Apple Silicon: The Complete Guide

[Brammy]

When I look at what Apple's SoC can do on devices without a fan, proper cooling and constrained size (even an iPad is constrained when you figure how much battery is in there), I think it is inevitable.

I would think the first device would be the old MacBook 12". That one port device is perfect for an Arm-based back. You can then start to move the MacBook Pros and iMacs there. Maybe the iMac Pro and Mac Pro remain Xeon based for a while.

 

[theluggage]

"I don't think Apple are going that way at the moment "

Which way do you think they're going?

Just my speculation - but they're putting a lot of effort into larger, more powerful iPads and adding more "power user"-ish features into iPadOS. I still don't see an iPad Pro (even souped up) as a credible replacement for a MacBook Pro - but iPad Pro (with keyboard cover) could be a serious competitor to a 12" ARM MacBook or entry-level ARM Air (and maybe that's why the 12" MB was dropped). I think that, if ARM Macs are imminent, the iPad focus would still be on mobile consumption.

They're also working with Adobe to bring "full" (for a particular value of "full") Photoshop to iPad.

"Catalyst" is also going to blur the distinction between iPad and Mac, potentially resulting in a lot of "low end" Mac software becoming near-identical to the iPad version, with just the minimum essential UI changes - again, reducing the incentive to buy a low-end Mac over an iPad.

They've also invested in a brand new Mac Pro built around the Xeon which commits them for supporting Intel for several more years (unless they want to become a joke in the Pro market).

So my guess is that, while I'm sure that they've got contingency plans for moving Mac to ARM (just as they had for PPC to Intel) if only to keep Intel on their toes, "plan A" is still to push the iPad further into Mac territory rather than vice-versa.

Remember that iPadOS locks people into the App Store and makes them more dependent on "services" than MacOS - cha-ching!. Trying to suddenly lock down MacOS would generate a lot of negative press - gradually migrating people to iPadOS would boil the frog more gently...

 

[Mago]

Why would they need 4-6? The ONLY reason for that now is Intel wants to have a tiered pricing structure so they can charge more for more performance. There’s no tiered pricing for either iPhone or iPad, they just keep old ones on the market for folks that want them.

You can't put the same cpu on the MacBook (low power low performance 4 cores), MacBook pro - Mac mini (low power, medium performance 6-8 cores), iMac (high performance 6-8 cores, no power restrictions), Mac Pro/iMac Pro (very high performance , many cores 8-28).

There is also a single thing thread performance gap growing too fast among ARM-intel/AMD ARM can't keep up, further now the CPU war is hot again by almost a decade AMD was irrelevant so Intel barely improved their CPU line, by 5-6 years IPC just grew about 28%, while past year only it grew 26% on AMD pressure, meanwhile ARM desktop struggles to keep up with ULwP-core CPUs.

FYI most applications require single thread is the most user sensitive benchmark.

 

[MikeZTM]

You can't put the same cpu on the MacBook (low power low performance 4 cores), MacBook pro - Mac mini (low power, medium performance 6-8 cores), iMac (high performance 6-8 cores, no power restrictions), Mac Pro/iMac Pro (very high performance , many cores 8-28).

There is also a single thing thread performance gap growing too fast among ARM-intel/AMD ARM can't keep up, further now the CPU war is hot again by almost a decade AMD was irrelevant so Intel barely improved their CPU line, by 5-6 years IPC just grew about 28%, while past year only it grew 26% on AMD pressure, meanwhile ARM desktop struggles to keep up with ULwP-core CPUs.

FYI most applications require single thread is the most user sensitive benchmark.

A13 is already running higher "IPC"/performance for same clock speed than Intel Sunny Cove/Icelake. It got 9900k level single core performance at 2.65GHz compared to 4.6-5.0GHz boost frequency.

They could create high performance core only and from quad core to dodeca core chip just like todays hex core and octa core version of A12/A12X

And A12X single core performance was already faster than Intel's ULP i5/i7 mobile chip.

This is based on SPEC2006 estimate and not related to Geekbench.

 

[Mago]

A13 is already running higher "IPC"/performance for same clock speed than Intel Sunny Cove/Icelake. It got 9900k level single core performance at 2.65GHz compared to 4.6-5.0GHz boost frequency.

They could create high performance core only and from quad core to dodeca core chip just like todays hex core and octa core version of A12/A12X

And A12X single core performance was already faster than Intel's ULP i5/i7 mobile chip.

This is based on SPEC2006 estimate and not related to Geekbench.

Guess which benchmark is useless, and which one is actually the industry std and why?

It's easy to cheat Spec2006, thus other benchmarks are based on common workloads involving out of order not easy to predict instructions streams, spec2006 is like to use x265 encoding as benchmark, it's an well understood algorithm predictable, and anything predictable can be executed in a single clock pulse, even multiple instances of any predictable algorithm.

Come here with A13's single core real world benchmark not a benchmark for 2006 algorithm.

There are also more things that single core benchmark making a difference, you have pcie lines (needed for peripherals as GPU s), memory channels and memory bus sustained bandwidth, these are expensive technically complex features dividing A13 from true desktop CPU, include this memory pagination, memory address width (Max memory size), DMA controllers (indispensable for efficient peripherals), a bunch of things a pc/sever does a phone or tablet don't care about.

Best A13 geekbench is 1334 single thread
Best x86 geekbench is AMD Ryzen 2700X at 1472, now tell me a joke.

If you want to discuss about vacuum cleaners don't argue with the sellers ads as arguments to an mechanical engineer that can teach you a word or two on fluid Dynamics and thermodynamics.

 

[Unregistered 4U]

You can't put the same cpu on the MacBook...

They could use the same 8>core processor for everything Mobile. They’ve already got a well performing 8 core processor in the iPad Pro, so maybe they’d start more at 16 cores. Even if you allow for an increase in heat for clocking it higher, the larger aluminum body could dissipate that readily. The higher clocked ones could go into the systems with more room for thermal control (or fans). Apple has been using laptop processors in iMacs, too. So, clock it higher for the iMac.

I could see a second processor being available for desktops OTHER than the iMac, again where it clocks determines what it goes into.

I could potentially see a third tier for higher performance, but even then, that’s still a far cry from the various features capabilities Intel offers just to have an excuse to charge more. And, unlike Intel, these would all have the same CPU level features, like the machine learning engine, fast memory subsystem, etc.

FYI most applications require single thread is the most user sensitive benchmark.

Consider this, though. For the VAST majority of people, their single threaded needs are well within the realms of both Intel’s low end AND Apple’s A-series chips. Intel’s single threaded speed could be seventy times ARM’s and that webpage isn’t going to refresh, or that YouTube video play or that the Facebook “like” button highlight, any faster.

 

[Mago]

They could use the same 8>core processor for everything Mobile. They’ve already got a well performing 8 core processor in the iPad Pro, so maybe they’d start more at 16 cores. Even if you allow for an increase in heat for clocking it higher, the larger aluminum body could dissipate that readily. The higher clocked ones could go into the systems with more room for thermal control (or fans). Apple has been using laptop processors in iMacs, too. So, clock it higher for the iMac.

I could see a second processor being available for desktops OTHER than the iMac, again where it clocks determines what it goes into.

I could potentially see a third tier for higher performance, but even then, that’s still a far cry from the various features capabilities Intel offers just to have an excuse to charge more. And, unlike Intel, these would all have the same CPU level features, like the machine learning engine, fast memory subsystem, etc.


Consider this, though. For the VAST majority of people, their single threaded needs are well within the realms of both Intel’s low end AND Apple’s A-series chips. Intel’s single threaded speed could be seventy times ARM’s and that webpage isn’t going to refresh, or that YouTube video play or that the Facebook “like” button highlight, any faster.

Ok smart kiddo, say that to Tim cook, but don't forget about to include r&d and/or IP costs for: 8-128 pcie lines for GPU and peripherals, plus interconnection fabric for 4-32 cores configurations, also add this the costs to migrate macOS to ARM, then account the winnings: a barely worse cpu than current Intel/amd pre-cpu war race (which just begun and to which arm isn't invited), another winning is apple liability on CPU issues, no chance to quickly migrate to an alternative provider's in a case of catastrophic failure event.

This is why computers are designed by engineers not journalists.

 

[Unregistered 4U]

Ok smart kiddo, say that to Tim cook,

I’m not saying anything to Tim Cook. Intel believes Apple will transition to ARM next year and I’m speculating how that might not require a plethora of chip variants.

This is why computers are designed by engineers not journalists.

Steve Jobs wasn’t an engineer OR a designer, he knew what he wanted out of technology and found engineers and designers that were NOT so full of themselves that they were able to make his vision a reality

 

[wtn]

A Mac with an arm? Or ARM? I'd rather the latter!

Right!! ARM is an acronym and should be capitalized as such in American English.

 

[InuNacho]

" The only professionals that still use macOS SHOULD be the ones that are using FCPX and Logic Pro"

And Xcode. iOS developers still constitute the largest base of professional users for Apple's ecosystem.

Let me know when Finder is on the iPad and I can play with any file type I want.

 

[derek]

Q: "When Might Apple Release an ARM-Based Mac?"
A: Never.

I think of this rumor as a zombie that eats both our time and brains. There are no viable reasons for Apple to dump Intel Macs. The reasons to not dump Intel Macs bury this undead rumor every time someone digs it up again. Let it stay dead. If you don't understand Computing 101, you're not going to comprehend why this rumor has no life. Digging it up again doesn't offer it life or promise thereof. Bereft of life, it rests in peace. This is a dead parrot.

 

[OxCxDx]

Let me know when Finder is on the iPad and I can play with any file type I want.

I am waiting for this, big time.

 

[Unregistered 4U]

I just wonder whether an arm based mac makes any sense in a market that is getting smaller and more niche.

I’ve been thinking about this... I agree that in a way, it doesn’t make sense to spend that money on ARMing the Mac, when, internally, Apple could have iPadOS builds (and hardware) right now that may be approaching or exceeding macOS levels of performance. Would Apple pour the R&D into producing ARM Macs that would soon be supplanted by iPadOS, OR focus on producing a version of iPadOS that would meet the needs of the folks buying the majority of Macs today? That being, folks that have never bought a Mac before and low end users.

What if, instead of doing some future bake-off between an Intel Mac and an Arm Mac, Apple instead does a bake-off between a Mac running FCPX and iPadOS running FCPX? I mean, this is the same company that some expect have the capability to make a very complex transition as simple as possible. CERTAINLY they have the ability to create a touch enabled version of FCPX running on, say, a 12.9 inch (or larger) iOS screen. I can’t even imagine how something as complex as FCPX would even work, but if they could figure it out... er, if they HAVE ALREADY figured it out, then maybe that “migration” that Intel thinks is going to happen next year is a much huger migration than even their contract numbers indicate. Interesting thought.

 

[Unregistered 4U]

There are no viable reasons for Apple to dump Intel Macs.

There IS one huge reason. Intel promised processors that would be suitable for the high end of Apple’s top sellers, the MacBook and MacBook Pro’s and failed year over year. Any company that has an opportunity to protect their top sellers from being impacted by failures of other companies should take steps to do so.

 

[Mago]

There IS one huge reason. Intel promised processors that would be suitable for the high end of Apple’s top sellers, the MacBook and MacBook Pro’s and failed year over year. Any company that has an opportunity to protect their top sellers from being impacted by failures of other companies should take steps to do so.

Intel didn't failed, they just keep their rival Pace, until AMD released Zen architecture Intel was alone in the CPU market no reason to justify investment, now AMD in part thanks to apple investment -in stock- and collaboration releasing Jim Keller (the real genius behind Apple SOC) then Keller designed the Zen architecture and trained a competitive r&d team before leaving for Tesla.

Now Intel is behind AMD, so they are investing in r&d again delivering 22% performance improvement in a single generation (that took 5-6 generations before), also halving prices (also beneficial for Macs)

 

[Mago]

Personally I think the Future of apple is to continue developing iPadOS so it can replace effectively the base MacBook (current iPad pro l the iMac21 (a desktop giant iPad), and Mac mini (an Apple TV "pro" with the iPad pro's soc) it will be consistent providing the general public with a friendly device suitable for theirs needs, but keeping the Professional Macs to address high performance users (app developers, video studios, AI, bigdata, servers, STEM) where s true compute rig is primordial.

 

[Unregistered 4U]

Intel didn't failed

Yes, they failed. They announced their roadmap (no one twisted their arm), but when it came time to deliver, they delayed critical parts Apple needed (primarily, laptop processors that can utilize LPDDR4 memory) for years. That “delay” is a failure of Intel to meet their commitments.

Apple’s iPhones have used LPDDR4 for years now, Intel STILL hasn’t shipped a processor that supports LPDDR4 AND would be suitable for a MacBook Pro. If Intel continues to fail to meet their own published roadmaps, then if low power, high performance RAM is important, it makes sense for Apple to hedge their bets.

 

[Mago]

Yes, they failed. They announced their roadmap (no one twisted their arm), but when it came time to deliver, they delayed critical parts Apple needed (primarily, laptop processors that can utilize LPDDR4 memory) for years. That “delay” is a failure of Intel to meet their commitments.

Apple’s iPhones have used LPDDR4 for years now, Intel STILL hasn’t shipped a processor that supports LPDDR4 AND would be suitable for a MacBook Pro. If Intel continues to fail to meet their own published roadmaps, then if low power, high performance RAM is important, it makes sense for Apple to hedge their bets.

Repeat, they didn't failed, no need to build this part as no rival produced it notwithstanding other markets already had it (for arm lpddr4 is Paramount, not the same for other platforms), apple didn't lose anything as no rival pc manufacturer also have access to it, only one that lose where those fanboy speculating about why there is no a 32gb mb with lp CPU.

Your reasoning is not the same as a product manager or engineer, just fanboyism

 

[CarlJ]

I would hope that someone’s not buying one to run open source software that could run on ANY platform, because performance per dollar isn’t in Apple’s favor. BUT, if someone has the money, and that’s their preference, then I’m not standing in the way of them being happy

Why? If you're going to be sitting in front of the machine 8+ hours a day for the next three years, even hundreds of dollars in upfront costs pales in comparison to many thousands of hours of dealing with the user interface.

How do you think employee productivity fares in a scenario where they're happy and the UI/UX of the OS stays out of their way, vs. a scenario where the tools you give them to use are a constant source of small annoyances?

Do you supply the highly-skilled/highly-paid employees the absolute cheapest tool available, or the one that equips them to be the most effective? (Assuming the difference in price isn't many orders of magnitude - and Windows/Linux vs Mac is not Toyota vs Lamborghini, pricewise.)

 

[smulji]

Personally I think the Future of apple is to continue developing iPadOS so it can replace effectively the base MacBook (current iPad pro l the iMac21 (a desktop giant iPad), and Mac mini (an Apple TV "pro" with the iPad pro's soc) it will be consistent providing the general public with a friendly device suitable for theirs needs, but keeping the Professional Macs to address high performance users (app developers, video studios, AI, bigdata, servers, STEM) where s true compute rig is primordial.

"Personally I think the Future of apple is to continue developing iPadOS so it can replace effectively the base MacBook "

I think the future of Apple is to continue developing iPadOS so it can effectively replace the entire Mac laptop line. To go along with that, I wouldn't be surprised if Apple releases a bigger iPad Pro (15"?).
[automerge]1573498493[/automerge]

Repeat, they didn't failed, no need to build this part as no rival produced it notwithstanding other markets already had it (for arm lpddr4 is Paramount, not the same for other platforms), apple didn't lose anything as no rival pc manufacturer also have access to it, only one that lose where those fanboy speculating about why there is no a 32gb mb with lp CPU.

Your reasoning is not the same as a product manager or engineer, just fanboyism

Where Intel failed was with their 10nm process manufacturing. That set them back a lot.

 

[Mago]

"Personally I think the Future of apple is to continue developing iPadOS so it can replace effectively the base MacBook "

I think the future of Apple is to continue developing iPadOS so it can effectively replace the entire Mac laptop line. To go along with that, I wouldn't be surprised if Apple releases a bigger iPad Pro (15"?).
[automerge]1573498493[/automerge]

Where Intel failed was with their 10nm process manufacturing. That set them back a lot.

Failed to timely invest, this is right, but it's CPU still on short supply, so ?

 

[Unregistered 4U]

Repeat, they didn't failed

So this is what an Intel fanboy looks like. cool!
[automerge]1573505396[/automerge]

Do you supply the highly-skilled/highly-paid employees the absolute cheapest tool available, or the one that equips them to be the most effective?

This is why I said, “if someone has the money, and that’s their preference, then I’m not standing in the way of them being happy” People can use whatever they want to do whatever job they want! I am in 100% agreement with what you said.

 

[Mago]

So this is what an Intel fanboy looks like. cool!
[automerge]1573505396[/automerge]

This is why I said, “if someone has the money, and that’s their preference, then I’m not standing in the way of them being happy” People can use whatever they want to do whatever job they want! I am in 100% agreement with what you said.

Actually I'm AMD/nVidia fanboy but I don't use Mac for r&d I do this now in Linux Ubuntu or the cloud, often with Opensource tools but I avoid those with Gpl2-3 only Opensource licenses I advise are Apache, MIT, BSD and lgpl2 if you only link libraries otherwise if you touch code only real true free non viral licenses as MIT BSD and Apache & similar.

 

[CyberBob859]

iPadOS has started to migrate into a more effective OS for tablets (still has a bit more to go). Can it be used as the basis of a clamshell system Yes! And thats what Apple needs to embrace!

What I foresee is a iPad clamshell filling in the MacBook slot. Thats the market Apple needs to enter. They are almost there now with the iPad Pro line with the keyboard covers. Many people just don't like the cover keyboards and the lack of a trackpad also forces people to touch the screen. A MacBook type of keyboard with a touchpad LCD surface with a full display above would be the ticket here!

When you consider the amount of work that has to be done to create an iPad Pro clamshell (iBook?) with modifications to iPadOS, versus an ARM-based Mac (creating and maintaining two Mac OS - ARM and Intel, getting developers on board, emulation of older Mac-programs and worries about performance, a bi-furcation of the product line, etc.), I think the clamshell iPad Pro is a lot more feasible to do.

Then consider programs written for the clamshell iPad Pro would be better candidates to convert to Mac programs via Catalyst and I'd prefer the clamshell iPad approach over an ARM Mac.

 

[MikeZTM]

Guess which benchmark is useless, and which one is actually the industry std and why?

It's easy to cheat Spec2006, thus other benchmarks are based on common workloads involving out of order not easy to predict instructions streams, spec2006 is like to use x265 encoding as benchmark, it's an well understood algorithm predictable, and anything predictable can be executed in a single clock pulse, even multiple instances of any predictable algorithm.

Come here with A13's single core real world benchmark not a benchmark for 2006 algorithm.

There are also more things that single core benchmark making a difference, you have pcie lines (needed for peripherals as GPU s), memory channels and memory bus sustained bandwidth, these are expensive technically complex features dividing A13 from true desktop CPU, include this memory pagination, memory address width (Max memory size), DMA controllers (indispensable for efficient peripherals), a bunch of things a pc/sever does a phone or tablet don't care about.

Best A13 geekbench is 1334 single thread
Best x86 geekbench is AMD Ryzen 2700X at 1472, now tell me a joke.

If you want to discuss about vacuum cleaners don't argue with the sellers ads as arguments to an mechanical engineer that can teach you a word or two on fluid Dynamics and thermodynamics.

iPhone is already using PCIe NVMe storage since A9.
I'm not saying making 20 lanes is as simple as copy paste this 4 lanes 5 times. But this definitely is not harder than bumping up single core performance by 50%.

And iPhone already have better than Intel IMC. A9 support LPDDR4 4years ago while Intel just paper-release their Icelake LPDDR4X support.

And also Apple have no reason to cheat SPEC. They are just building the chip to run iOS and optimizing the chip to do the best for that. You can not call software driven hardware development cheating.
[automerge]1573576311[/automerge]

Failed to timely invest, this is right, but it's CPU still on short supply, so ?

Their CPU is on short supply because of their fail in 10nm.
If 10nm was delivered in 2015 then current 14nm capacity is totally enough for chipset and low end CPU manufacture. Right now Intel is manufacturing Z390 on same node as their hight end CPUs and this reduced their fabs' output.

Short supply doesn't means success.

MDS/Spectre almost killed Intel's Hyper Threading Technology. I do not think any customer of Intel is riding on their chip without a plan B.