Future Apple Silicon Macs Will Reportedly Use 3nm Chips With Up to 40 Cores

[grilledcheesesandwich]

I have not forgotten about legacy windows/x86 software. But legacy software support does not mean that Microsoft can’t offer ARM software or the entire planet has to stay on x86/64 for them to service that market. I doubt Microsoft will stop making Intel versions of windows and office or any of their other core software in any of our lifetimes, but that does not mean these will be the only versions they will make. Companies that have been reliant on legacy software, until now, have been doing so while still getting generation over generation performance increases due to hardware advances and may start reevaluating those decisions when that train starts to slow. Legacy software entrenchment in business has also been eroding on many fronts over the years and operating system support alone is not going to save it. Nor should it. Even Microsoft has done a lot with Windows 11 to make a demarcation point as to what they considered unsupportable cruft like nixing NTVDM support and other things that make continuing to use 16 and 32-bit software painful.

It won’t take a monumental shift in the market for Intel to be in trouble. It will just take enough people liking light, fan-less, all-day-running notebooks that run windows and office respectfully well alongside a few common programs to sour their fortunes. Over a year ago Intel stock fell 6% on rumors that Microsoft was going to design their own server and laptop chips. Now imagine how bad that would be if it were Dell or HP and you will understand why Intel is spending million on ads to downplay MacBooks.

I doubt Microsoft will simultaneously support two different enterprise architectures. Windows 11 compatibility requirements remain worlds apart from breaking x86 functionality. Many Fortune 500 businesses reliant on legacy pipelines would vote against your presumptions with their wallets. Like it or not, Intel's existential problems lie in the desktop & server markets, not mobile computing.

I don't think any company wants to lose 10% of one of their major markets. Anyone who thinks Intel is OK with losing Apple as a customer really doesn't understand how business works. Apple is about 1/2 of Dell for notebooks.

I don't think you understand Intel's business model, which heavily depends on desktops & servers. Long-term OEM contracts & the relative competitiveness of Intel's mobile chips versus desktop ones mean Intel has bigger fires to fight, i.e., AMD. In this context, Apple's business remains a rounding error to Intel, albeit a prestigious one, which explains the recent campaign to woo Cupertino back.

 

[mdriftmeyer]

Zen 5 will use 3nm as well, but TSMC has postponed 3nm back to 2023.

 

[LIVEFRMNYC]

I doubt Microsoft will simultaneously support two different enterprise architectures. Windows 11 compatibility requirements remain worlds apart from breaking x86 functionality. Many Fortune 500 businesses reliant on legacy pipelines would vote against your presumptions with their wallets. Like it or not, Intel's existential problems lie in the desktop & server markets, not mobile computing.


I don't think you understand Intel's business model, which heavily depends on desktops & servers. Long-term OEM contracts & the relative competitiveness of Intel's mobile chips versus desktop ones mean Intel has bigger fires to fight, i.e., AMD. In this context, Apple's business remains a rounding error to Intel, albeit a prestigious one, which explains the recent campaign to woo Cupertino back.

IDK about that. The enterprise world seems to be moving on too. There maybe only several fields as a whole that still choose to completely reply on MS.

 

[sideshowuniqueuser]

Future Macs will also use 1nm with up to 80 cores. Then after that, they move on to -1 nm chips.

Future Macs will have the screen printed on the inside of your eyeballs, and the M37 Pro Max Pro Extreme Super Pro chip embedded in your chest.

 

[grilledcheesesandwich]

IDK about that. The enterprise world seems to be moving on too. There maybe only several fields as a whole that still choose to completely reply on MS.

Sure, Linux dominates cloud servers & other applications independent from legacy tooling, but Windows still dominates enterprise IT largely because grandfathered software remains a PITA for businesses to overhaul. "If it ain't broke, don't fix it" very much applies.

 

[visualseed]

I doubt Microsoft will simultaneously support two different enterprise architectures. Windows 11 compatibility requirements remain worlds apart from breaking x86 functionality. Many Fortune 500 businesses reliant on legacy pipelines would vote against your presumptions with their wallets. Like it or not, Intel's existential problems lie in the desktop & server markets, not mobile computing.

Microsoft's interests and Intel's are no longer as aligned as they once were. Enterprise is no longer Intel's bread and butter as it is for Microsoft and Microsoft is already making ARM software. So I don't see a reality where change fearing IT managers are going to stop that from spilling over into the general computing market where Intel is more vulnerable to competition from different chip architectures. Microsoft doesn't have a technological upside in supporting legacy software on Windows other than it keeps the door open to selling their own enterprise software and keeps windows the dominate OS in the market. The sooner an outside paradigm shift in computing can help them move off that square without a backlash that jeopardizes their upsides, the better.

Intel is losing the profitability battle to AMD in the datacenter and their own less profitable chips on the desktop. Think about how many businesses are buying barebones systems with min spec Core i5s because they are good enough to run windows and office. Just last month Intel, again, announced an earnings miss because of underwhelming server chip sales. Mobile CPUs (notebooks) which is now, primarily, a consumer driven market, the only market they are doing well in (I think it was 41% or their revenue and +60% of their profits in 2021) and the only reason they kept overall revenues above water. This is why they are frantically trying to protect it.

 

[lysingur]

Sure, Linux dominates cloud servers & other applications independent from legacy tooling, but Windows still dominates enterprise IT largely because grandfathered software remains a PITA for businesses to overhaul. "If it ain't broke, don't fix it" very much applies.

Microsoft has pulled the plug on 32-bit applications in Windows 11. Whether they like it, eventually those enterprises and the public sector have to get with the times too. In some cases, it's not so much that the management doesn't want to move on but the union worries about IT personnel who haven't upgraded their skills since Windows XP losing their jobs and prevents progress. You can get new machines but you can't fire people who don't know how to service them properly.

 

[grilledcheesesandwich]

Microsoft has pulled the plug on 32-bit applications in Windows 11. Whether they like it, eventually those enterprises and the public sector have to get with the times too. In some cases, it's not so much that the management doesn't want to move on but the union worries about IT personnel who haven't upgraded their skills since Windows XP losing their jobs and prevents progress. You can get new machines but you can't fire people who don't know how to service them properly.

Again, deprecating 32-bit remains in a different ballpark from obsoleting x86. Blaming tech support unions for the slow progress of enterprise IT sounds bizarre to me.

 

[sideshowuniqueuser]

Today's Apple Silicon chips & form factors vs those rumored to be out in 2022.
​

Form Factor​	Mac Pro iMac Pro​	Mac mini Pro Mac Pro iMac Pro​	MBP 14" MBP 16" Mac mini Pro iMac 24" iMac Pro​	MBP 14" MBP 16" Mac mini Pro iMac 24"​	MBA MBP 13" iMac 24" Mac mini 2021 iPad Pro​	iPhone 13 iPhone 13 mini iPhone 13 Pro iPhone 13 Pro Max 2021 iPad Mini​
Apple M1 Max chip​	M1 Max Jade-4C​	M1 Max Jade-2C​	M1 Max​	M1 Pro​	M1​	A15​
Launch​	Q2 or Q4 2022​	Q2 or Q4 2022​	Q4 2021​	Q4 2021​	Q4 2020​	Q3 2021​
# of dies​	4​	2​	1​	1​	1​	1​
CPU​	40​	20​	10​	10​	8​	6​
performance cores​	32​	16​	8​	8​	4​	2​
efficiency cores​	8​	4​	2​	2​	4​	4​
GPU core​	128​	64​	32​	16​	8​	5​
Neural Engine core​	64​	32​	16​	16​	16​	16​
memory bandwidth​	1,600GB/s​	800GB/s​	400GB/s​	200GB/s​	68.2GB/s​	68.2GB/s​
Max Memory​	256GB​	128GB​	64GB​	32GB​	16GB​	6GB​
Hardware-accelerated H.264, HEVC, ProRes, and ProRes RAW​	4​	2​	1​	1​	-​	-​
Video decode engines​	4​	2​	1​	1​	-​	-​
Video encode engines​	8​	4​	2​	1​	-​	-​
ProRes encode and decode engines​	8​	4​	2​	1​	-​	-​
Transistors​	228 Billion​	114 Billion​	57 Billion​	33.7 Billion​	16 Billion​	15 Billion​
Estimated Die Size​	17.004cm²​	8.502cm²​	4.251cm²​	2.513cm²​	1.19cm²​	1.116cm²​
Estimated AMD Ryzen 9 5950X Performance​	4x​	2x​	1x​	-​	-​	-​
Estimated RTX 3080 Performance​	4x​	2x​	1x​	-​	-​	-​

Watch the video below for an explanation how Apple will possibly use TSMC's updated chiplet tech to attain such insane transistor counts.

As early as 2025 we can see TSMC's 2nm node process be applied to iPhones then the rest of Apple silicon devices. If 5nm & 2nm are not marketing terms but actual measures then expect the chips to shrink by ~60%.

Wow, what a wonderful table, thanks for publishing! Could I politely suggest that you could maybe put it up on a Google Sheets spreadsheet, and give it public read-only access? That way, as you update it, other people can bookmark it and refer to it. PS - it's a free service, in case you weren't already aware.

I do find your predictions of which machines will have which chips quite interesting. I would be surprised, however, if the Mini Pro has a more powerful chip than the MBP, but interesting idea. The idea that an iMac Pro will have the same chip options as the Mac Pro is also an interesting thought. Exciting times!

 

[grilledcheesesandwich]

Microsoft's interests and Intel's are no longer as aligned as they once were. Enterprise is no longer Intel's bread and butter as it is for Microsoft and Microsoft is already making ARM software. So I don't see a reality where change fearing IT managers are going to stop that from spilling over into the general computing market where Intel is more vulnerable to competition to different chip architectures. Microsoft doesn't have a technological upside in supporting legacy software on Windows other than it keeps the door open to selling their own enterprise software and keeps windows the dominate OS in the market. The sooner an outside paradigm shift in computing can help them move off that square without a backlash that jeopardizes their upsides, the better.

Intel is losing the profitability battle to AMD in the datacenter and their own less profitable chips on the desktop. Think about how many businesses are buying barebones systems with min spec Core i5s because they are good enough to run windows and office. Just last month Intel, again, announced an earnings miss because of underwhelming server chip sales. Mobile CPUs (notebooks) which is now, primarily, a consumer driven market, the only market they are doing well in (I think it was 41% or their revenue and +60% of their profits in 2021) and the only reason they kept overall revenues above water. This is why they are frantically trying to protect it.

If you have data showing enterprise no longer underwrites Intel's business, please do share. Microsoft's resources for Windows remain overwhelmingly dedicated to x86, especially so for enterprise.

Again, the relative competitiveness of Tiger Lake versus Rocket Lake proves Intel has bigger fish to fry in the desktop & server markets versus mobile. The brisk timeline for Raptor Lake & Meteor Lake proves Intel remains far more "frantic" regaining market share from AMD than worrying about preexisting long-term OEM contracts for mobile.

Nobody thinks Pat Gelsinger will turn Intel around easily, but ADL looks like a good start.

 

[smulji]

And it's been happening at Apple since the 90's

That's false. It happened since Tim Cook took over as CEO.

 

[smulji]

This happens across the entire industry, not just at Apple.

Who gives a s**t. Two wrongs don't make a right.

 

[quarkysg]

I still dont see how this Chiplet "rumour" will work without redesigning the M1 Max in the first place due to NUMA and memory access. But given this rumours is now literally reported again and again and it fast booming fact. We will see what happen when it launch.

Cache coherency between all the chiplets will be handled by the internal interconnect fabric. All IPs in all chiplets will see the same memory image. macOS will not need NUMA support.

 

[smulji]

They were doing it anyway

there wasn't any diversity program under Steve Jobs. He didn't have patience for those type of things. His philosophy on building a great team was hiring "A" players because he knew "A" players would hire more "A" players. He hired based on talent not based on some politically correct nonsense

 

[nathansz]

Apple is gonna lay waste to Intel and the rest of the PC industry. Hard to see how Apple does NOT double its market share in 3-5 year time

by not running windows or linux, that’s how

 

[Shirasaki]

by not running windows or linux, that’s how

Tbf, Linux community is trying real hard to let one of their distros running on M1 Mac natively. However, apple is frankly against the idea so Linux community is completely on their own. Windows… I’d rather emulate X86 version of windows than waiting for ARM version of windows to arrive on apple silicon Mac. Oh wait, I already HAVE a windows PC and it works well.

 

[ratspg]

Why? It was a machine for pros, making money with their computer. If they were really pros, then their machines will have gotten them 3 good years, and when the time comes for an upgrade, it'll presumably be a nice bump. If their work shows a benefit in moving to one of the M1 Max or Pro chips now, then they'll still have gotten 2 good years of revenue from their machines. Now if you're talking about a hobbyist who really wasn't the market for the Mac Pro but they plunked down the money anyway... once again, it's not like the ARM Mac Pros came out right after they made their purchase. In fact, they're still a year away. So your comparison is silly. Future chips rule. But current chips is how a pro actually earns his or her living. It's not about fanboyism. It's about doing your job right now with the tools available.

I made money just fine with my Power Mac G5 it was a great tool. But it wasn’t worth the money spent. The Intel Mac Pro purchased to replace it lasted 3x as long for me. The G5 had a very short life span and wasn’t as valuable as the Intel Mac that replaced it quickly. I had a custom PC built at one point that was for professional use and outlasted both the G5 and the Intel Mac I had. Of course technology improves, but regardless the Power Mac G5 was overhyped and underdelivered. Speaking from personal experience using that product.

 

[Adarna]

Mac: M3 - *Hardware accelerated RT (Part 1)

It is nothing new. AMD uses something similar: Yes, Apple would need a super fast bus connections between chips. No, it is cost effective solution. AMD proved this. Architecture updated with cross connections 🙂

blenderartists.org

View attachment 1903387

This is is possible through TSMC”S chiplet tech.

The estimated die size of that 5nm node process M1 Max Quadra may be 17.004cm².

By 2025 when TSMC”s 2nm node process is offered it would allow to shrink this by 60%.

 

[dogslobber]

The successor to the Mac Pro needs to be able to hit at least 256GB of memory minimum. Probably more likely 512GB. Anything less is a toy.

 

[Adarna]

The successor to the Mac Pro needs to be able to hit at least 256GB of memory minimum. Probably more likely 512GB. Anything less is a toy.

You buying one?

 

[Adarna]

Wow, what a wonderful table, thanks for publishing! Could I politely suggest that you could maybe put it up on a Google Sheets spreadsheet, and give it public read-only access? That way, as you update it, other people can bookmark it and refer to it. PS - it's a free service, in case you weren't already aware.

I do find your predictions of which machines will have which chips quite interesting. I would be surprised, however, if the Mini Pro has a more powerful chip than the MBP, but interesting idea. The idea that an iMac Pro will have the same chip options as the Mac Pro is also an interesting thought. Exciting times!

Thank you! I'd love to make one using Google Sheets in the future. Pls be aware that the data I provided is largely conjencture at best.

I call it the "Mac mini Pro" for three reasons

• The curent Mac mini has a 150W PSU that was carried over from the Mac mini Core i5 & Core i7 with max power consumption of 85W & 122W respectively. Apple disclosed that its peak power load for the Mac Mini M1 is only 39W​
• The MBP 14" M1 Max & M1 Pro can be powered full tilt by the bundled 96W charger. So when we remove its mini LED display's power overhead it would be possible for that Mac mini's 150W PSU to power a M1 Max, "M1 Max Duo" or even a "M1 Max Quadro". At top spec'd BTO it could cost beyond $6,000.​
• There's rumor of the return of the Mac Cube​

Nvidia, AMD and Intel will look really bad when Apple delivers a pro desktop with 2-4x the performance of an AMD Ryzen 9 5950X desktop CPU & 2-4x the performance of a NVidia RTX 3080 dGPU at a fraction of the power draw and fraction of operating noise.

 

[dogslobber]

You buying one?

Probably for my MIL.

 

[Adarna]

Probably for my MIL.

lucky in-law.

 

[Apple_Robert]

Apple is gonna lay waste to Intel and the rest of the PC industry. Hard to see how Apple does NOT double its market share in 3-5 year time

Whether or not that happens depends (in part) on what kind of official cross platform support Microsoft offers.

 

[Adarna]

There was just an article on MR the other day saying Apple's notebook market share had 10% YOY growth and 10% market share.

MacBook Air Drives 6.5 Million Apple Laptop Shipments in Q3 2021

I don't think any company wants to lose 10% of one of their major markets. Anyone who thinks Intel is OK with losing Apple as a customer really doesn't understand how business works. Apple is about 1/2 of Dell for notebooks.

View attachment 1903536

Apple does not buy the cheap stuff & they are the trend setter where most of the market aspire to be.

Apple's all time high market share was 13% of the worldwide desktop/laptop market.

What if Apple expands that market to 20%?

Cornering 80% of the $999 & higher desktop/laptop market?

Would be awesome when yearly worldwide shipments goes from 22.5 million to 55 million of 275 million annual worldwide laptops/desktops shipped.