Rumored pro chip

[CWallace]

So this sounds like it would be an S1.

"S" is used for the Apple Watch System on a Chip. The Apple Watch Series 6 uses the S6 SoC.

Truly. If they're reworked so much that we only need 2 of them here, then why would the report claim we're still getting 4 on the M1 successor and even 8 on whatever will go in the Mac Pro? It's really bizarre.

Because the 4+4 will be used on the iPad Pro and the MacBook Air and in those applications, battery life is as important as overall performance and four efficiency cores improves battery life.

The MacBook Pros will have larger batteries so while only having two efficiency cores will increase battery drain when jobs have to be off-loaded to the performance cores, the larger battery capacity will help offset this.

Desktops arguably do not need efficiency cores since they are always on main power, but there are side benefits with lower power consumption and lower heat generation when they can do the job instead of handing it off to a performance core.

According to Mark Gurman:

For the new MacBook Pros, Apple is planning two different chips, codenamed Jade C-Chop and Jade C-Die: both include eight high-performance cores and two energy-efficient cores for a total of 10, but will be offered in either 16 or 32 graphics core variations.

Are these the codenames for the SOCs or the CPUs?

This should be the codename for the SoC. In late 2020, the China Times released the rumored codenames for the A14, A14X (which became the M1 on release) and a more powerful SoC for the iMac and MacBook Pro they called "A14T". The codename for the "A14T" was "Mt. Jade". Note, this implies that Jade-C could still be on the A14 and an "M1 class" SoC.

I am presuming the SoC with 32 GPU cores is "Jade-C Die" and then "Jade-C Chop" is the one with 16 cores, either by electrically disabling half the cores of the 32 model or using a different mask that only has 16 cores.

The name of the performance core for the A15 is said to be "Avalanche" (the A14's are called "Firestorm").

 

[cmaier]

What's that mean? I thought the way people were using M2 vs. M1x is the number referred to the generation, and the letter referred to the "series". Thus an M2 would be the next-gen low end chip, and the M1x would be the first-gen midrange chip. Given this, wouldn't the 10-core chip be an M1x and not an M2?

Not that I like that nomenclature There should be a different letter for each series, following by a number for the generation. Something like this would be much cleaner (gotta have an S in there somewhere, because all the carriage trade companies to do it):

iPad: A#
low-end Mac: M#
MBP: S#
high-end iMac: T#
Mac Pro: Z#

So this sounds like it would be an S1.

M2 means a different core microarchitecture than M1. M1X means same core microarchitecture as M1. That is all. You don’t name chips based on what products you put them in.

 

[Andropov]

Just as a quick note, these things are actually explicitly mentioned in the official documentation: https://developer.apple.com/news/?id=vk3m204o

Yes, but the details of how QoS relates to the efficiency/performance cores is pretty vague. Apple's documentation mentions that background processes may run on the efficiency threads, but I would have expected them to switch to the performance cores if the efficiency cores were at 100% capacity. But they seem to be very reluctant to do so.

It may be just me, but the fact that an API from 2014 (QoS) synergizes so well with the heterogeneous architecture of the new cores is insanely cool. Apps making full use of GCD and QoS-labeled queues get quite an improved experience for 'free' on M1 (vs. using other alternatives as OpenMP or the like).

 

[NT1440]

Yes, but the details of how QoS relates to the efficiency/performance cores is pretty vague. Apple's documentation mentions that background processes may run on the efficiency threads, but I would have expected them to switch to the performance cores if the efficiency cores were at 100% capacity. But they seem to be very reluctant to do so.

It may be just me, but the fact that an API from 2014 (QoS) synergizes so well with the heterogeneous architecture of the new cores is insanely cool. Apps making full use of GCD and QoS-labeled queues get quite an improved experience for 'free' on M1 (vs. using other alternatives as OpenMP or the like).

I think it speaks volumes to the Longview Apple has had on this transition. Some people here act like it’s a new and risky initiative when the reality is Apple has been laying the groundwork on the software front for Apple Silicon for the better part of a decade before any hardware was close to production. I knew the second the 64 but A series was announced that it was just a matter of time. They’ve done an incredible job in planning and preparation for this.

 

[Neodym]

An Apple-based 128-core GPU will most likely be the fastest GPU on the market, save for the dual-chip systems.

Only half-serious: Now if Apple would enter the (gaming) GPU market with that … As prices and availability of the usual suspects’ offers are currently very unattractive, it could be a perfect time for a competitor to show up.

In reality, though, Apple probably couldn’t keep up in the Software department. But it’s a nice daydream, isn’t it? ;-)

 

[thingstoponder]

I don’t think 2 M1 efficiency cores is going to be a problem. This chip won’t be geared for as low of power usage as the M1. I think 2 cores will still be good enough for background tasks.

 

[cmaier]

I don’t think 2 M1 efficiency cores is going to be a problem. This chip won’t be geared for as low of power usage as the M1. I think 2 cores will still be good enough for background tasks.

High efficiency cores are not just for background tasks, and there are a lot of threads at any given time that could benefit from them. Which makes me think these are very different cores than those in the M1.

 

[Zdigital2015]

Remember that intel isn't standing still either. And I don't think it'll be faster than the CPU's you suggest, but it'll be close, but we'll have to wait and see the clocking specs and the benchmarks. The fastest i9's (X and XE parts) are well over twice as fast as the M1...

Single-core on the M1 is 15% faster than the 11980HK already, even if the M1 is lagging behind by 45-50% on multi-core. You have to be seriously out of touch with reality to suggest that an 8p/2e setup in an M1X/M2 won’t resoundingly crush the 11980HK on multicore. Intel isn’t standing still but they aren’t taking big steps in the kid they find themselves mired in. Why do so many apologize for Intel when it’s clearly being bested on so many fronts? Karma is a bitch and they have reaped the whirlwind.

 

[sublunar]

I think two efficiency cores will be enough to handle the general background tasks and low-power needs, allowing more die space to be used to for the additional performance cores.

As for the GPU, there are the rumors of the "Lifuka" GPU so this could be an "external" GPU to keep the die size down on the main SoC by removing the on-die GPU cores as found on the M1. This GPU could still be on-package (like the DRAM of the M1) so it would benefit from physical proximity and high-bandwidth connections. We could also see a mix of on-die and on-package GPU cores like we have now with the Intel iGPU and AMD dGPUs.

I wonder if M1X (if based on the A15 cpu) is able to use more powerful cores because all of the cores are just more power efficient? It would be like the transition between A12 and A13 vs between A13 to A14.

I put my theory in the other thread on the main news page so I am loathe to repeat myself but in an effort to summarise, Apple can probably deliver performance and efficiency improvements year on year which is why this year's A15 should be slightly more powerful and slightly more efficient.

But the even number A series CPUs might be where they deliver the process shrink which leads to a more substantial process improvement and power efficiency. And couple with the fact that availability of LPDDR5 and Thunderbolt 4 could see a CPU called M2 debut next year based on the A16 when it goes to 4nm - I think we might only find major Mac and iPad updates every 2 years - and only in even numbered years.

So M2 could well be where Apple deliver the quantum leap in benchmarks and specs, but it'll be for next year's Macs in my current opinion.

 

[bobcomer]

Single-core on the M1 is 15% faster than the 11980HK already, even if the M1 is lagging behind by 45-50% on multi-core.

Since I haven't seen the 11980HK in the benchmarks I look at, I can't comment about that, but I can say single core doesn't mean anything to me, it's just a benchmark thing, multicore does. And since we both have not seen a benchmark for the M2, perhaps we should wait and see. And besides, the 11980HK isn't intel's fastest chip, at least the 10980HK isn't even close, as it's a slow mobile chip. (A chip I'd *never* buy!)

 

[Joelist]

Alas we don't totally know how Apple is planning to advance the SOC line now that there are multiple families using the same core microarchitecture. Maybe the successor to Firestorm will launch on M series then go to A series?

 

[theorist9]

M2 means a different core microarchitecture than M1. M1X means same core microarchitecture as M1. That is all. You don’t name chips based on what products you put them in.

And yet names are based on architecture, architecture is based on use, and use is based on what class of products they go into. And besides, Intel did label chip series based on intended product class, e.g., "Core i7 Mobile."

I was thinking of calling them low power consumer /mid power consumer/high power consumer/pro, but thought it might be more understandable if I illustrated the different series with examples.

In any event, I think having letters designate the series, and numbers designate the generation, would be a nice, logical naming scheme. Then again, I'm a chemist, so I'm used to pleasingly logical nomenclature.

 

[cmaier]

And yet names are based on architecture, architecture is based on use, and use is based on what class of products they go into. And besides, Intel did label chip series based on intended product class, e.g., "Core i7 Mobile."

I was thinking of calling them low power consumer /mid power consumer/high power consumer/pro, but thought it might be more understandable if I illustrated the different series with examples.

In any event, I think having letters designate the series, and numbers designate the generation, would be a nice, logical naming scheme. Then again, I'm a chemist, so I'm used to pleasingly logical nomenclature.

Apple doesn’t sell CPUs. Intel does. Big difference.

 

[theorist9]

No, I didn't, but it makes sense. An Apple-based 128-core GPU will most likely be the fastest GPU on the market, save for the dual-chip systems.

As you know, the Mac Pro video cards are intended for tasks like video rendering rather than data center use. That means this hypothetical 128-core GPU will be competing with the NVIDIA RTX A6000 and the AMD Vega Pro II Duo. I don't know if the ATX is internally multi-chip, but it is packaged as a single card, and has claimed peak FP32 of 38.7 TFLOPS, which beats the AMD Pro Vega II Duo's claimed peak FP32 of 28.4 TFLOPS.

How might a 128-core Apple GPU compare with these?

 

[theorist9]

Apple doesn’t sell CPUs. Intel does. Big difference.

If anything, the fact that Apple sells finished product makes it more likely they would to want to label their chips by what they go into, rather than less, since that improves product line differentiation. You get better product differentiation by having the chips in the low-end Macs be M-series, and those in the MBP be, say, P-series (or whatever letter of the alphabet you want to use), etc.

Furthermore, in instances when different chips are going into different products, this appears to be exactly what Apple is already doing: A-series for iPhone/iPad, H-series for headphones/earbuds, S-series for Apple Watch, M-series for low-end MacOSdevices, etc.

 

[cmaier]

If anything, the fact that Apple sells finished product makes it more likely they would to want to label their chips by what they go into, rather than less, since that improves product line differentiation. You get better product differentiation by having the chips in the low-end Macs be M-series, and those in the MBP be, say, P-series (or whatever letter of the alphabet you want to use), etc.

Furthermore, in instances when different chips are going into different products, this appears to be exactly what Apple is already doing: A-series for iPhone/iPad, H-series for headphones/earbuds, S-series for Apple Watch, M-series for low-end MacOSdevices, etc.

That there is backwards thinking. Apple goes the other way. “Buy this ipad, and you get the same power and speed as this desktop computer!”

 

[EntropyQ3]

In any event, I think having letters designate the series, and numbers designate the generation, would be a nice, logical naming scheme. Then again, I'm a chemist, so I'm used to pleasingly logical nomenclature.

??

Anyway, the rumours are nice food for speculation, but I would like to hear something about the memory subsystems that are supposed to keep the beasts fed. The rumours are always in a "number of cores" format, with easy factor of two scaling and could just as well be super simplistic extrapolation by just about anyone exposed to Android cell phone marketing. I haven’t picked up any supply chain whispers about anything regarding new Macs apart from some difficulties with mini-LED backlighting supply issues, which is frustratingly irrelevant to architecture.
When the only bone you have to gnaw on is "it will have cores", it gets a bit dry.

I wish Apple would let their engineers have some public space where they could talk about what they’ve achieved and how. Let the geeks have some time, if not in the sun, then at least out of the closet, where they could talk about data paths, interlock avoidance, handling differences in access patterns to shared memory of different functional units, latency minimizing and hiding, power control, all the glorious nuts and boltery of their products.

"Our greatest *** ever!" and "Now even thinner!", really doesn’t inspire much.
It’s thin…and has colours? The intellectual level is very modest indeed.
As if the technology that enables the products is embarrassing or something.

 

[theorist9]

That there is backwards thinking. Apple goes the other way. “Buy this ipad, and you get the same power and speed as this desktop computer!”

And that there is limited thinking. You're missing that when you get into that territory, you also have: "I'm paying all this extra for a Mac desktop, but it only has the speed and power of an iPad!" Which is why Apple doesn't play that game. Instead, Apple seeks to maintain differentiation between their product lines. Why else would they offer two entirely different color schemes for the iPhone 12 and the iPhone 12 Pro? It's certainly not for any functional reason! Maybe you should stick to your legal briefs. This seems to be beyond you.

And what's gotten into you anyways? You always used to be collegial in your responses. That's part of why I enjoyed interacting with you. Now you're descending to personal insults like "backwards thinking". Yeah, I did the same in this post, but that was just me swinging back.

 

[theorist9]

??

Anyway, the rumours are nice food for speculation, but I would like to hear something about the memory subsystems that are supposed to keep the beasts fed. The rumours are always in a "number of cores" format, with easy factor of two scaling and could just as well be super simplistic extrapolation by just about anyone exposed to Android cell phone marketing. I haven’t picked up any supply chain whispers about anything regarding new Macs apart from some difficulties with mini-LED backlighting supply issues, which is frustratingly irrelevant to architecture.
When the only bone you have to gnaw on is "it will have cores", it gets a bit dry.

I wish Apple would let their engineers have some public space where they could talk about what they’ve achieved and how. Let the geeks have some time, if not in the sun, then at least out of the closet, where they could talk about data paths, interlock avoidance, handling differences in access patterns to shared memory of different functional units, latency minimizing and hiding, power control, all the glorious nuts and boltery of their products.

"Our greatest *** ever!" and "Now even thinner!", really doesn’t inspire much.
It’s thin…and has colours? The intellectual level is very modest indeed.
As if the technology that enables the products is embarrassing or something.

It would be nice to hear the engineers give more technical details in a way that's not marketing driven (i.e., where the technical details really are technical, and not just another form of marketing). But that's not been Apple's culture--they like to play things close to the vest. There have been some presentations about the M-series chips, but you can tell there are a still a lot of fuzzy areas. Granted, much of this is proprietary. But there probably are areas where Apple could talk in more depth without giving anything away, yet they choose not to, because they're Apple. And you can understand why they continue with this approach, because it has worked for them very well thus far (usually).

 

[thingstoponder]

High efficiency cores are not just for background tasks, and there are a lot of threads at any given time that could benefit from them. Which makes me think these are very different cores than those in the M1.

What are they for then? My understanding is that they power the background tasks like photoanalysisd and also kick in when the computer is sleeping for low power tasks.

I dont see why next gen cores will be any different anyways other than just being a little faster.

 

[quarkysg]

What are they for then? My understanding is that they power the background tasks like photoanalysisd and also kick in when the computer is sleeping for low power tasks.

I dont see why next gen cores will be any different anyways other than just being a little faster.

Well, you could look at it this way.

If Apple is confident enough to reduce the HE cores from 4 to 2, and assuming that 4 HE cores are actually required for the M1, then the new HE cores would be at least 100% more performant compared to the M1 HE cores to take the impact of a 50% reduction in cores to at least maintain the same performance level.

 

[theorist9]

What are they for then? My understanding is that they power the background tasks like photoanalysisd and also kick in when the computer is sleeping for low power tasks.

I dont see why next gen cores will be any different anyways other than just being a little faster.

This might help paint a more detailed picture of how they are used: https://developer.apple.com/news/?id=vk3m204o

 

[leman]

As you know, the Mac Pro video cards are intended for tasks like video rendering rather than data center use. That means this hypothetical 128-core GPU will be competing with the NVIDIA RTX A6000 and the AMD Vega Pro II Duo. I don't know if the ATX is internally multi-chip, but it is packaged as a single card, and has claimed peak FP32 of 38.7 TFLOPS, which beats the AMD Pro Vega II Duo's claimed peak FP32 of 28.4 TFLOPS.

How might a 128-core Apple GPU compare with these?

An 128-core Apple GPU with the same specs as an M1 would have 16384 ALUs and peak throughput of 41.6 FP32 TFLOPS.

 

[theluggage]

Are these the codenames for the SOCs or the CPUs?

Simply going on the names "chop" vs. "die" (for whatever that is worth) it would sound like one is a single-die "true" SoC while the other has separate CPU and GPU dies (possibly still in the same package, so have fun discussing whether it is still a SoC).

 

[leman]

Simply going on the names "chop" vs. "die" (for whatever that is worth) it would sound like one is a single-die "true" SoC while the other has separate CPU and GPU dies (possibly still in the same package, so have fun discussing whether it is still a SoC).

It was speculated that ”chop” is a “die” with 16 GPU cores cut off. There is some logic to that I suppose…

https://twitter.com/i/web/status/1395706013286809600