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

[senttoschool]

This will probably be a similar situation to the iPad Pros (2020) and iPad Airs. The Air's single core performance beat the Pros, while the multi-core and GPU performance beat the Airs.

Doesn't hold back the Pro buyers at all. They need the 75%+ greater multi-core performance and will pay for it.

In a fall M2 Air vs M1X Pro situation, they'll have an unbeatable "highest performance" low(er)-cost laptop in the market. At the same time, Macbook Pro buyers will have have an unbeatable 75%+ higher multi-core, w/2-4X GPU performance with 32/64GB memory option at twice the price, with the M1X.

This is Apple's SoC roadmap, not Intel. Apple's single-core performance can increase by as much as 20% with each generation, instead of the 1-5% that Intel churns out. It makes zero sense for the $1000 Macbook Air to have 20% faster ST performance over the $2400 16" MBP for 8 months every year.

Single-core performance matters a lot for me because a lot of my professional work-flow depends on single thread.

This is why I and few others are thinking that the upcoming MBPs will actually be based on something else other than the M1. It resets the cycle so that the MBPs will always have the highest performing SoCs for both ST and MT.

 

[dieselm]

This is Apple's SoC roadmap, not Intel. Apple's single-core performance can increase by as much as 20% with each generation, instead of the 1-5% that Intel churns out. It makes zero sense for the $1000 Macbook Air to have 20% faster ST performance over the $2400 16" MBP for 8 months every year.

Single-core performance matters a lot for me because a lot of my professional work-flow depends on single thread.

This is why I and few others are thinking that the upcoming MBPs will actually be based on something else other than the M1. It resets the cycle so that the MBPs will always have the highest performing SoCs for both ST and MT.

If a given workflow benefits from maxing out cores, gpu power, screen size, memory or ports, slightly lower ST performance won't stop pros from finding a reasons to buy it (see: 2019 Mac Pro).

The tradeoff in power (higher clock) or yield (brand new process) is probably not worth the cost to Apple.

 

[leman]

If a given workflow benefits from maxing out cores, gpu power, screen size, memory or ports, slightly lower ST performance won't stop pros from finding a reasons to buy it (see: 2019 Mac Pro).

The tradeoff in power (higher clock) or yield (brand new process) is probably not worth the cost to Apple.

Pros will still buy it if it meets certain minimal expectations. But Apple wants to make the best possible product. They have been very serious about single-threaded performance and they haven’t been happy with Intel recently because it didn’t allow them to build products they wanted to build.

I am quite confident that Apple will try as hard as they can to build the best CPU they can. For them it’s a matter of pride and they are not known to compromise in these areas.

 

[senttoschool]

If a given workflow benefits from maxing out cores, gpu power, screen size, memory or ports, slightly lower ST performance won't stop pros from finding a reasons to buy it (see: 2019 Mac Pro).

The tradeoff in power (higher clock) or yield (brand new process) is probably not worth the cost to Apple.

Macbook Pros are supposed to be superior to Macbook Airs in every way. Mac Pros are a completely different class of products to Macbook Pros.

My guess is that most people buying Mac Pros heavily depend on big GPUs and many CPU cores. More cores just trump everything in their workflows. Examples include high-end video editing and rendering.

Conversely, most people buying Macbook Pros are usually professionals whose workflow depends on a mix of ST, MT, and GPUs, like me. Examples include coding, design.

Again, the Macbook Air having 20% faster ST every cycle for 8 months is just wrong. There's simply no good reason for it.

 

[Kung gu]

Macbook Pros are supposed to be superior to Macbook Airs in every way. Mac Pros are a completely different class of products to Macbook Pros.

My guess is that most people buying Mac Pros heavily depend on big GPUs and many CPU cores. More cores just trump everything in their workflows. Examples include high-end video editing and rendering.

Conversely, most people buying Macbook Pros are usually professionals whose workflow depends on a mix of ST, MT, and GPUs, like me. Examples include coding, design.

Again, the Macbook Air having 20% faster ST every cycle for 8 months is just wrong.

The 2020 intel iMac has better single core than Mac Pro. I think the M1X MacBook Pro will have less single core than the M2 Air.

It’s the same with intel, their laptop chips have higher single core than desktop.

the more high end the computer the better multi core is important.

the m1x MacBook Pro will have same single core as M1.

 

[neilw]

I won't bother to guess what the chips in the forthcoming new machines are going to be. At this point I would say it could go either way.

But I will agree that it makes sense from a marketing standpoint if they can get a cadence going where the pro/"x" variant of the chip is first one out of the gate, and the lower end version follows 6 months (or so) later. This would ensure that the pro machines always have the highest performance chips in all dimensions.

I don't know if it makes sense from an engineering or production standpoint, though. In either case, I would expect that this round of machines will establish the pattern that they will follow for at least the next few years... although we don't know for certain that they're necessarily going to upgrade the entire chip line every year. I tend to think they will, at least for the first few years, but no guarantee.

 

[quarkysg]

Macbook Pros are supposed to be superior to Macbook Airs in every way. Mac Pros are a completely different class of products to Macbook Pros.

My guess is that most people buying Mac Pros heavily depend on big GPUs and many CPU cores. More cores just trump everything in their workflows. Examples include high-end video editing and rendering.

Conversely, most people buying Macbook Pros are usually professionals whose workflow depends on a mix of ST, MT, and GPUs, like me. Examples include coding, design.

Again, the Macbook Air having 20% faster ST every cycle for 8 months is just wrong.

I think the reason why most folks think this way is because of the constraints imposed by Intel and also AMD’s CPU (and also GPU) offerings. We’re used to thinking that thin and light notebooks must be under powered to have any hope of a decent battery life.

I am now thinking that with Apple Silicon, we’ll probably see three classes of performance levels: base, pro and workstation.

Each performance class will come with different form factors with the same performance level. I think Apple has delivered the entire base performance class Macs with the M1 Macs. So whether you want thin and light or heavier but longer lasting notebook, you have it and they both have similar performance level. Similarly with their desktop offerings.

Next week with WWDC21, we’ll see the start of the pro class Macs being introduced, again with the same performance level but different form factors.

This is probably what Apple wanted from Intel all along.

 

[senttoschool]

The 2020 intel iMac has better single core than Mac Pro. I think the M1X MacBook Pro will have less single core than the M2 Air.

It’s the same with intel, their laptop chips have higher single core than desktop.

the more high end the computer the better multi core is important.

the m1x MacBook Pro will have same single core as M1.

Weird. All "x" versions of Apple's chips have had better single-core performance than non "x". For example, iPad Pro A10X had a 14% higher ST than the A10 in the iPhone.

Why do you think it will be different for Macs?

 

[senttoschool]

I won't bother to guess what the chips in the forthcoming new machines are going to be. At this point I would say it could go either way.

But I will agree that it makes sense from a marketing standpoint if they can get a cadence going where the pro/"x" variant of the chip is first one out of the gate, and the lower end version follows 6 months (or so) later. This would ensure that the pro machines always have the highest performance chips in all dimensions.

I don't know if it makes sense from an engineering or production standpoint, though. In either case, I would expect that this round of machines will establish the pattern that they will follow for at least the next few years... although we don't know for certain that they're necessarily going to upgrade the entire chip line every year. I tend to think they will, at least for the first few years, but no guarantee.

It actually makes perfect sense from an engineering standpoint.

When it comes to chips, you want to manufacture the biggest chips first that will go into your top products. Then you take the defective chips and cut them down for use in lower-end products.

For example, if the Macbook Pro 16" comes standard with an 8-core CPU, then you use up all the chips with 8 working CPUs first. Then you take the chips with only 6 working CPU cores and maybe you put them in an upgraded Macbook Air or something. Very few chips go to waste this way.

 

[Kung gu]

Weird. All "x" versions of Apple's chips have had better single-core performance than non "x". For example, iPad Pro A10X had a 14% higher ST than the A10 in the iPhone.

Why do you think it will be different for Macs?

Because I think Apple will not increase the clock speed but rather add more cores.

increasing clock speed means more power draw and more heat. Better spend that heat on more cores.

of course I could be wrong and Apple may choose to add cores and increase clock speed.

 

[senttoschool]

Because I think Apple will not increase the clock speed but rather add more cores.

increasing clock speed means more power draw and more heat. Better spend that heat on more cores.

of course I could be wrong and Apple may choose to add cores and increase clock speed.

Then maybe explain why A10X has more cores and has higher ST than the A10?

 

[Kung gu]

Then maybe explain why A10X has more cores and has higher ST than the A10?

???

 

[CWallace]

I agree that Apple wouldn't make that tradeoff. That's why I think the new 8+2 SoC has a new core design

A15 is entering production to support the iPhone so presuming M2 uses the same cores as A15, then it is possible that instead of M1X we will get M2(X) with the new efficiency ("Son of Icestorm") and performance (Avalanche) cores, but I expect this new SoC to use the A14 Firestorm and Icestorm cores and not the A15's Avalanche and "SoIS".

Again, the Macbook Air having 20% faster ST every cycle for 8 months is just wrong. There's simply no good reason for it.

There is a good reason - chipmaking is not "snap your fingers and it's done!" easy.

If your workload is heavily-biased to single core then you buy a MacBook Air. If your workload is predominately/significantly multi-core than you buy a MacBook Pro and you wait the few extra seconds/minutes on the occasions you need to use single-core.

Weird. All "x" versions of Apple's chips have had better single-core performance than non "x". For example, iPad Pro A10X had a 14% higher ST than the A10 in the iPhone.

Why do you think it will be different for Macs?

Because the various Mac models will (likely) be on different generations of SoCs - think iPad Pro on A10X and iPhone on A11. A11 had better SC performance than A10 and A10X.

 

[Andropov]

There is a good reason - chipmaking is not "snap your fingers and it's done!" easy.

If your workload is heavily-biased to single core then you buy a MacBook Air. If your workload is predominately/significantly multi-core than you buy a MacBook Pro and you wait the few extra seconds/minutes on the occasions you need to use single-core.

Precisely because chipmaking is hard, I think it would be easier for Apple to have the newest, most expensive and hardest to make chips go to the lower volume models. The production rate of new node processes is limited initially, so it would be easier for Apple to have their most expensive computers (low volume) on the newest core/node process.

Now that Apple manages its own chip timeline, it would give a bad look to have the 'Pro' computers stay on an outdated core design for most of the year. Maybe it will happen now that they're in the middle of a transition, but I doubt it'll become norm.

Also, professional workloads are usually mixed, with some steps being single-core only and others being heavily multithreaded, so it's not as easy as choosing a MacBook Air if the workload in heavily single core. And MacBook Airs are limited in other ways (maximum RAM, GPU...).

 

[SteveManila1960]

Precisely because chipmaking is hard, I think it would be easier for Apple to have the newest, most expensive and hardest to make chips go to the lower volume models. The production rate of new node processes is limited initially, so it would be easier for Apple to have their most expensive computers (low volume) on the newest core/node process.

Now that Apple manages its own chip timeline, it would give a bad look to have the 'Pro' computers stay on an outdated core design for most of the year. Maybe it will happen now that they're in the middle of a transition, but I doubt it'll become norm.

Also, professional workloads are usually mixed, with some steps being single-core only and others being heavily multithreaded, so it's not as easy as choosing a MacBook Air if the workload in heavily single core. And MacBook Airs are limited in other ways (maximum RAM, GPU...).

Being at the leading edge is not an obvious strategy. The newest will likely have low yields for a while and back end assembly issues exacerbating initial low wafer volumes. Sometimes it's better to choose a sweeter spot in process technology even at lower volumes.

 

[CWallace]

Precisely because chipmaking is hard, I think it would be easier for Apple to have the newest, most expensive and hardest to make chips go to the lower volume models.

MacBook Pro is not a "low volume" product, however. Prior to M1, comments from Apple strongly imply the MacBook Pro family was the best-selling Mac followed by the MacBook Air (non-Retina).

The production rate of new node processes is limited initially, so it would be easier for Apple to have their most expensive computers (low volume) on the newest core/node process.

The first SoCs of a new generation will always be the "simplest" (least number of cores, lowest amount of RAM, etc.) due to production ramping. That is why each new "A" series launched on iPhone and iPad and then they scaled it up with more performance and GPU cores for the iPad Pro once the production yields and capacity were sufficiently mature to handle more complex designs.

We're seeing the same with the Mac SoCs - A14 launches with 6 CPU (4E+2P) and 4 GPU cores and then it is upgraded to 8 CPU (4E+4P) and 8 GPU cores with M1. And now an even more complex model is on the way with 10 (2E+8P) and 16/32 GPU cores as the FABs are now mature enough to handle it.

And we're going to see the same pattern this Fall - A15 will likely launch with the same 6+4 CPU/GPU configuration in iPhone 13 and then a more powerful 8+8 "M2" will launch for MacBook (Air) either in Nov/Dec or Q1 2022. And then the iPad Pro will get it next Spring and a more powerful "M2X" will likely launch in late Summer / Fall for the Macs.

Now that Apple manages its own chip timeline, it would give a bad look to have the 'Pro' computers stay on an outdated core design for most of the year. Maybe it will happen now that they're in the middle of a transition, but I doubt it'll become norm.

It was common practice on Intel to have the latest and greatest Generation in the entry-level models because they also had to launch it on the least-complex chips (sub-25W). Nobody was losing their minds because a MacBook Air was on 11th Generation Intel with 2 or 4 cores and their iMacs were on 10th Generation with 8 or 10 cores because the high-core count 10th Gen chips would destroy the 11th in pretty much every category - just as "M1X" will destroy "M2" in pretty much every category.

 

[Andropov]

MacBook Pro is not a "low volume" product, however. Prior to M1, comments from Apple strongly imply the MacBook Pro family was the best-selling Mac followed by the MacBook Air (non-Retina).

I said lower, not low. Nothing Apple makes is low volume. Well, maybe the HomePod was. Either way, I imagine the most popular MacBook Pros are the entry level models, not the $2400 16" MacBook Pros. And the entry level MacBook Pros wouldn't get the 'M1X' version.

The first SoCs of a new generation will always be the "simplest" (least number of cores, lowest amount of RAM, etc.) due to production ramping. That is why each new "A" series launched on iPhone and iPad and then they scaled it up with more performance and GPU cores for the iPad Pro once the production yields and capacity were sufficiently mature to handle more complex designs.

To this, I have to agree. Although sometimes the launches of the A and AX versions have come very close in time (A12/A12X were released within a month of each other).

It was common practice on Intel to have the latest and greatest Generation in the entry-level models because they also had to launch it on the least-complex chips (sub-25W). Nobody was losing their minds because a MacBook Air was on 11th Generation Intel with 2 or 4 cores and their iMacs were on 10th Generation with 8 or 10 cores because the high-core count 10th Gen chips would destroy the 11th in pretty much every category - just as "M1X" will destroy "M2" in pretty much every category.

Well, no one complained because a new Intel generation meant a 3-5% increase in single core performance at best, and this was massively offset by the lower core count of the mobile chips. So it was impossible for a MacBook Air with a 11th Gen chip to do anything faster than an iMac with a 10th Gen chip. Now Apple improves the single core performance 20-25% YoY, so the chip generation does make a difference now.

In fact, some of the latest Intel updates were so meaningless that Apple didn't bother to update their Macs with them at all.

 

[Krevnik]

As of now, Apple seems to be confining tasks with a low QoS to run in the efficiency cores only.

Keep in mind that only 1 QoS level is restricted to the efficiency cores, out of 4. And for the description Apple provides, I actually agree that it makes sense to put tasks of that QoS level on the efficiency cores. These sort of tasks shouldn’t even be visible to the user, should be something that can be delayed or interrupted at will. Nothing in this QoS should require a performance core to do. For the examples given in this thread, “utility” makes a lot more sense as the QoS level, not “background”.

Both are entirely possible, but hey, this stuff is funky

Which was my point, honestly. Nothing in the Bloomberg report specified how Jade 2C-Die and 4C-Die are supposed to be packaged, just that they are Jade C-Die doubled and quadrupled.

 

[dieselm]

Keep in mind with 10 cores and proper scheduling, regardless of actual ST performance, "M1X" should feel more responsive than M1.