A14X Chip for First Apple Silicon Mac and New iPad Pro to Enter Mass Production in Fourth Quarter

[demodave]

Does anyone know (On average) how many (usable) chips that that are likely to get per wafer?

5 nm is on 300 mm ("12 inch") wafers, so there is a hell of a lot of surface area available, and I've found an Anandtech report showing 2600 good die on a test wafer, with 600 failed die and some number of edge die (die at the edge fo the wafer are generally discarded as a quality protection against field returns.

I would assume that TSMC would not release a die to production with less than 90-955 yield, so, being teh lazt mathematics engineer that I am, I would expect 3000 good die per wafer.

www anandtech com /show/15219/early-tsmc-5nm-test-chip-yields-80-hvm-coming-in-h1-2020

This needs a major caveat: the test die that was referenced by Anandtech could be either smaller or larger than an Apple-designed die, so the good die per wafer yield is still very much up in the air. WHere I work, and given our technology node, we can get 25,000 die printed on a 15 mm ("6 inch") wafer. For our more modern devices, it's more like 25,000 on a 200 mm ("8 inch") wafer. But our die are relatively small, so I would't expect that an Apple Axxx would fit that many die, even on a 300 mm wafer.

(Yes, I am a semiconductor manufacturing engineer, so my comments on edge die and yield expectations are generally valid, but I really don't have any secret insight into TSMC's process for Apple Axxx ships. So my insight is not worth much more than Livverpool's "Depends, we talking chocolate, vanilla or strawberry wafers? ??".)

 

[cmaier]

5 nm is on 300 mm ("12 inch") wafers, so there is a hell of a lot of surface area available, and I've found an Anandtech report showing 2600 good die on a test wafer, with 600 failed die and some number of edge die (die at the edge fo the wafer are generally discarded as a quality protection against field returns.

I would assume that TSMC would not release a die to production with less than 90-955 yield, so, being teh lazt mathematics engineer that I am, I would expect 3000 good die per wafer.

www anandtech com /show/15219/early-tsmc-5nm-test-chip-yields-80-hvm-coming-in-h1-2020

This needs a major caveat: the test die that was referenced by Anandtech could be either smaller or larger than an Apple-designed die, so the good die per wafer yield is still very much up in the air. WHere I work, and given our technology node, we can get 25,000 die printed on a 15 mm ("6 inch") wafer. For our more modern devices, it's more like 25,000 on a 200 mm ("8 inch") wafer. But our die are relatively small, so I would't expect that an Apple Axxx would fit that many die, even on a 300 mm wafer.

(Yes, I am a semiconductor manufacturing engineer, so my comments on edge die and yield expectations are generally valid, but I really don't have any secret insight into TSMC's process for Apple Axxx ships. So my insight is not worth much more than Livverpool's "Depends, we talking chocolate, vanilla or strawberry wafers? ??".)

I agree.

 

[Makosuke]

Apple's first Mac processors will have 12 cores, including eight high-performance cores and at least four energy-efficient cores, according to Bloomberg. Apple is said to be exploring Mac processors with more than 12 cores for further in the future, with the company already designing a second generation of Mac processors based on the A15 chip.

If this prediction is accurate and the per-core performance of the A14 is even close to the A13, this chip will be extremely fast.

Looking at Geekbench 5 scores--which is great because it's the exact same benchmark on both Mac and iOS, we already can say confidently that if you put an A12Z into a 13" MacBook Pro today, it would perform roughly on par with the current top-of-the-line i7 13". So a 4-high-performance-core A14X would make for a nice top-of-the-line 13"-ish MBP.

But 8 high-performance cores would change the equation entirely. The A13 scores about 1325 in single-core performance and about 2.5 times that in multi-core with its 2 high-performance cores and 4 efficient ones. The A12Z, with four of each, has a multi-core score about 4.1 times its single-core score.

The MacOS has the same architecture as iOS, so this would point to it being relatively reasonable to assume that the multi-core performance will be at least the single-core times the number of high-performance cores. Maybe less once you get to even higher numbers of high-vs-low performance cores, but it's probably at least in the ballpark.

Which is to say that an 8-high-performance-core version of the A13, with the same clock speed and no modifications other than more cores, could be reasonably expected to have single-core score of 1325 and mutli-core somewhere in the ballpark of 10,600.

In terms of mid-2020 Macs, that would mean it would have about 6% better single-core performance than any Mac available (the fastest is the 3.6GHz Core i9 in the iMac), and multi-core about 27% faster than that same 8-core i9 iMac, Apple's fastest current non-pro desktop. Basically, it would be a solid upgrade in the top-of-the-line 27" iMac but (apparently) in the thermal envelope of a 13" laptop.

Comparing this hypothetical to the current top-of-line Mac laptop, the 8-core i9 MBP 15", it would have 18% better single-core performance and about 52% better multi-core performance.

All this is if there are no clock speed bumps, and the per-core performance is identical to the A13. If there is even a modest 10% improvement in per-core performance over the A13, this A14x would be expected to slightly outperform a 10-core Intel Core i9-10900K desktop CPU in both single and multi core performance, which currently has the best single-core performance of any Intel or AMD CPU available and I believe is the lower end of Intel's high-end desktop line. It also has a TDP of 125W.

No guarantees that all this will actually turn out that way, but if Apple can pull off this chip the way they've been doing the A-series so far, it's going to be a heck of a thing.

 

[MareLuce]

I’d love a lighter 12.9” iPad Pro. Does anyone believe Apple Silicon probably means everything will be lighter?

 

[chucker23n1]

I’d love a lighter 12.9” iPad Pro. Does anyone believe Apple Silicon probably means everything will be lighter?

The iPad Pro is already on Apple Silicon.

 

[Unregistered 4U]

Interesting that all Mac processors will have 12 cores. If they're planning multiple Macs, it will be interesting how they'll differentiate them.

For iOS, Apple makes a small number of processors and since every member of that family gets the same processor, they can spread costs over the entire line. So, it could be similarly to how they differentiate the current iOS products. All of the MacBook Air’s get the same processor (CPU/GPU/system ram), you choose how much storage, what ports (if there are options), annnnnnd the color

Higher-level products probably need a more flexible naming convention. Something like A14M.1 (13-inch Pro), .2 (16-inch Pro, low-end iMac, Mac mini), .3 (high-end iMac) through A14M.4 (Mac Pro), maybe.

The current state is due to the fact that Intel wants to make money across a wide range, so there are features unique to high end processors that aren’t available in lower end. OR, charging more for .7 increase in speed. Apple MAY replicate this in laptops, but to me, it makes more sense to do it the way Apple does it with their iOS/iPadOS devices. This way, even the BOTTOM of the range gets performance similar to the top of the range.

If there was no competition for iPhone, we would all be lining up to buy iPhone 5S this fall.

There is effectively no “competition” as far as speed in concerned. Yet, Apple has continued to drive themselves to the next level. Even now, nothing in the mobile space can touch Apple’s solution, yet they’re still driving year over year performance increases and feature additions.

 

[Laurent François]

Depends, we talking chocolate, vanilla or strawberry wafers? 🤔😂

Plain wafers, with cholate chips and whipped cream.
Thanks

 

[MareLuce]

The iPad Pro is already on Apple Silicon.

Thank you! I knew that! ;-). Darn... Probably with AR it won’t get lighter...

 

[fokmik]

@MacRumors if this is kind of serious, it should be posted on the main page as leaks/rumours

 

[bluecoast]

Well beating the cooling of the 2020 MBA will be easy, but that's not Intel's fault. Now I understand why Apple didn't even bother putting a heat pipe on the 2020 MBA.... so they can rave about how much cooler the AS MBA's run

Agreed. The reports about heat issues when buying a 2020 MBA with anything other than the entry level i3 were concerning.

 

[chucker23n1]

The current state is due to the fact that Intel wants to make money across a wide range, so there are features unique to high end processors that aren’t available in lower end. OR, charging more for .7 increase in speed. Apple MAY replicate this in laptops, but to me, it makes more sense to do it the way Apple does it with their iOS/iPadOS devices. This way, even the BOTTOM of the range gets performance similar to the top of the range.

Unless Apple wants to compress the Mac product line, they’ll want to approach this similarly. A Mac Pro whose CPU is basically the same speed as that of a MacBook Air is a poor value proposition, even if you ignore thermals. I don’t think core count alone will work.

 

[Unregistered 4U]

Perhaps a custom intel chip in a ipad with software specifically coded for it would be fast also?

No. Intel is making general purpose processors for a general audience. Apple Silicon will always have a benefit when running macOS as it’s designed specifically for the task. Plus, there’s no way Intel would ever actually release something with the performance of an i7 in a small power efficient package like an i3. They’d gut their i7 and i5 sales if they did.

Does Apple have more smart and clever people that can overcome these same issues both these other companies have struggled with for years and years?

The real issues to overcome aren’t at the chip level. The big ones are at the architectural level. When your platform (Wintel) is broad enough for many companies to build custom solutions on, there are inefficiencies as a part of the trade-off to get to wide acceptance. Vary from that and you may get better performance, but at the cost of being less compatible. Apple has an opportunity to design the entire system such that any corners that can be cut to execute macOS code can be made understanding that the hardware will 100% support the software that runs on it. Those two things working together will provide benefits much greater than an incremental increase in clock speed ever would.

Unless Apple wants to compress the Mac product line, they’ll want to approach this similarly. A Mac Pro whose CPU is basically the same speed as that of a MacBook Air is a poor value proposition, even if you ignore thermals. I don’t think core count alone will work.

Compressing the Mac product line compared to Intel’s offerings, yes. Instead of 4 different processors for the Air, just one (no need to offer a “cheaper” i3 type solution). For the MacBook Pro and iMac, instead of 6 to 10 options between them, just one. And then 1 for the potential iMacPro and MacPro.

 

[robinp]

There are major engineering challenges to multi chip that make it not worth their effort. I designed opteron which was designed to work that way. It was a necessity because we couldn’t put more cores on a die, but it is a pain and not worth it given the external bus speeds.

You're actually an expert here by the sounds of it which is a really unusual thing in these forums. Thanks for your input.

One thing I would say though, is that many tasks that require lots of parallel processing power work very well over networked clusters of machines where the CPUs are interconnected with much slower connections than those within a die or across a motherboard. Perhaps, for many of these workloads, it needn't be so difficult?

 

[theorist9]

If the first AS laptop they release is a 12"/13" MacBook, and if its CPU/GPU performance exceeds that of the i7/Iris combo offered in the contemporaneous top-of-the-line small (14"?) Intel MBP, then what happens to sales of both that machine (which could cost >2x the MacBook), and the Air?

 

[MandiMac]

Agreed. The reports about heat issues when buying a 2020 MBA with anything other than the entry level i3 were concerning.

Still the question is: If Intel had their $h!t together, we wouldn't have thermal problems either way. If Apple Silicon can do it, why not Intel?

 

[chucker23n1]

Compressing the Mac product line compared to Intel’s offerings, yes. Instead of 4 different processors for the Air, just one (no need to offer a “cheaper” i3 type solution). For the MacBook Pro and iMac, instead of 6 to 10 options between them, just one. And then 1 for the potential iMacPro and MacPro.

Maybe. But why? Just one CPU on the Mac Pro seems… needlessly limiting. Some people want the $2399 16-inch MacBook Pro, and some want to spend a little/a lot extra on more cores.

 

[MandiMac]

If the first AS laptop they release is a 12"/13" MacBook, and if its CPU/GPU performance exceeds that of the i7/Iris combo offered in the contemporaneous top-of-the-line small (14"?) Intel MBP, then what happens to sales of both that machine (which could cost >2x the MacBook), and the Air?

Apple cannibalizing its own sales is a regular phenomenon. So, people who need/want Intel and Bootcamp will get that one, and others will get the newest AS hardware.

 

[theorist9]

Apple cannibalizing its own sales is a regular phenomenon. So, people who need/want Intel and Bootcamp will get that one, and others will get the newest AS hardware.

That used to the be the case under Jobs, but it's been much less so under Cook (see, e.g., https://www.imd.org/research-knowle...ales-show-importance-of-self-cannibalization/).

Indeed, one of the arguments I've heard for why we'll never see the elusive Mac tower* (aka "xMac") is that it would cannabilize sales from both the iMac and the Mac Pro. [High-end, e.g., Intel i9-10900K, but still consumer-grade.]

 

[MandiMac]

Indeed, one of the arguments I've heard for why we'll never see the elusive mid-range Mac tower (aka "xMac") is that it would cannabilize sales from both the iMac and the Mac Pro.

Or: Mid-range is not the thing that Apple wants to do. You either get good products (iPhone SE), better products (11) or best (11 Pro). Same with iMac with good (21.5), better (27) or best (iMac Pro).

 

[LastStarfighter]

After trying and failing to get an iPad Pro to be my one single mobile computer, I've fallen back on a MacBook Pro 13".
If the rumoured 12" MacBook with Apple Processor has what we expect it to have (longer battery life than the existing MacBook Pro, hopefully built in cellular, better graphics performance) I'm going to be in the queue for one of these as soon as possible. If they make it a 2-in-1 I'll be even more happy with it.

 

[MandiMac]

After trying and failing to get an iPad Pro to be my one single mobile computer, I've fallen back on a MacBook Pro 13".
If the rumoured 12" MacBook with Apple Processor has what we expect it to have (longer battery life than the existing MacBook Pro, hopefully built in cellular, better graphics performance) I'm going to be in the queue for one of these as soon as possible. If they make it a 2-in-1 I'll be even more happy with it.

I'm interested: What was the roadblock about iPad Pro? I have a 12.9 inch (2017) as main computer for 3 years now.

 

[theorist9]

Or: Mid-range is not the thing that Apple wants to do. You either get good products (iPhone SE), better products (11) or best (11 Pro). Same with iMac with good (21.5), better (27) or best (iMac Pro).

Figured there was a chance you might be confused by that, so I was in the process of editing it. What I meant by mid-range wasn't middle-end, it was mid-range within the (limited) lineup of Macs that come without displays, i.e., higher than the mini, but lower than the Mac Pro. The headless Mac would offer, e.g., the Intel i9-10900K, which is a higher-end processor than the i9-10900 in the iMac. I.e., it would be Apple's version of a fast desktop PC.

 

[MandiMac]

Figured there was a chance you might be confused by that, so I was in the process of editing it. What I meant by mid-range wasn't middle-end, it was mid-range within the Mac lineup, i.e., higher than the mini, but lower than the Mac Pro. So the headless Mac would offer, e.g., the Intel i9-10900K, which is a higher-end processor than the i9-10900 in the iMac. I.e., it would be Apple's version of a fast desktop PC.

Gotcha.

But still: You have a Mac mini, you have an iMac, and you have a Mac Pro. Not quite sure there fits a headless Mac in there that can do things neither the best Mac mini nor the base Mac Pro can.

 

[theorist9]

Gotcha.

But still: You have a Mac mini, you have an iMac, and you have a Mac Pro. Not quite sure there fits a headless Mac in there that can do things neither the best Mac mini nor the base Mac Pro can.

That's not the idea behind the headless Mac. It's to do things the (thermally limited) Mac mini can't, without having to shell out the $6k needed for a base Mac Pro (which, in its base configuration, is a poor value).

Apple could probably produce an i9-10900K headless Mac, at its current profit margins, for ~$3k. That would be a beast of a machine for people that need power but don't need, or have the budget for, Xeons. That's the basis of the interest in such a machine. It would be of interest to hobbyists, graphics/video pros that are independents and don't have the budget for a Mac Pro, scientists wanting to supply their staff with Macs for development work, etc. etc.

I should add the other characteristics of the desired headless Mac that aren't available in the Mac Mini, iMac, or iMac Pro are these standard tower features:
1) Full modularity, with replaceable, and thus upgradeable, RAM, SSD's, GPUs, and CPUs (like in the G5).
2) Slots to accommodate two GPU's, and two or more storage drives.
3) A significantly expanded thermal envelope, enabling the machine to run on all cores without significant thermal constraint, and also (see no. 2), to accommodate up to two powerful GPUs
[No. 3 *might* be mooted by AS; we shall see.]

 

[MandiMac]

That's not the idea behind the headless Mac. It's to do things the (thermally limited) Mac mini can't, without having to shell out the $6k needed for a base Mac Pro (which, in its base configuration, is a poor value).

Apple could probably produce an i9-10900K headless Mac, at its current profit margins, for ~$3k. That would be a beast of a machine for people that need power but don't need, or have the budget for, Xeons. That's the basis of the interest in such a machine. It would be of interest to hobbyists, graphics/video pros that are independents and don't have the budget for a Mac Pro, scientists wanting to supply their staff with Macs for development work, etc. etc.

I get what you're saying, but I think Apple believes that the iMac is for them. It's exactly the case for better processors and GPU than the Mac mini for your budget, but it happens to have a state-of-the-art display with it in 5K resolution.