TSMC starts production of A15 for iPhone 13. Macbook Pros will probably get SoCs based on A15.

[senttoschool]

Everything is lining up for the upcoming Macbook Pros to get chips based on the A15.

A15 is already in production for the iPhone 13 according to a new report. This means the A15 design was completed long ago already. Rumors are that the new Macbook Pros will have an 8/2 big little CPU configuration which suggests a new SoC design that is different from the M1.

Another poster suggested that if the Macbook Air 2021 gets an M2 SoC, it'll have faster single-thread performance than any M1-based 2021 Macbook Pros for as many as 8 months every cycle which is "embarrassing" for the Pro moniker. Thus, it's likely that the upcoming Macbook Pros will skip A14 cores and go straight to A15 to reset the cycle.

A final reason for Macbook Pros to get the newest SoC designs first over the Air/iPad Pro/iMac is because you can bin the top chips for Macbook Pros and then use the defective chips or lower clocked chips for lower-end devices. This is exactly what AMD/Nvidia/Intel do. The highest-end GPUs/CPUs will always get released first. Then the defective chips start to accumulate for midrange or low-end products with disabled cores and/or lower clock speeds.

 

[dugbug]

I'd argue that single core performance differences have been the norm for years. Xeons for example have always had slow single core performance.

That said I sure hope the new imac 30"/mbpros all get an "m2x"

 

[leman]

I'd argue that single core performance differences have been the norm for years. Xeons for example have always had slow single core performance.

Which is why Apple has been using special Xeons with higher turbo boost.

Balanced CPU performance was always something Apple refused to compromise upon. They continued using 28W CPUs where everyone moved to 15W, consistently shipped highest clocked available models and always used thermal throttling rather than power throttling (which is the norm) to squeeze all possible performance from the CPU. They were obviously rather unhappy with zingele roadmap in last 4-5 years that failed to deliver any significant efficiency improvements and instead relied on increased power consumption (either implicitly or explicitly) to deliver improved performance.

Bottom line is that now when Apple designs their own silicon, there is little doubt that they will focus on single-threaded performance as much as they can. And combined with their power efficiency (that’s 2-3 years at least ahead of anyone else), this also translates to excellent multi-threaded performance. I would be very surprised if Apples Pro Silicon was not at least 10% faster than fastest desktop CPU cores.

 

[Lemon Olive]

Nope. There is no evidence to support what you're saying.

By all accounts the chip going in to the next MBP is the same architecture as the M1 but with a different arrangement of cores, additional graphics cores, and more memory options.

The timing is also not evidence.

The M1X has been completed for months and went into mass production early last month. The A15 went into production for iPhone today.

And if there is an M2 processor at all in 2021, it will be based on A15, and will still be a low end chip made for the low end products. It won't magically outperform a 10 core version of the M1.

 

[leman]

By all accounts the chip going in to the next MBP is the same architecture as the M1 but with a different arrangement of cores, additional graphics cores, and more memory options.

What accounts? There are no direct accounts making concrete claims, and no direct evidence. All we have are rumors and circumstantial evidence. Based in the later, I believe Pro Apple Silicon to be based on a different microarchitecture than M1.

 

[senttoschool]

I'd argue that single core performance differences have been the norm for years. Xeons for example have always had slow single core performance.

That said I sure hope the new imac 30"/mbpros all get an "m2x"

New SoCs often get as much as 20% faster ST speed improvements. That's a huge difference for 8 months (assuming Airs get released in November and Pros in the Summer).

This isn't Intel's roadmap where new CPUs offer 1-5% improvement.

 

[CWallace]

Someone (Bloomberg?) was reporting a month or so ago that TSMC was starting production of the M2, which most of us assume will be based on the A15 (as the M1 was based on the A14).

Rumor mill says M2 will go into the next MacBook Air and 13.3" MacBook Pro with the 16" MacBook Pro and a new 14" model getting the "M1X" which will have more performance than an M2 thanks to more performance CPU cores (8 instead of 4) and more GPU cores (16 or 32 instead of 8).

M2 will likely have better single-core thanks to having the newer Avalanche core (as opposed to the M1X's Firestorm) and of course it will be more power-efficient with 4 efficiency cores vs. the 2 on the "M1X".

 

[nylon]

.....
Rumor mill says M2 will go into the next MacBook Air and 13.3" MacBook Pro with the 16" MacBook Pro and a new 14" model getting the "M1X" which will have more performance than an M2 thanks to more performance CPU cores (8 instead of 4) and more GPU cores (16 or 32 instead of 8).
.....

They aren't going to have 16", 14" and 13.3" MBP's. The 13.3" screen size will be scrapped for the MPB range.

 

[Lemon Olive]

What accounts? There are no direct accounts making concrete claims, and no direct evidence. All we have are rumors and circumstantial evidence. Based in the later, I believe Pro Apple Silicon to be based on a different microarchitecture than M1.

Those are the accounts. And they are very clear.

 

[leman]

Those are the accounts. And they are very clear.

Sorry, clear how? So far, we’ve heard rumors that the new machines will be based on M1X, on M2, and everything in between. The only really credible rumor (because it has multiple sources and has surfaced multiple times) is the name of the prosumer platform (Jade Die). We also can assume that it will have at least 8 high-performance cores and at least 16 GPU cores because anything less doesn’t make any sense. That’s about it. We have no information whatsoever about the microarchitecture or the features.

 

[CWallace]

They aren't going to have 16", 14" and 13.3" MBP's. The 13.3" screen size will be scrapped for the MPB range.

One would think - and I do hope - the 13.3" 2-port model would go away to be replaced by the 14" multi-port model as even if the 14" starts at $1799 with 16/512, there will be the MacBook (Air) to anchor "the low end" and they can add a fan to it to keep it cool and at $1449 with 16/512 it would be $350 cheaper than the MBP 14".

But Tim likes to keep models around to hit price points so I would not be surprised if Apple does keep the 13.3" around at $1299/1499.

 

[Jorbanead]

Someone (Bloomberg?) was reporting a month or so ago that TSMC was starting production of the M2, which most of us assume will be based on the A15

Nobody knows what it is called - Bloomberg AFAIK stated it could also be called the M1X or something else. We don’t know if this “M2” chip is just an M1X chip as people don’t know what to call it. M2 simply may have made for better headlines as it’s easier for people to understand.

A15 is already in production for the iPhone 13 according to a new report. This means the A15 design was completed long ago already.

Correct, but that’s how it’s been basically every year for the last decade and those processors got released in sept. (Minus 2020 for obvious reasons).

Right now, I don’t see any strong indication that the M1X (again who knows what it’ll be called) will use 5nm+. Honestly the only real evidence right now is that A15 started production, but if we look at the history of Apple silicon through the years this simply tracks with a typical year and doesn’t really show either way anything about the M-series.

Honestly just judging off of Apple silicon history, the X-variant of the A-series has always come after the iPhone chip. So there’s just as much (if not more) evidence to show that Apple would first announce iPhone chips, then low-tier Mac chips, then higher-tier chips.

As for binning, I think each class of chip will get binned: meaning M1X may have a few binned versions, but that wouldn’t filter down to an M2 or whatever it’s called. I think of it this way:

M1, M2, M3 - all low power, lower core count chips (2 binned versions)

M1X, M2X, M3X - higher power, higher core chips (maybe 2-3 binned versions)

So binning would take place within their class of chip. Sort of how xeons can be binned, but you don’t take a Xeon and then bin it to become an i7 or something. They’re in different classes.

 

[Krevnik]

But Tim likes to keep models around to hit price points so I would not be surprised if Apple does keep the 13.3" around at $1299/1499.

Apple can keep around the price point without keeping the 13.3” as-is, too.

A new design for the MBA that can scale up, or a low end 14” MBP can both slide into that spot at some point. I wouldn’t expect it to be replaced when the 14” is announced unless it is a price-cut 14” that can hit the 1299$ price point, but rather replaced down the road during a refresh when the low-end M2 chips are available.

Keep in mind that the M1 Mini and 13” M1 MBP provide important hardware options during the transition and the lack of a refreshed design to have them available on day 1 makes sense. Specifically, if I need a bunch of M1 build machines to release a large project out of a build lab, the M1 Mini is really my only option (even more if I depend on places like MacStadium to provide access to machines to build/test on), and without it, delivering ARM builds of larger projects is difficult and extremely painful. The 13” MBP serves a similar purpose of providing developer-class hardware for smaller developers where they need a laptop (or a few). Neither of these devices are going to be as important 18-24 months from now once the transition is complete, as they have been in the last 6 months to get stuff ported over and tested.

Correct, but that’s how it’s been basically every year for the last decade and those processors got released in sept. (Minus 2020 for obvious reasons).

But it’s also the case that these chips were primarily for the iPhone where launching in fall is part of timing things with the market. New era with a different platform means that we should be open to new schedules. Nothing says Apple has to keep this schedule at this point. So I don’t really think either approach is more likely than the other.

Honestly just judging off of Apple silicon history, the X-variant of the A-series has always come after the iPhone chip. So there’s just as much (if not more) evidence to show that Apple would first announce iPhone chips, then low-tier Mac chips, then higher-tier chips.

And an A14X never got made this time around (neither did an A13X for that matter, which in hindsight was a sign that Apple was spending energy on a bigger move).

Honestly, there’s also been a compelling bit of information on the forums that Apple won’t use the “X-variant” labeling with the Mac. And after having that discussion, I’m inclined to believe that it has merit. Especially since Apple is putting out core counts in the M1 spec sheets, which they didn’t do with the A series, and still don’t.

That said, there’s no reason Apple can’t jump straight to M2 with the higher end version. It does make sense that if they knew the timing of things well enough in advance, that they could jump to the new cores in time for a WWDC announcement with July availability. It wouldn’t completely surprise me that the memory and I/O controller work they had to do is built alongside Avalanche and the like, and we start seeing more “trickle down” from the higher end M2 to the low end M2. Especially if Apple goes the route that Bloomblerg outlined with Jade C-Die / 2C-Die and 4C-Die to scale up the die to Mac Pro levels, since that’d mean that we’d be looking at Jade C-Chop and Jade C-Die variants for everything but what currently has an M1. And the low-end M2 could be based on the Jade C-Chop for all we know.

 

[Lemon Olive]

Sorry, clear how? So far, we’ve heard rumors that the new machines will be based on M1X, on M2, and everything in between. The only really credible rumor (because it has multiple sources and has surfaced multiple times) is the name of the prosumer platform (Jade Die). We also can assume that it will have at least 8 high-performance cores and at least 16 GPU cores because anything less doesn’t make any sense. That’s about it. We have no information whatsoever about the microarchitecture or the features.

The rumors we have indicate an M1 variant for the MBP and an M2 with new core design for the MacBook Air.

 

[leman]

The rumors we have indicate an M1 variant for the MBP and an M2 with new core design for the MacBook Air.

Gurman’s report does not make any claims about the microarchitecture used, only the code names of the chips used. All you write here is just your interpretation with very little factual background.

 

[jeanlain]

The rumors we have indicate an M1 variant for the MBP

Which rumours? There were leaks about number of cores and code names. I don't remember any rumour from a reputable source mentioning an "M1 variant".

 

[Lemon Olive]

Which rumours?

All of them.

 

[Lemon Olive]

Gurman’s report does not make any claims about the microarchitecture used, only the code names of the chips used. All you write here is just your interpretation with very little factual background.

None of the changes mentioned by Bloomberg indicate a new chip architecture, and they all are consistent with the already reported information by Bloomberg, that M1 variants exist and were planned for Pro products in late 2021.

However the changes for the MacBook Air do in fact indicate a new chip architecture.

Also, common sense indicates that the M2 is not debuting ahead of the A15, nor it is debuting as a Pro product. This simply is not how Apple does things.

 

[Jorbanead]

That said, there’s no reason Apple can’t jump straight to M2 with the higher end version.

Fully agree - Apple could jump straight to an M2X or something, but I just don’t think there’s any real indication of this happening based off of what we know. I’m not saying it wouldn’t happen, but that it’s just as likely that we don’t get a 5nm+ chip in two weeks and instead it’s just a high-core count M1

 

[jeanlain]

None of the changes mentioned by Bloomberg indicate a new chip architecture, and they all are consistent with the already reported information by Bloomberg, that M1 variants exist and were planned for Pro products in late 2021.

No one said that new new chips will use a new microarchitecture, we just said that there is no evidence that they will use the same as the M1.
The latest Bloomberg report doesn't say anything about it. I don't remember that previous reports did.

 

[quarkysg]

Now that we have all the various M1 Macs released together with the M1 iPad Pro, it got to me thinking that maybe the M1 is supposed to be the A14X but Apple was surprised by the performance they got out of it that they quickly bolted on the PCIe + whatever circuitry required to ship the base Macs with it. Craig Federighi did say that they were surprised by the M1's performance. That may explain why the I/Os (including external display support) are limited in the existing M1 Macs.

I suspect the next round of the Mx SoC are the actual SoC meant for the Macs.

 

[EntropyQ3]

Now that we have all the various M1 Macs released together with the M1 iPad Pro, it got to me thinking that maybe the M1 is supposed to be the A14X but Apple was surprised by the performance they got out of it that they quickly bolted on the PCIe + whatever circuitry required to ship the base Macs with it. Craig Federighi did say that they were surprised by the M1's performance. That may explain why the I/Os (including external display support) are limited in the existing M1 Macs.

I suspect the next round of the Mx SoC are the actual SoC meant for the Macs.

Err, no. ?
There is no chance of Apple being surprised by the performance of the chip presently known as "M1" and deciding to bolt on a bit of I/O on it. That's not how chip design works. You could argue that it was held back by the fact that it was supposed to serve several product lines, and Apple held back it's I/O capabilities both to save a tiny amount of die area, and to reinforce tiers in their product matrix. But whatever their reasons, the limited I/O of the current M1 Macs is a conscious decision on Apples part, not an afterthought.

 

[leman]

None of the changes mentioned by Bloomberg indicate a new chip architecture, and they all are consistent with the already reported information by Bloomberg, that M1 variants exist and were planned for Pro products in late 2021.

However the changes for the MacBook Air do in fact indicate a new chip architecture.

Also, common sense indicates that the M2 is not debuting ahead of the A15, nor it is debuting as a Pro product. This simply is not how Apple does things.

All these are conjectures. Maybe it’s common sense to you, but it doesn’t make it universally true. The simple fact is that we don’t know. One can argue one way or the other, there are plenty of good arguments for either side. My intuition tells me that the Pro Silicon will be very different from M1, because M1 microarchitecture simply doesn’t cut it for prosumer chips.

 

[senttoschool]

All these are conjectures. Maybe it’s common sense to you, but it doesn’t make it universally true. The simple fact is that we don’t know. One can argue one way or the other, there are plenty of good arguments for either side. My intuition tells me that the Pro Silicon will be very different from M1, because M1 microarchitecture simply doesn’t cut it for prosumer chips.

Agreed. It seems like M1 is basically just a rebranded A14X which was going to be made for the iPad Pro anyways.

The SoCs going forward for Macs will actually be designed for Macs from the start.

 

[gpat]

Apple is obviously going to milk the M1X name before going to M2 and M2X.