 6:32 am wrote:
>
> But it's better to compare it with Windows system results only.
> clang results in another systems are faster than Visual C++ results.
> Just don't mix Visual C++ results and clang results in one comparison.

I'm not sure how much of it is the compiler per se, and how much of it is "system issues". But yes, I very much agree that you shouldn't treat GB as a "cross system" benchmark.

We've seen this before: cellphones tend to have simpler libraries that are statically linked, and at least iOS uses a page size that would not be relevant or realistic on a general purpose desktop setup.

At the other spectrum of issues, cellphones obviously tend to have thermal limits that may not be relevant in other form factors, although some of the "benchmark mode" tweaks might hide some of that.

But those phones will have statically linked libraries and a different page size in real-world use as well.

He's maybe arguing that you can't draw conclusions on how fast the mobile CPU would be in a desktop, but what people are mainly doing is comparing the overall system, and by his argument, that comparison is still valid.

 

[chucker23n1]

Not sure what you mean by that. Of course those apps exist for iOS -- I have them on my 1st gen 11" iPad Pro. But do we know if Chrome & Office run on macOS for ARM? And how about VPN software? Will that run on macOS for ARM without modification?
Nope -- won't know until the machine is released or Apple publishes more info on macOS for ARM.

Chrome already runs on ARM on Windows 10, Ubuntu and others, and Apple itself listed Chromium as one of the projects they've helped port:

So yeah, that'll happen.

Similarly for Office. It already runs on iPads. I can see Office for ARM not including all features the x86 version has in its initial version (stuff like VBA).

VPN software: Big Sur adds some limitations here, but OpenVPN already runs on ARM.

 

[Arctic Moose]

But do we know if Chrome & Office run on macOS for ARM?

Microsoft Office has already been previewed on Apple Silicon, so I would be very surprised if it isn’t available at launch.

And how about VPN software?

There’s nothing particularly special about VPN software, so I don’t see why not. Even if the particular product you use lags behind it should be fine on Rosetta 2 until it is updated.

 

[xraydoc]

Microsoft Office has already been previewed on Apple Silicon, so I would be very surprised if it isn’t available at launch.
...
There’s nothing particularly special about VPN software, so I don’t see why not. Even if the particular product you use lags behind it should be fine on Rosetta 2 until it is updated.

That's all very good to know, as even Microsoft's own Surface Pro X (which runs Windows 10 for ARM) has a lot of problems with both 64-bit software, Windows extensions like Dropbox's Windows Explorer integration and certain VPN software. I came very close to buying one for work but discovered these fatal (for me) limitations. The Dropbox and VPN issues completely prevent me from using it for my work purposes.

 

[deevey]

Microsoft Office has already been previewed on Apple Silicon, so I would be very surprised if it isn’t available at launch.

There’s nothing particularly special about VPN software, so I don’t see why not. Even if the particular product you use lags behind it should be fine on Rosetta 2 until it is updated.

VPN's run perfectly on iOS, cant why would there be issues with Apple Silicon natively.

 

[theorist9]

I don't think the skepticism has borne out thus far. Apple's CPUs simply happen to be surprisingly fast. (Add to that that Intel, especially in the 14nm era which still hasn't concluded, has done a poor job competing.)



But it does.

Don't want Geekbench? Fine, use Google Octane.

• The 2019 16-inch MacBook Pro (14nm, 45W) scores 51,235
• The 2020 MacBook Air (10nm, 10W) scores 37,229
• The iPhone 11 (7nm*, ~5W) scores 48,216

*) roughly like Intel's 10nm

And if that's still too synthetic to that, just try Safari on a Mac and an iPhone.



But we know that 2019 Apple chips already beat 2019 Intel chips (Ice Lake) despite half the TDP. And when you compare against Comet Lake instead, which you have to, because most Intel chips aren't 10nm yet, it's an utter bloodbath, which Intel won't be able to resolve until at least late 2021.

We don't technically know what an Apple chip in a Mac will be like: the thermals will differ, and odds are the core count and clock will differ. Heck, they might change some features. Less Neural Engine, perhaps. Beefier GPU instead. We also don't know how the A14 compares to the A13.

If your issue here is the polemic use of "trounce", I don't think your priorities are right. We're not trying to get published in a scientific journal.

I mean, yeah, people here get overly excited sometimes. I don't think we'll see a "10x boost" some are expecting.



But those phones will have statically linked libraries and a different page size in real-world use as well.

He's maybe arguing that you can't draw conclusions on how fast the mobile CPU would be in a desktop, but what people are mainly doing is comparing the overall system, and by his argument, that comparison is still valid.

It doesn't matter how many synthetic benchmarks you throw at this, it doesn't argue against the general principle I've been articulating:

I acknowledge these results suggest AS under MacOS will be faster than Intel under MacOS, with the apps we actually run. What bothers me is that people are speaking as if we know this (e.g., "[AS] trounces Intel"). But we don't yet know this, nor will we know this, until we have independent real-world data comparing a wide range of apps in AS on MacOS and AS on Intel. That's all I've been saying all along. This seems obvious, so I don't know why it's so controversial.

Yes, AS performs significantly faster running benchmarks in iOS than Intel does running benchmarks in MacOS. But that doesn't necessarily mean the apps we actually use will run faster in AS under MacOS than they do in Intel under MacOS.

I'm thinking about this the same way any scientist would:

The problem with the current comparisions is that they introduce two possible confounding variables, the app and the OS. If you want to know that a certain app is faster in AS under MacOS than it is in Intel under MacOS, you need to compare the same app in the same OS, changing only the chip.

E.g., just because synthentic benchmarks in iOS on AS are signficantly faster than those benchmarks in MacOS on Intel, that doesn't necessarily mean Mathematica in MacOS on AS will be significantly faster than Mathematica in MacOS on Intel.

 

[johngwheeler]

It doesn't matter how many synthetic benchmarks you throw at this, it doesn't argue against the general principle I've been articulating:

I acknowledge these results suggest AS under MacOS will be faster than Intel under MacOS, with the apps we actually run. What bothers me is that people are speaking as if we know this (e.g., "[AS] trounces Intel"). But we don't yet know this, nor will we know this, until we have independent real-world data comparing a wide range of apps in AS on MacOS and AS on Intel. That's all I've been saying all along. This seems obvious, so I don't know why it's so controversial.

Yes, AS performs significantly faster running benchmarks in iOS than Intel does running benchmarks in MacOS. But that doesn't necessarily mean the apps we actually use will run faster in AS under MacOS than they do in Intel under MacOS.

I'm thinking about this the same way any scientist would:

The problem with the current comparisions is that they introduce two possible confouding variables, the app and the OS. If you want to know that a certain app is faster in AS under MacOS than it is in Intel under MacOS, you need to compare the same app in the same OS, changing only the chip.

E.g., just because synthentic benchmarks in iOS on AS are signficantly faster than those benchmarks in MacOS on Intel, that doesn't necessarily mean Mathematica in MacOS on AS will be significantly faster than Mathematica in MacOS on Intel.

You are demonstrating a dangerous and subversive tendency to use scientific methodology to validate hypotheses with empirical data, in a calm rational manner, devoid of emotion and hyperbole. Just stop it!

This is 2020 - Fanatical belief and zealotry, and a complete aversion to considering the available evidence before broadcasting your opinion should be all you need! ;-)

 

[Arctic Moose]

E.g., just because synthentic benchmarks in iOS on AS are signficantly faster than those benchmarks in MacOS on Intel, that doesn't necessarily mean Mathematica in MacOS on AS will be significantly faster than Mathematica in MacOS on Intel.

A reddit post on an AppleInsider article and YouTube video (both removed, but mirrors available in the post) including Geekbench results running on Big Sur on the DTK:

https://www.reddit.com/r/apple/comments/hi5k1d

From the comments on the post:

This wasn't a missed opportunity, this was a failure of note. Why are there no compilation tests, why isn't there anything for power apps like Final Cut? Why didn't they test Rosetta 2 with other things?

Yeah, that.

 

[chucker23n1]

The problem with the current comparisions is that they introduce two possible confounding variables, the app and the OS. If you want to know that a certain app is faster in AS under MacOS than it is in Intel under MacOS, you need to compare the same app in the same OS, changing only the chip.

Yes, obviously. I do think people overestimate the difference between iOS and macOS, though. It’s basically a different UI framework at this point. Tons of frameworks underneath after the same.

 

[theorist9]

A reddit post on an AppleInsider article and YouTube video (both removed, but mirrors available in the post) including Geekbench results running on Big Sur on the DTK:

https://www.reddit.com/r/apple/comments/hi5k1d

From the comments on the post:

This wasn't a missed opportunity, this was a failure of note. Why are there no compilation tests, why isn't there anything for power apps like Final Cut? Why didn't they test Rosetta 2 with other things?

Yeah, that.

Not sure how much it tells us, but it is another fun data point. Essentially, running GB5 in MacOS on both an A12Z and a Core i3, both within Mac Mini cases, they found the A12Z's single core performance is ~10% slower than the Core i3's (845 vs 949)—but with the caveat that the GB5 running on the A12Z was auto-recompiled by Rosetta 2.

It's hard to go from that to how a production A14X would compare to an i7/i9, where Rosetta 2 is not used.

The problem with AI testing "power apps" like Final Cut at this point is that they don't have access to the AS-optimized code that will eventually be released for Final Cut/Adobe CC/Mathematica etc., only to X86-optimized code that would then be recompiled by Rosetta2.

Screenshots from the video (ignore the multi-core score for the iPad, since it uses both its low-power and performance cores, while the DTK has its low-power cores disabled):

 

[Yebubbleman]

Funny how we see things differently.
To me it makes complete sense to have the MB as the first Mac using Apple silicon. The recent MBs have been entirely about form factor, lightness and battery duration. A perfect fit for an A14x processor.

As for more powerful Macs, they will need more powerful processors, specifically designed for Macs (as opposed to iPhones and or iPads) and those are not ready yet.

You have a ton of skeptics out there that are not wholly convinced that switching to the ARM64 architecture in Apple Silicon will be a performance improvement when moving from x86-64 in Intel. You're not going to disprove that easily when producing a Mac that there isn't a current Intel equivalent to compare it to. If Apple's rumored Apple Silicon 13" MacBook Pro, let's say, provides performance on par with the current entry level 16" MacBook Pro, then that disproves the skeptics. Similarly, if an Apple Silicon MacBook Air outperforms the highest-end Intel 2020 4-port 13" MacBook Pro with it's 10th Gen quad-core U-series processors, then that disproves skeptics and shows that Apple Silicon SoCs can beat out Intel processors ON THE SAME PRODUCTS.

People tend to forget that sluggish performance was only ONE of the many things that were disliked about the 12" Retina MacBook.

 

[chucker23n1]

I don’t think pleasing skeptics ranks highly among Apple’s priorities.

 

[supertiffany]

so what happened to this.