Will Apple offer fewer processor performance tiers in its Apple Silicon Macs?

[leman]

Really? Not that their contextless chart proves much, but it did lean toward lower power usage across the board. I can't imagine they'd like to keep the 16" at 70-80W if they can get better performance at lower power (and I think they can). You think they'd still keep it at 70-80W max?

I’m guessing it’ll be downgraded primarily because Apple will redefine what the power levels CAN be on a performant laptop. Apple’s not going to use 80 watts or 70 watts if they don’t need to.

Maybe you are right. We will know soon enough

I don't think that we will see a significant change from the chassis TDP levels, simply because these are the figures that make sense for the even form factors. Apple is still going to get good battery life upgrades due to their system architecture, so why would they want to sacrifice performance for a lower TDP?

While Binning is used. It isn't as pervasive as you are implying across broad CPU product line ups . Post 23 above already linked a link that Coffee lake had 3 dies just in the i3-i7 space.

True, but it doesn't mean that Apple has to follow the same strategy 100%. The interesting thing is that Intel has traditionally relied on binning more heavily. I would speculate that their stagnation has pushed them to mix their chips more — for example the 8-core i9 chips are the same as used in the desktop systems, but binned for higher energy efficiency.

 

[theorist9]

• 14" & 16" MacBook Pro would have similar CPU performance with 16" having higher GPU option, thus providing small device option for those who do not need higher graphical performance.

I really like your table, but I don't think you are right about this particular point. Apple has traditionally offered beefier CPUs in the larger MBP, and I don't think they will stop doing it now. In particular, the system performance of the 13" model was limited to 30Watts and that of the 15-16' model was limited to around 70-80 watts. I find it unlikely that Apple will choose to downgrade the TDP of the larger model.

Really? Not that their contextless chart proves much, but it did lean toward lower power usage across the board. I can't imagine they'd like to keep the 16" at 70-80W if they can get better performance at lower power (and I think they can). You think they'd still keep it at 70-80W max?

I’m guessing it’ll be downgraded primarily because Apple will redefine what the power levels CAN be on a performant laptop. Apple’s not going to use 80 watts or 70 watts if they don’t need to. Heck, they’re getting near i9 performance at less than 20 watts! So, if the MBP comes in at 30 watts, it’s won’t be because they made compromises, it’s because that’s all that’s needed for now. The same applies for the future Apple Silicon MacPro. Folks toss around multiple hundred watts ONLY because that’s what’s expected from the world as we know it now. If Apple can provide near Xeon performance in 90 watts, then that’s what they’ll do.

Interesting discussion about the possible TDP of the 16" AS MBP. I don't know what Apple will do, but here's what would make sense to me:

The 16" MBP is supposed to be a serious pro-level machine. Apple has been moving more strongly in that direction since they added their Pro Workflow Team, and had them contribute to the design of the 16".

Thus, if taking full advantage of the current form factor's ~75W TDP capability allows them to put in a CPU/GPU with shockingly high performance, they should certainly do that.

OTOH, if they find that, above a certain TDP (say, 50W, just to pick a number) they get diminishing returns on performance with added power, then it would make sense for them to stay at that lower TDP, and make use of the lower TDP to offer a machine that is thinner, cooler, quieter, and has better battery life.

In sum, I think what determines whether a 75W TDP 16" AS MBP makes sense is the kind of performance scaling they are able to get out of their chip.

[I'm assuming that Apple has decided how many cores make sense for their 16", and thus isn't going to increase the number of cores to meet a 75 W TDP if their processor doesn't scale well above, say, 50 W with the number of cores they originally planned.]

 

[theorist9]

A large factor will be just how much money Apple leaves on the table for not winking in 2-4 more cores. If they can slap another $200-300 dollars on the price of the system, is Apple really going to pass those fatter margins up? If bin testing cost another $60 and they can get $200, that is $140 they are walking away from per machine. ( 1.5M units and that is walking away from $210M. )
[ apple charges around $200-300 for processor upgrade on Mini. $100-250 upgrades for MBA , $200-300 upgrades for MBP 13". Intel pricing isn't driving most of that. apple is only giving fractional credit for standard processor. So a real question us just how hooked Apple is on BTO margin 'crack candy'? Pretty good chance that they are addicted on at least a subset of the Mac product line up. ]

Not sure I'm following your numbers here. If bin testing costs $60, and they find that, say, 20% of chips qualify for the upgraded model, that's $300 in costs to get the extra $200 - $300 you mention (they'd need to test five chips for each one that is upgradeable). I.e, unless I'm missing something, it doesn't seem like your analysis accounts for the failure rate.

 

[ChrisA]

You are all thinking of this as if the only change is from Intel to Arm CPU. That is not what Apple is doing they are moving from Intel to "Apple Silicon", Apple Silcon is not Arm. Rather Arm is one of the componets of Apple Solicon. There are other componets like the GPU but there are also MANY MORE and potentially in t future there might be a dozen different kinds of cores.

Today we saa a GPU and a CPU as the only kinds of processors. But now Apple will have The Arm CPU, some kind of GPU and also Something Googlecalls a "TPU" or others call anueral network processor or "AI Chip", there is also a motion sensor and this list will grow overtime. What Apple can do now is if there is any part oftheir software that is slow, they can buid a hardware core just for that. So if transcoding video format is slow in software Apple can build a "vidio transcoding core".

Now to diffrentiat a low-end MacBook from a Mac Pro that place 8 Arm cores in the MacBook and do the video processing in software. In the mac Pro the have 4 vidio processing cores. The Mac Promight have a better Nueralnetwork processor too.

Apple Silicon is a collection of processors andthey can change the mix of them on the different Macs

 

[Yebubbleman]

Currently (at least as far as I can tell), Apple doesn't offer a choice of processor or graphics performance tier within any of its Apple silicon (AS) form factors (the iPads and iPhones).

I.e., there is only one CPU/GPU option in the 11" iPad Pro, only one in the 10.5" iPad Pro, only one in the iPad, only one in each iPhone form factor (indeed, all current iPhone form factors have identical processors), and so on*.

By contrast, its Intel Macs offer the following number of CPU and GPU choices within each form factor (not necessarily independently) (given as CPU; GPU):
Mini (3; 1), MBA (3; 1), 13" MBP (4; 2), 16" MBP (3; 4), 21.5" iMac (5; 4), 27" iMac (5; 4), Mac Pro(5; 8).

Clearly, there are reasons independent of processor type (Intel or AS) to offer various performance tier choices for each Mac. I.e., the reason there are choices for the Intel Macs but not for the AS i-Devices isn't merely that the former are Intel and the latter are AS.

Yet, at the same time, I am wondering if Apple might decide to offer fewer processor and graphics peformance tiers in each future AS Mac form factor than it does now for its Intel Macs.

For instance, since Intel and AMD already make these processors, there is not much added development cost to offer several processor tiers. But now that Apple is making the chips itself, it may become expensive to offer several different CPU-GPU combinations for each form factor.

Yes, it may just be a matter of changing clock speed, and turning on extra cores; but each time that's done, extensive additional testing is required. Further, Apple may not have as much variation available in its architectures to offer so many choices. E.g., will Apple have (or does it even want to have) the AS equivalent of i3, i5, i7, and i9?

One option that may go away is that of independently selecting GPU memory (as can be done now for the Radeon Pro 5500M in the the 16" MBP), since (IIUC) the AS chips will use unified CPU/GPU memory.


*From Wikipedia:

iPad: 2.32 GHz quad-core (two used) 64-bit (underclocked)

iPad Mini: 2.49 GHz Hexa-core (2× high performance Vortex + 4× high efficiency Tempest)

iPad Air: 1.4 GHz dual-core 64-bit

iPad Pro:

• 12.9-inch 2G, 10.5-inch: 2.34 GHz hexa-core 64-bit
• 12.9-inch 3G, 11-inch 1G: 2.49 GHz octa-core 64-bit
• 12.9-inch 4G, 11-inch 2G: 2.49 GHz octa-core 64-bit

iPhone:

• 11 / 11 Pro / 11 Pro Max and SE (2nd): 64-bit hexa-core Apple A13 Bionic

Back in the PowerPC days, you only had one processor with different frequencies (as stated many times in this thread, a property of binning). Hell, in the Intel Core 2 Duo days (and Core Duo days if you weren't talking about the Early 2006 Mac mini) you only had a single processor with different frequencies. I suspect Apple will revert back to this. No 3-level, 5-level, 7-level, 9-level tiering like you have with Ryzen and Core i. Similarly, as others have suggested, there will probably be either GPU variants to each version (the entry level option will have x graphics, the mid-level option may also have x graphics or y graphics, and the high-end option will have z graphics). At least, given Apple's track record with the Mac spanning the time before Core i3, Core i5 and Core i7 (and certainly Core i9) Macs ever existed, that's what I'm guessing will be the case.

 

[Unregistered 4U]

I'm assuming that Apple has decided how many cores make sense for their 16", and thus isn't going to increase the number of cores to meet a 75 W TDP

Agreed, and remember, they really only have to beat the 2020 MacBook Pros, which I think will be handily done well below whatever Intel’s TDP was/will be.

 

[aeronatis]

Generally in agreement, but, (and this is wishful thinking, certainly ) if we look at the iPad Pro for clues, we see that both the small and large 2020 iPad Pro’s got the same processor with the same graphical capabilities, same RAM and same storage options. The reason customers would pick one over the other is size. So, I think it might be possible that the MacBook Pro 14 and 16 could have the same CPU/GPU option as well. This would mean either the 16 inherits the 14’s 16 core or, even cooler, the 14 would inherit the 16’s 24 core.

In other words, in my brain view, the only difference between MBP’s would be the size, I think even the port options would be the same.

That would even be more amazing for me as I always use my 16" MacBook Pro attached to a Thunderbolt display. Therefore, 14" giving me the same graphical performance at home when I use Final Cut Pro & Motion and being more portable when I need to take it outside would be a dream come true. However, I still believe they will keep at least GPU core count different on the larger model. I would be very glad to be proven wrong, though.

I really like your table, but I don't think you are right about this particular point. Apple has traditionally offered beefier CPUs in the larger MBP, and I don't think they will stop doing it now. In particular, the system performance of the 13" model was limited to 30Watts and that of the 15-16' model was limited to around 70-80 watts. I find it unlikely that Apple will choose to downgrade the TDP of the larger model.

It's not about being right. I am only thinking wishfully. We cannot know for sure before we see the actual products released. Right now they use 45-watt Intel CPU and 50-watt Radeon graphics card on the 16" and it is definitely higher than ideal for that chassis. If they manage to give similar performance with a SOC that has a total TDP of 50-watt of so, they would totally do that, thus giving us silent devices which is actually suitable for video/music production.

We know that CPU/GPU performance doesn't scale infinitely with more power draw, so it is a matter of how scalable their architecture is. As 14" and 16" having similar performance would definitely cannibalise the sales of the 16", I, too, believe that 16" will have more performance at least GPU wise. It would be nice to have similar CPU performance for many use cases like music production, photo editing, code writing etc.

 

[leman]

It's not about being right. I am only thinking wishfully. We cannot know for sure before we see the actual products released. Right now they use 45-watt Intel CPU and 50-watt Radeon graphics card on the 16" and it is definitely higher than ideal for that chassis. If they manage to give similar performance with a SOC that has a total TDP of 50-watt of so, they would totally do that, thus giving us silent devices which is actually suitable for video/music production.

We know that CPU/GPU performance doesn't scale infinitely with more power draw, so it is a matter of how scalable their architecture is. As 14" and 16" having similar performance would definitely cannibalise the sales of the 16", I, too, believe that 16" will have more performance at least GPU wise. It would be nice to have similar CPU performance for many use cases like music production, photo editing, code writing etc.

Let's speculate a bit more about the 16" Apple Silicon model. Based on what we know about current Apple CPUs, they are about 1.5 times faster than Intel CPUs per clock. With 8 high-performance CPU cores running at 3 ghz (+couple of efficiency cores), Apple will be confidently outperforming any current 8 or 10 core design from Intel or AMD, by a large margin. Expect TDP of such a CPU cluster: around 8*5 ~ 40W (5 watts per core). GPU-wise, Apple has a much more efficient graphical architecture, but TBDR doesn't really help with compute. If they want to keep up with dGPUs here, they need to offer a larger GPU cluster. Now, there are some interesting tricks that Apple can do here (and it seems they are doing it, looking at how much faster A14 is in compute tasks), but I don't see much reason for them to hold back. A 32-core GPU should be able to match or outperform something like Pro 5600M in compute, while being at least 50-60% faster in graphics — with the TDP of around 35 watts.

Overall, I think there is little doubt that Appel Silicon Macs will run cooler than current ones. Not only Apple's hardware is more efficient in itself (they could do with 40watts what Intel needs 90watts+ to do), it will offer a lot of opportunities to have power-usage-related optimizations (no more separate RAM for CPU and GPU, no more separate power converters etc.). And finally, no more ridiculous CPU power usage spikes like with Intel, making power management more predictable.

 

[ksec]

Yes, it may just be a matter of changing clock speed, and turning on extra cores; but each time that's done, extensive additional testing is required. Further, Apple may not have as much variation available in its architectures to offer so many choices. E.g., will Apple have (or does it even want to have) the AS equivalent of i3, i5, i7, and i9?

A lot of the discussion are focusing on the MacBook, and didn't take into account of iMac and Mac Pro. That is from 200W+ to 600W of CPU plus GPU combined.

The MacBook Pro isn't quite as interesting, which would be something like doubling of A14X. I expect the A14X to be able to perform up to 30W if not more. You could do a A14X(X) for 80W. And simply bin Cores, / Clock and Yield for each product segment. The additional cost is actually minimal compared to the cost saving from throw away die.

What is remains to be seen is their Desktop Offering.

 

[KushagraSharma]

Really? Not that their contextless chart proves much, but it did lean toward lower power usage across the board. I can't imagine they'd like to keep the 16" at 70-80W if they can get better performance at lower power (and I think they can). You think they'd still keep it at 70-80W max?

I Think having big little technology (heterogeneous processing) they can have max TDP upto 70-80 w but the power cores would become dormant when not required thus increasing efficiency

 

[deconstruct60]

If bin testing cost another $60 and they can get $200,

Not sure I'm following your numbers here. If bin testing costs $60, and they find that, say, 20% of chips qualify for the upgraded model, that's $300 in costs to get the extra $200 - $300 you mention (they'd need to test five chips for each one that is upgradeable). I.e, unless I'm missing something, it doesn't seem like your analysis accounts for the failure rate.

The $60 bin test is extra on top of the basic qualifying testing they would be doing. That why it is "another $60" , not that "testing is $60".

The cost of failed chips from the wafer and basic baseline testing are all build into the baseline price charging for the "lowest" priced option they be would be offering. Even with Apple's you only get one speed and one core configuration they still are doing testing for failures and chucking those that don't pass. That cost is already accounted for in the pricing. This is all extra work past having a basically working chip that brings extra revenue.


So if the die design has 8 cores then can do testing of the full die working. It it all works then it goes down the line. Most of the dies are going down this path. if turning dies off then there isn't a huge testing cost of just avoiding parts that already proven work.


If there is a defect then there is where there is lots of branching and more involved testing. The stuff coming out in defect mode are much smaller subset of the wafer processed. It is only there where truly trying to 'bin' into subcategories with testing. Is the defect recoverable? Can it be cleanly isolated? Retest in the isolated new isolated mode. Still pass ? etc. [ just more involved work than passes completely or chuck it. ]

[ If doing overclockiing testing wouldn't have to select all of them as candidates either. ]

 

[deconstruct60]

....
True, but it doesn't mean that Apple has to follow the same strategy 100%.

Apple cannot follow the strategy 100% because they don't relatively have substantive enough volume. After get away from the top 2 most popular laptops and the iMac ... there isn't much volume to motivate lots of dies without resorting to prices as high or higher than what Intel is charging.

The interesting thing is that Intel has traditionally relied on binning more heavily. I would speculate that their stagnation has pushed them to mix their chips more

Intel hasn't used binning more heavily in the Xeon E5 and now W space. I think you are mixing up where in the line up that Intel has put a low cap on core count. Yeah when there was only 1, 2, or 4 cores in the whole die line up for mainstream products there wasn't much to change. So there is substantively more binned on clock speed. And when AMD was busy shooting themselves in the foot with large caliber weapons, Intel didn't need to have a broader performance range.

— for example the 8-core i9 chips are the same as used in the desktop systems, but binned for higher energy efficiency.

The main target of the Server/Workstation class products are late at moving away from 8 cores as the normal entry point. But it really isn't higher energy efficiency when the effective TDP keeps going up. Right now they are just finding the ones that still run OK when run hotter and making modes to package design to dissipate the additional heat better.

 

[MrGunnyPT]

I’m only worried about GPU.

I think it’s finally the time to give the 13/14 small form factor a better GPU with the transition to Apple Silicon

 

[deconstruct60]

I’m only worried about GPU.

I think it’s finally the time to give the 13/14 small form factor a better GPU with the transition to Apple Silicon

Those form factors would have gotten a substantively better GPU if Apple just put a Gen 11 Intel processor in them with Xe-LP graphics. Apple also seems to have done several evaluations of the advanced AMD APU/Mobile GPUs coming. ( Renoir , Van Gogh , etc. support has popped up hidden in macOS versions. Apple probably got reference boards on the pretense that they'd be switching x86 vendors and probably did "intelligence gathering" on them. ). Apple is probably out to at least keep up with the "Joneses" there.

Some iGPU that is aiming at the midrange of the dGPU class though is probably coming later when there are more 5nm wafer starts available though.

 

[Unregistered 4U]

Those form factors would have gotten a substantively better GPU if Apple just put a Gen 11 Intel processor in them with Xe-LP graphics.

I don’t think they had an option to use a Gen 11 Intel processor before now, though. When are systems with them expected to be generally available?

 

[aednichols]

Dell has 11th gen XPS laptops available already.

Historically Apple takes waaaay longer to ship new Intel chips (they release on a cycle that suits them instead of harmonizing to Intel)

 

[deconstruct60]

I don’t think they had an option to use a Gen 11 Intel processor before now, though. When are systems with them expected to be generally available?

Next month. Probably will arrive before Apple's stuff does.


P.S. for example
" ... The HP Envy x360 13 convertible is available now at Best Buy for a starting price of $949.00, and will hit HP.com in November. The Envy 13 laptop will hit HP.com later this October for $899.99. ... "

HP's New Spectre x360s Include 14-inch Size, 5G Option

HP’s selling a new 14-inch Tiger Lake convertible with a 3:2 display, plus a 10th gen 5G 13-inch.

www.tomshardware.com

 

[raknor]

Dell has 11th gen XPS laptops available already.

They have them on the site but they don't ship till OCT 26th.

Historically Apple takes waaaay longer to ship new Intel chips (they release on a cycle that suits them instead of harmonizing to Intel)

That's not true. The versions Apple uses don't ship till later. When Apple used the iris plus series IGPU SKUs (with EDRAM) those were not released when the rest of the vendors would use the HD versions.

 

[aednichols]

They have them on the site but they don't ship till OCT 26th.

Fair point. But not that long.

The versions Apple uses don't ship till later. When Apple used the iris plus series IGPU SKUs (with EDRAM) those were not released when the rest of the vendors would use the HD versions.

Apple is still shipping 8th gen CPUs in the two lower-end 13" MBP SKUs. They made a business call not to rev the machines because in their concept of value, adopting the CPU bump was not worth it.

 

[raknor]

Apple is still shipping 8th gen CPUs in the two lower-end 13" MBP SKUs. They made a business call not to rev the machines because in their concept of value, adopting the CPU bump was not worth it.

That's because Intel doesn't really have anything much better. The 10th gen SKU at 15Ws is not that much of an upgrade if at all. The 8th gen SKU Apple uses with the 128 MB EDRAM is about as fast as the Ice Lake version.

 

[leman]

Dell has 11th gen XPS laptops available already.

As mentioned above, everyone has 11th gen laptops available — unfortunately those laptops are not for sale

Historically Apple takes waaaay longer to ship new Intel chips (they release on a cycle that suits them instead of harmonizing to Intel)

Apple needs many more of these chips than Dell or anyone else. The XPS is the premium laptop of the Dell brand, the bulk of their laptops are cheaper Inspirons and Latitudes. Given Intel's obvious problems with 10nm production volume, they can serve a customer like Dell, but they won't be able to deliver enough chips to sever Apple. In fact, I wouldn't be surprised if the lower-end MBP still uses 8-th gen simply because Intel doesn't have enough Ice Lake chips to fill the demand.

Intel's inability to ship large quantities of CPUs is probably one of the main reasons Apple takes longer these days to update their CPUs. Before Intel started having production issues Apple was consistently one of the first manufactures to adopt new CPUs.

 

[iPadified]

I really hope Apple does not copy Intels bloated processor lineup.

 

[aednichols]

Intel's inability to ship large quantities of CPUs is probably one of the main reasons Apple takes longer these days to update their CPUs

Hadn't thought of that.

Does not explain previous cases where they e.g. declined to update the Mac Pro and skipped entire generations of Xeons. Those chips ship in vast quantities to server customers so the scarcity explanation does not apply. Likewise the Mac mini uses the same 14nm desktop chips as the iMac.

 

[Unregistered 4U]

Before Intel started having production issues Apple was consistently one of the first manufactures to adopt new CPUs.

And even recently, there have been chips that show up in new Apple laptops prior to even showing up at ark.intel.com.

 

[deconstruct60]

....

Intel's inability to ship large quantities of CPUs is probably one of the main reasons Apple takes longer these days to update their CPUs. Before Intel started having production issues Apple was consistently one of the first manufactures to adopt new CPUs.

In 2012 Apple skipped the Xeon E5 v1. That wasn't Intel's fault at all. Every other major workstation vendor shipped a Xeon E5 system while Apple was doing retreads for another 18 months. Skipped more than a few after the Xeon E5 v2 also for little good reason.

Similar with the Mini and the 2012 -> 2014 gap. That isn't Intel. Really isn't Intel on the 2014 -> 2018 gap either but close enough to where they were eyeball in problems to that I'll wave that one off.

Apple spends are fair amount of time designing themselves into a corner . That also plays a major role in the 1-2 major system upgrades per year of Mac renewals. There is extremely little evidence that Apple is running concurrent , parallel , deep pipelines development across the whole Mac line up. If they focus on one product then some another lags has been the presented evidence.