Solved! ARM Apple High-End CPU - Intel replacement

[Richie Rich]

There is a first rumor about Intel replacement in Apple products:

• ARM based high-end CPU
• 8 cores, no SMT
• IPC +30% over Cortex A77
• desktop performance (Core i7/Ryzen R7) with much lower power consumption
• introduction with new gen MacBook Air in mid 2020 (considering also MacBook PRO and iMac)
• massive AI accelerator

Source Coreteks:

 

[Doug S]

i, of course, have no inside knowledge about what sort of glue logic they would, or could, use. We know that there have been several recent introductions of cache coherent high speed interconnects that have been introduced recently. Any one of them could work. I don’t think that the choice is important, as long as it meets their needs.

As for why? It’s all a matter of volume. Apple doesn’t move anywhere near enough high end (greater than 8 core processor) processors to justify a “large” core design that is restricted to those machines. However, to remain relevant in that space, Apple needs to have a decent product. I have previously suggested that Apple is going to offer cloud processing power as a service through extensive cloud integration in a future version of Mac OS. However, if they don’t go that route, they need something.

the premise of these threads is that one A series core is dramatically faster than an x86 core. If we accept that, then it’s reasonable to suppose that fewer Mac Pro products can be better with fewer cores, but that 8 isn’t enough. If that’s a reasonable assumption, then Apple needs more cores, but still wants to make money, and therefore won’t make a chip at a loss for a low volume product. Therefore, a reasonable assumption is that their highest end A series processor will have the ability to be used in a 2p configuration.

16 super fast A series cores should be more computational power than anyone could ever need! /s

There's no way they'll go multi socket. They can put multiple chips in the same package like AMD has done. Less board space, less latency between cores for cache coherency etc. and they wouldn't need to be limited to "only" 16 cores.

I don't think they will go monolithic but I wouldn't totally rule it out. Apple pays Intel and AMD a lot of money for the CPU/GPU going into it now. That leaves Apple with a ton of room to play with as far as designing something special for the Mac Pro / iMac Pro only if they wish, and still have it cost less than what they pay Intel/AMD now.

It all depends on what they are trying to do with it. Do they just want to beat the performance of the current Mac Pro and call it a day, or do they want to make a statement? I don't know, but if they want to make a statement then I wouldn't rule anything out if they get a blank check from above to make that statement heard across the industry.

 

[Eug]

I don't see Apple going too crazy with all this. Why not:

One SoC for phone and tablet big.LITTLE 2+4
One SoC for Pro tablet, iMac and MacBook Air big.LITTLE 4+4
One SoC for MacBook Pro, iMac Pro, Mac Pro. Big all around 8(4+4), glue logic for 2P, expanded I/O.

While that may be fewer cores on the top end, with how "massively superior" Apple's A series cores are to x86 cores, end users will be completely blown away with the power that their systems have!

Probably more:

2+4 for iPhone and the lower end iPads
4+4 for iPad, Mac Mini and Macbook/Air
8+4 for MBP and iMac

There might be some overlap. Apple could for instance make a OSX 2-in-1 that sort of replaces both the iPad Pro and the Air.

Assuming the designs and core numbers of the A14 series chips follow those of A12, A12X, and A13:

I’m kinda surprised so many people think Apple is going to equip the MacBook / MacBook Air with an SoC that will be as fast as the one in the 2021 iPad Pro (A14X 4+4). There isn’t really a need for it in terms of performance. Remember, even A14 (non-X 2+4) performance will probably be competitive with A12X. A14 will likely be much faster than A12X single-core, and roughly as fast as A12X multi-core. IMO, their biggest hurdle here is not performance. Their biggest hurdle is marketing, but Apple has lots of experience with that as we all know.

Also, 8+4 is overkill for lower end iMacs and lower end MacBook Pros. 4+4 would be sufficient for those. Even if they wanted more than 4+4 in lower end MBPs and iMacs, Apple could release a 6+4 variant, limiting 8+4 to the highest end MacBook Pro and iMac models.

 

[LightningZ71]

There's always room for die recovery models in any of the stacks. To me, the bigger differentiators for the pro lines are the I/O features. Granted, they are still a bit stingy in some models, but there are usually more ports there. There's no need to waste silicon on that sort of thing for lower end parts.

Is 2p harder than 1p? Of course. But, it's certainly not a show stopper. Remember, these are going to be custom motherboards confirming to just one standard, Apple's. Going 2P deals with several issues. Lower end devices don't need more than the equivalent of dual channel DDR for memory bandwidth. There's no need to waste the silicon for that memory controller there. They still need it on higher end products.

Could they use some sort of MCM or other type of multi-die package? It's definitely possible, but it's going to take a lot of development for them to start from square one there, and it will have issues with Z height that they currently don't really have to deal with now. Apple is somewhat obsessed with thinness in their mobile devices, and an MCM makes that difficult to achieve. Cooling would also be an issue. They could also decide to do two different packaging solutions, making a slim single die version and a big MCM, but that's a bunch more expense added for a low volume project.

In the places where they would need to use multiple processors, they have "plenty" of space. None of those devices are mobile focused.

This isn't reinventing the wheel. This is just changing the type of engine that turns it.

 

[soresu]

Could they use some sort of MCM or other type of multi-die package? It's definitely possible, but it's going to take a lot of development for them to start from square one there, and it will have issues with Z height that they currently don't really have to deal with now. Apple is somewhat obsessed with thinness in their mobile devices, and an MCM makes that difficult to achieve.

You are conflating MCM with chip stacking.

Technically all chip stacks are MCM, but not all MCM are chip stacks - some like Epyc 1 and TR1/2 are just a handful of chips connected by short interconnects in a smallish area single package.

OTOH Zen4 is supposed to be going down the route of either multi logic, or multi memory and logic dies in a stack.

HBM stacks supposedly also have a basic logic die in each stack, probably a microcontroller or something.

Chip stack z height is also dependent on the thickness of the independent dies and their interconnects.

Samsung's VNAND has not just advanced state of the art in layer quantity, it has also pushed thinner and thinner layers for each level of flash in the device, ensuring that each successive jump in layer quantity does not significantly increase full die height - which is important as their enterprise SSD's stack up to 32 of them at once.

 

[name99]

Assuming the designs and core numbers of the A14 series chips follow those of A12, A12X, and A13:

I’m kinda surprised so many people think Apple is going to equip the MacBook / MacBook Air with an SoC that will be as fast as the one in the 2021 iPad Pro (A14X 4+4). There isn’t really a need for it in terms of performance. Remember, even A14 (non-X 2+4) performance will probably be competitive with A12X. A14 will likely be much faster than A12X single-core, and roughly as fast as A12X multi-core. IMO, their biggest hurdle here is not performance. Their biggest hurdle is marketing, but Apple has lots of experience with that as we all know.

Also, 8+4 is overkill for lower end iMacs and lower end MacBook Pros. 4+4 would be sufficient for those. Even if they wanted more than 4+4 in lower end MBPs and iMacs, Apple could release a 6+4 variant, limiting 8+4 to the highest end MacBook Pro and iMac models.

Apple has been forced to cripple Macs for years because of Intel's choices. Why are you so certain they will continue that crippling once Intel is out of the picture?

The story of the iPhone (or iPad or aWatch) since the day Apple took control ofd the silicon is that they keep ramping up performance because they can use that to keep adding functionality. To assume that they don't have similar plans for additional functionality that can be added to the mac seems short-sighted.
The "appropriate" roles and tasks for a PC, as defined by Intel and Microsoft (who last showed imagination some time in the 90s), are not accepted by Apple.

 

[Eug]

Interesting. Metal developer tools are now available for Windows.

Metal Overview - Apple Developer

Metal powers hardware-accelerated graphics on Apple platforms by providing a low-overhead API, rich shading language, tight integration between graphics and compute, and an unparalleled suite of GPU profiling and debugging tools.

developer.apple.com

What’s New in Metal
Tools for Windows
The Metal Developer Tools for Windows enables Metal Shading Language (MSL) compilation on Windows into Metal Library Objects targeting Apple platforms.
Download now

Apple has been forced to cripple Macs for years because of Intel's choices. Why are you so certain they will continue that crippling once Intel is out of the picture?

The story of the iPhone (or iPad or aWatch) since the day Apple took control ofd the silicon is that they keep ramping up performance because they can use that to keep adding functionality. To assume that they don't have similar plans for additional functionality that can be added to the mac seems short-sighted.
The "appropriate" roles and tasks for a PC, as defined by Intel and Microsoft (who last showed imagination some time in the 90s), are not accepted by Apple.

I don't think I've ever met anyone in person who has bought a MacBook Air and has complained that the CPU was too slow for their needs when they first bought it. Surely they exist, but they're probably a lot fewer than many tech nerds assume. Even over time, CPU speed usually isn't the first issue of concern for the vast majority of that target market. Complaints about speed have usually been due to storage speed (hard drives) and RAM. OTOH, I have met a number people who have complained about their cooling fans. So, in that context, I would expect Apple would rather prioritize thermals and battery life than raw CPU performance, in machines aimed at the lower end, for students taking notes for classes and executives working in Word and mom storing recipes, etc.

I will admit that I thought the first fanless MacBook in 2015 was a tad slow, but I didn't think the same thing with the 2017 model. I still use it and have no desire to upgrade it. However, even the A14 SoC with 2 performance cores would blow that 2017 MacBook out of the water in terms of raw CPU performance. Seriously, it's no contest. Furthermore, A14 would likely represent a decent incremental speed bump over the current fastest Intel chip in the MacBook Air, the Core i7-1060NG7.

For people wanting more speed, they can buy MacBook Pros with 4-6 performance cores. And for those wanting even more speed, they can buy the top tier models with 6-8 performance cores. And again, thermal characteristics matter here, especially with the form factors Apple likes to release.

This would be perfect product differentiation. Light and fanless? Go with MacBook Air with A14. What more power and more ports? Go with MacBook Pro with more cores.

 

[Glo.]

https://twitter.com/x/status/1281994827345911811

Apple

-considering- selling their ARM chips

to laptop/desktop manufacturers

im hearing

Ahem.

If this turns out to be true it is a complete game changer from top, to bottom.

 

[Doug S]

I don't think I've ever met anyone in person who has bought a MacBook Air and has complained that the CPU was too slow for their needs when they first bought it.

No one bought an iPhone Xr/Xs and complained the CPU was too slow, but they made it faster for the 11. Why are they bothering with redoing the iPhone SoC every year when they could do it every other year and no one is going to complain? Heck they even used the latest n greatest A13 in the new SE, when surely they could have saved a few bucks by using a couple year old A11 made on an older, cheaper 10nm process. Why not use the SoC the iPhone 8 came with if you are going to make the new SE look like the iPhone 8?

Apple doesn't have to make the Air fast, but do you really think they are going to intentionally slow down or limit it other than as may be necessary if they want it to operate without a fan vs higher end laptops that have fans? There are plenty of other ways to segment between the Air, the Macbook non-Pro, and the Macbook Pro without requiring huge steps in performance between them.

If they continue to solder RAM & NAND on the laptops they will have to offer a few different configurations since there's no one size fits all. But that makes sense, because it costs them less if there is less RAM and flash, so they can justify charging less. Just like it costs them less to buy a Core i3 versus a Core i7. However, when they drop Intel and all SoCs used in the Macbook Air cost them the same, maybe they don't segment that way on CPU/GPU and all Macbook Airs have the same performance?

The only possible cost advantage to segmenting on the SoC within a model would be as a way to use dies with a failed cores or allow clock binning, but its unclear whether they'll feel that's worth it. One could argue the same thing with the iPhone - why not clock bin and give the iPhone Pro faster clocks or use the slower ones in the new SE? Since they haven't - and that's a lot higher volume product where they need extra features to justify the higher price- I don't think we should assume they will do it on the Mac. Maybe they will, maybe they won't. Other than the Mac Pro, I think they are certain to do it there, because the difference between e.g. 16 and 32 cores is pretty significant and the customers would be willing to pay a lot more to double their performance.

I imagine there also might be some motivation for the hardware guys to want to really go crazy on the Mac Pro performance on the high end, just to show up all the naysayers who are claiming Apple won't be able to compete with x86 performance on that scale.

 

[Doug S]

https://twitter.com/x/status/1281994827345911811

Ahem.

If this turns out to be true it is a complete game changer from top, to bottom.

Some random nobody without any track record of inside information on Apple isn't going to leak the biggest scoop about the whole ARM transition.

And Apple learned their lesson about competing with themselves during the dark days of the 90s when they licensed the Mac, I'd be more likely to believe someone saying that Apple was buying GF and going to fab their own chips than to believe this nonsense.

 

[name99]

I don't think I've ever met anyone in person who has bought a MacBook Air and has complained that the CPU was too slow for their needs when they first bought it. Surely they exist, but they're probably a lot fewer than many tech nerds assume. Even over time, CPU speed usually isn't the first issue of concern for the vast majority of that target market. Complaints about speed have usually been due to storage speed (hard drives) and RAM. OTOH, I have met a number people who have complained about their cooling fans. So, in that context, I would expect Apple would rather prioritize thermals and battery life than raw CPU performance, in machines aimed at the lower end, for students taking notes for classes and executives working in Word and mom storing recipes, etc.

I will admit that I thought the first fanless MacBook in 2015 was a tad slow, but I didn't think the same thing with the 2017 model. I still use it and have no desire to upgrade it. However, even the A14 SoC with 2 performance cores would blow that 2017 MacBook out of the water in terms of raw CPU performance. Seriously, it's no contest. Furthermore, A14 would likely represent a decent incremental speed bump over the current fastest Intel chip in the MacBook Air, the Core i7-1060NG7.

For people wanting more speed, they can buy MacBook Pros with 4-6 performance cores. And for those wanting even more speed, they can buy the top tier models with 6-8 performance cores. And again, thermal characteristics matter here, especially with the form factors Apple likes to release.

This would be perfect product differentiation. Light and fanless? Go with MacBook Air with A14. What more power and more ports? Go with MacBook Pro with more cores.

An iPad Pro doesn't have a fan. If your argument is "MBA won't have an A14X with 4+4 cores because that would force it to have a fan" the argument makes little sense to me.
The issue of fan-ness is somewhat orthogonal to the precise SoC used (as I've mentioned the iPad Pro with an A10X doesn't have a fan; the aTV does).

My guess would be that in terms of user perception, a user would expect that a low end MacBook should basically match an iPad Pro and Apple will cater to that. They won't match along the stuff that costs, and that Apple can't control (mainly iPad Pro will have better screen) but for stuff that Apple can control like core count, why not? It's no big deal to throttle the big core frequencies back if the device does start to overheat while using all of them.

Now could Apple decide that they want a dramatically segmented mac portable line, somewhat like the iPad line, so
MacBook cheap
MacBook mid
MacBook pro
and that's achieved (somewhat like iPads) by using (essentially) iPhone SoC for the first, iPad SoC for the second, Mac SoC for the third?

We could make arguments like that but we hit the problem of IO. Are you going to put a USB controller or two on the iPhone SoC and make every iPhone pay that price? (Maybe, if iPhone moves to USB-C? Maybe if Apple is ready with USB4/TB4 and introduces that across the entire line going forward?)
How low is the target price point? MacBooks competing with ChromeBooks?

I don't think it's an insane vision, but it's one that's difficult to weigh in probability. EVEN SO, I still think that will be a lower end device, the mythical MacBook Vacuum. The midrange MacBook (say at ~$1000) I expect to have the iPad Pro SoC with 4+4.

 

[DrMrLordX]

If this turns out to be true it is a complete game changer from top, to bottom.

Depends on whether they'll require OEMs to buy the software along with the hardware.

 

[Eug]

No one bought an iPhone Xr/Xs and complained the CPU was too slow, but they made it faster for the 11. Why are they bothering with redoing the iPhone SoC every year when they could do it every other year and no one is going to complain? Heck they even used the latest n greatest A13 in the new SE, when surely they could have saved a few bucks by using a couple year old A11 made on an older, cheaper 10nm process. Why not use the SoC the iPhone 8 came with if you are going to make the new SE look like the iPhone 8?

Apple doesn't have to make the Air fast, but do you really think they are going to intentionally slow down or limit it other than as may be necessary if they want it to operate without a fan vs higher end laptops that have fans? There are plenty of other ways to segment between the Air, the Macbook non-Pro, and the Macbook Pro without requiring huge steps in performance between them.

I agree that there should be incremental performance increases each year, but that's just it: A14 non-X will already be an incremental performance increase over their current MacBook Air CPUs.

An iPad Pro doesn't have a fan. If your argument is "MBA won't have an A14X with 4+4 cores because that would force it to have a fan" the argument makes little sense to me.
The issue of fan-ness is somewhat orthogonal to the precise SoC used (as I've mentioned the iPad Pro with an A10X doesn't have a fan; the aTV does).

I was thinking they could run A14 at higher clock speed in the MacBook Air if they wanted. Yes, clock binning.

My guess would be that in terms of user perception, a user would expect that a low end MacBook should basically match an iPad Pro and Apple will cater to that.

Well, that isn't the case now. Even the 2018 iPad Pro is much, much faster than the 2020 MacBook Air.

It would be a pleasant surprise though if they started MacBook Air performance at 2021 iPad Pro levels.

We could make arguments like that but we hit the problem of IO. Are you going to put a USB controller or two on the iPhone SoC and make every iPhone pay that price? (Maybe, if iPhone moves to USB-C? Maybe if Apple is ready with USB4/TB4 and introduces that across the entire line going forward?)

The iPhone already supports USB. You just need a dongle. As of iOS 13 you can connect third party USB devices to iPhones. (Technically you could before, but it was really just limited to cameras and stuff for transferring images.)

And the Lightning cables connect to USB anyway.

The main problem is power. The iPhones don't supply enough power for most external storage devices, but they work with the same storage devices just fine if you supply external power. I know, because I've already tried it: USB thumb drives do work with my iPhone... if you supply external power. And stuff like USB keyboards and mice work fine too, even without external power.

The advantage of the USB-C endowed iPad Pro is that it doesn't require external power for these same devices.

 

[name99]

I agree that there should be incremental performance increases each year, but that's just it: A14 non-X will already be an incremental performance increase over their current MacBook Air CPUs.


I was thinking they could run A14 at higher clock speed in the MacBook Air if they wanted. Yes, clock binning.


Well, that isn't the case now. Even the 2018 iPad Pro is much, much faster than the 2020 MacBook Air.

It would be a pleasant surprise though if they started MacBook Air performance at 2021 iPad Pro levels.


The iPhone already supports USB. You just need a dongle. As of iOS 13 you can connect third party USB devices to iPhones. (Technically you could before, but it was really just limited to cameras and stuff for transferring images.)

And the Lightning cables connect to USB anyway.

The main problem is power. The iPhones don't supply enough power for most external storage devices, but they work with the same storage devices just fine if you supply external power. I know, because I've already tried it: USB thumb drives do work with my iPhone... if you supply external power. And stuff like USB keyboards and mice work fine too, even without external power.

The advantage of the USB-C endowed iPad Pro is that it doesn't require external power for these same devices.

"Well, that isn't the case now. Even the 2018 iPad Pro is much, much faster than the 2020 MacBook Air."
The issue is one of PERCEPTION...

Re USB: I was being somewhat flip because I misunderstood the audience.

(a) Do you know how many USB controllers there are in say a MacBook? It's something astonishing like 12 or so, used for all manner of connection that you wouldn't think of, like BT and WiFi.

Now does the iPhone do the same thing? Appear to have a single "port" (which can speak USB2 -- but not as far as I know USB3) but in fact it has 6 other USB controllers in there to connect to WiFi, BT, some debugging interface, and god knows what else? I honestly don't know. (At least USB used to be used to connect to stuff like that back 20 years ago, when I last looked at this in detail. Today?)

(b) As you say, the USB "connection" in the iPhone/iPad doesn't have to deal with power delivery, which I expect requires somewhat physically larger transistors. (But I honestly don't know.)

All these sorts of things kinda blur two issues together. In terms of "Can Apple do x, y ,z" the answer is surely yes because Apple is doing some variant of it already. But a different question is "Can Apple do x, y, z to extent expected of a Mac using the silicon appropriate for a phone" and that's a rather different question.
Now it's not "does Apple know how to make a USB controller?" it's "does it make sense in terms of area and tradeoffs to put as much USB firepower (number of controllers, and power delivery) on a phone as in a mac".
My guess would be no, but I will be the first to admit that there are so many variables here that I have no idea what the tradeoffs are.

For example, is connecting to things like BT and WiFi via USB a sensible choice? Is it inertia in PC-land, that Apple will ditch for something more efficient now that they control the SoC? Can they persuade their WiFi and BT partners to go along? (Hell, does Apple Silicon encompass BT and WiFi yet? Of course we know they have low-level versions of this, as evidenced by Apple Watch and Airpods, but good enough for the phone and mac?) So many unknowns!

 

[Eug]

The iPad Pro 2018 does USB 3 with proper power delivery. Flash drives work fine.

The iPad Pro 2017 does USB 3 with limited power delivery. Flash drives require external power.

The iPad does USB 2 with limited power delivery. Flash drives require external power.

The iPhones do USB 2 with limited power delivery. Flash drives require external power.

Apple uses Broadcom (and others) for BT & WiFi. Cellular will be Qualcomm in 2020.

 

[Doug S]

Apple uses Broadcom (and others) for BT & WiFi.

Whether Apple uses Broadcom chips for wifi or BT, someone else's, or designs their own is orthogonal to how they are connected to the SoC. It would be connected via some sort of serial protocol to minimize the number of I/Os and complexity. The question is whether there is a reason to 'roll your own' or use USB or at least some cut down version of it. USB is clearly fast enough, but there may be reasons (such as lower voltage I/Os) to prefer something else...I really don't know.

Perhaps the datasheet for the specific wifi or BT chip Apple is using in one of the last few iPhone generations might say something, and we can probably find out exactly which chip from the teardowns that iFixit does, or the sort of firms that also generate BOMs with cost estimates and provide more detailed information (perhaps down to the level of knowing whether they are connected via USB or something else) if you pay them thousands of dollars.

 

[Eug]

If it helps, I looked it up and Apple uses the Cypress CYPD2104 USB-C/3.1 controller in the iPhone.

https://www.cypress.com/part/cypd2104-20fnxit

https://www.cypress.com/file/140151/download

 

[soresu]

Apple uses Broadcom (and others) for BT & WiFi. Cellular will be Qualcomm in 2020.

Wait, what??!!

I thought they had been doing their own BT/WiFi for years, and their own modem for a bit too.

Kind of amusing that they are addressing CPU and GPU vertical integration for desktop before they have such comm basics integrated into mobile SoC's.....

 

[defferoo]

Wait, what??!!

I thought they had been doing their own BT/WiFi for years, and their own modem for a bit too.

Kind of amusing that they are addressing CPU and GPU vertical integration for desktop before they have such comm basics integrated into mobile SoC's.....

i think the reason is differentiation. controlling wifi, bt, cellular chips don’t allow them to differentiate. those are table stakes, every phone has it and they’re all about the same. CPU and GPU + other dedicated hardware like NPU is where the special sauce lives.

 

[name99]

The iPad Pro 2018 does USB 3 with proper power delivery. Flash drives work fine.

The iPad Pro 2017 does USB 3 with limited power delivery. Flash drives require external power.

The iPad does USB 2 with limited power delivery. Flash drives require external power.

The iPhones do USB 2 with limited power delivery. Flash drives require external power.

Apple uses Broadcom (and others) for BT & WiFi. Cellular will be Qualcomm in 2020.

Eug, you keep seeing everything I write as some sort of competition, rather than a collaboration. It's fine for you to clarify what I said if you think others in the conversation are confused; but it becomes tiresome when the clarifications are written as corrections, or are simply wrong.

For example Apple has its own WiFi and BT right now. Apple W1, W2, W3 provide Bluetooth and a somewhat limited form of WiFi. (2.4GHz, 1x1 only, only up to 802.11n).
So, yes, Apple's *known* WiFi offerings are not good enough for an iPhone right now, let alone an iPad Pro or a desktop. BUT the point, which should have been obvious to "one skilled in the art", is the question of whether Apple plans to keep it that way.

Given that Apple plan to design their own cellular modem, designing the other RF parts doesn't seem a particularly crazy assumption. BT is easiest, then WiFi, then Cellular, so it would make sense to work on a full-fledged WiFi-6/e chip, and have it ready, before cellular is ready.
Will those designs ship this this year in the new iPhones, iPads, and ARM Macs? Well, who knows?

==================

OK, to add to this for soresu.
The W parts provide the BT/WiFi for Apple Watch. So far Watch 3, 4, 5 jas all used W3.
The first round of AirPods used the W1, subsequent Airpods have used the H1 which (along with improving the BT part of the design, apparently drops the WiFi part as irrelevant to AirPods functionality).

Why haven't Apple been more aggressive about creating their own high-end WiFi chip? My guess is that everything is priorities. As long as Broadcom was reasonable in accommodating Apple's requests, whatever they were, Apple has plenty of work dealing with all the other silicon they've been creating and have lined up.
My personal theory (with zero evidence to support it!) is that once they were committed to their own cellular modem, the WiFi is a reasonable goal along the way. You have to create the same sort of primitives (Viterbi and ECC, Fourier, MIMO, now with one of the WiFi6 modes even the same sort of frequency/time blocks). You can use WiFi as a slightly less demanding learning project, with a bonus that at the end you have one more piece of silicon you can control. (The Project Zero write-up on an exploit that used the Broadcom driver is terrifying in what it reveals about Broadcom software practices...)

 

[Eug]

Wait, what??!!

I thought they had been doing their own BT/WiFi for years, and their own modem for a bit too.

Kind of amusing that they are addressing CPU and GPU vertical integration for desktop before they have such comm basics integrated into mobile SoC's.....

i think the reason is differentiation. controlling wifi, bt, cellular chips don’t allow them to differentiate. those are table stakes, every phone has it and they’re all about the same. CPU and GPU + other dedicated hardware like NPU is where the special sauce lives.

Probably has to do with patents. My understanding is that is why they bought Intel’s modem business, for the intellectual property.

Similarly, that is why they signed a contract again with Imagination Technologies after all, after a big public breakup.

BTW, the story about the modem is that they were counting on Intel for 5G in 2020 but Intel failed to deliver. So all of a sudden Apple kissed and made up with Qualcomm and signed a multi-billion dollar deal to put Qualcomm 5G chips into the iPhone and iPad Pro. Apple at the same time bought out Intel’s modem business, with the hope of creating their own modem silicon after a few years, but probably not anytime soon.

 

[Doug S]

Wait, what??!!

I thought they had been doing their own BT/WiFi for years, and their own modem for a bit too.

Kind of amusing that they are addressing CPU and GPU vertical integration for desktop before they have such comm basics integrated into mobile SoC's.....

Your statement makes no sense. You think they should wait on going ARM on the desktop until they have integrated BT/wifi on mobile? Should they start making their own DRAM and flash too, and OLED display panels, and hey while they're at it mine and refine their own aluminum as well? How vertical do they need to be on phones before you think they are ready to proceed onto the desktop?

Maybe you aren't able to do two things at once, but companies the size of Apple are.

 

[DrMrLordX]

BTW, the story about the modem is that they were counting on Intel for 5G in 2020 but Intel failed to deliver. So all of a sudden Apple kissed and made up with Qualcomm and signed a multi-billion dollar deal to put Qualcomm 5G chips into the iPhone and iPad Pro. Apple at the same time bought out Intel’s modem business, with the hope of creating their own modem silicon after a few years, but probably not anytime soon.

Too right. Was wondering when someone would bring that up, since it was kind of a big deal when it all went down.

 

[soresu]

Your statement makes no sense. You think they should wait on going ARM on the desktop until they have integrated BT/wifi on mobile? Should they start making their own DRAM and flash too, and OLED display panels

Last time I checked none of those are integrated into the main SoC - plenty of SoC's have BT/WiFi integrated, including Exynos.

As for them making those things, well why not?

Apple had a stake in a display panel manufacturer a while ago, so it's technically true already to an extent.

Either way it wasn't meant to make sense - I said it was amusing, I wasn't going for a Pulitzer on vertical integration strategy.

 

[soresu]

Maybe you aren't able to do two things at once, but companies the size of Apple are.

Well, first off true human multi tasking is a fallacy - we actually 'multi task' by switching between serial tasks quickly and some basic application of muscle memory.

Secondly. the average company the size of a large hotel can potentially do dozens of things at once - that's because they have dozens of employees.

So you are not exactly blowing anyones mind by claiming* that a huge, trillion dollar valued company the size of Apple, with tens of thousands of employees can do 2 things at once.

Point of fact I do think that it is hilarious a company "the size of Apple" can't stay ahead of the competition themselves on basic comms IP at this point in the game, or buy it for themselves if necessary - it's not like Broadcom wasn't actually bought out in the last decade either.

Between that and the fact that they never bit the bullet and acquired IMG Tec, I'm inclined to think that they are squeamish about acquisitions for some reason - they often do make joint ventures but very seldom actually acquire big companies (not that IMG Tec was particularly big at any point mind you).

*albeit incorrectly in this case as ARM Mac will basically mean Apple reusing a lot of work across mobile to desktop/laptop divisions, not to mention that WiFi/BT is likely to be in their desktop/laptop systems too as it can be for desktop PC's if you buy the right motherboard.

 

[Doug S]

Well it is obvious that Apple could have been doing their own BT/wifi for years now had they made it a priority like they made CPU a priority. But you are assuming that just because something CAN be done that it should be done, and it reflects a failing on Apple's part to have not done so.

Consider why vertical integration is attractive to Apple:

1) they can set their targets for features and performance, not be stuck with what others provide
2) they can save money by not paying someone else's profit margin

For BT/wifi they CAN'T do #1, because as standard protocols they can't say "we're doing wifi 7 now" and get ahead of everyone else. All they can do is do what everyone else is doing. They can beat everyone else on the CPU, they can't for standards-based protocols like BT/wifi.

They can save money, but they've never had a problem with Broadcom. Unlike Qualcomm they don't charge licensing fees based on a percentage of the sales price of the whole phone, so Apple would save a lot of money doing their own cellular (but like BT/wifi it is standards based so they can only do as well as others, they can't get ahead) they would save little money doing their own BT/wifi since Broadcom isn't making near the profit on that that Qualcomm does on cellular. Thus BT/wifi hasn't been a priority for them for either #1 or #2.

Yes they are doing their own for the watch, and maybe that indicates they will do it someday for phones/tablets/Macs. Or maybe not, if the reason they did it for watches and Airpods is because they wanted to use as little power as possible (as accomplished that by using a cut down feature set) or because the additional integration was necessary in the VERY space constrained environments of a watch and an earphone, while the iPhone is by comparison roomy and can afford the space required for a separate chip.