Could we see M3 before A17?

[theorist9]

As for M2, I'm not sure if TSMC increase fab capacity for 5nm, but I think it is unlikely for them to increase by much? So the M2 is likely facing the same constraint as the M1.

Just found some more info on this. According to a January 2023 Investor Earnings Call, TSMC says they expect to be generally under-capacity in 2023, but for N3, specifically, customer demand is expected to exceed supply:

https://forums.macrumors.com/threads/apples-huge-3nm-chip-orders-for-iphone-15-pro-and-m3-macs-helping-to-maintain-tsmcs-fab-utilization-rates.2383200/page-3?post=32023176#post-32023176

 

[Xiao_Xi]

I've no idea how much faster a 3 nm M3 would be. Based on https://www.anandtech.com/show/1735...n3e-in-2024-n2-in-2026-major-changes-incoming , the process improvement from N4P to N3B should give maybe 10% at equal power, but then there's also the improvement from the newer M3 architectiture, and I have no basis for estimating that. Also depends on how they clock it.

M2 (4P/4E) gets around 9000 [1] on multicore GB5, and M2 Pro (6P/4E) around 12000 [2]. M3 would have to improve 1.33x to reach the less powerful M2 Pro. Could it be feasible with TSMC N3?

If it isn't feasible, at what performance improvement could M3 start to get M2 Pro sales?

[1] https://browser.geekbench.com/v5/cpu/20840928
[2] https://browser.geekbench.com/v5/cpu/20839387

 

[Mago]

ASi Mac Pro is more concerned on soc scalability/interfacing than thermals (which is the only improvement n3 provides over n5+).

That's why a mature process has sense for Mac Pro SOC solution, if inFO-LSI represents a challenge with n5+ with new m3u likely more complicated and unpredictable.

So, unless something below radar is happening there won't be an n3 (neither m3) based Mac Pro, which likely to adopt for it's 2nd generation (m4) n3+.

 

[leman]

ASi Mac Pro is more concerned on soc scalability/interfacing than thermals (which is the only improvement n3 provides over n5+).

That's why a mature process has sense for Mac Pro SOC solution, if inFO-LSI represents a challenge with n5+ with new m3u likely more complicated and unpredictable.

So, unless something below radar is happening there won't be an n3 (neither m3) based Mac Pro, which likely to adopt for it's 2nd generation (m4) n3+.

This is a good argument. At the same time, Apple currently does not have any CPU or GPU tech that would let them compete in the high-end workstation market, unless they go very wide. An M2-Ultra Mac Pro would hardly be interesting and I do not that we will see Extreme version this iteration.

 

[dmccloud]

Let's suppose Apple does release an M3 Air on N3 sometime between now and WWDC. Rumors are that the 14"/16" MBP's won't get a chip update for about a year from when the M2's came out (in Jan). Wouldn't it ber a problem to have their flagship laptops linger so long without the newest tech? Might Apple thus do a mid-year refresh?

Apple may be able to compensate for this in an M2 Mac Pro by significantly increasing the CPU and GPU clocks (thus exceeding base M3 per-core performance). But for the M2 MBP's, the horse has left the barn.

I think that any product using a 3nm SoC would come first to a product that does not sell as much as the Air. Given that current 3nm wafer production at TSMC is less than 1/3 of 5nm production, there may not be enough working chips to go around (yields of usable chips always starts a bit lower, then improves as manufacturing processes get refined and adjusted). I could see 3nm coming to a new Mac Pro and possibly the Mac Studio before it hits either the MBA or 13" MBP (which are Apple's best selling laptops).

 

[theorist9]

I think that any product using a 3nm SoC would come first to a product that does not sell as much as the Air. Given that current 3nm wafer production at TSMC is less than 1/3 of 5nm production, there may not be enough working chips to go around (yields of usable chips always starts a bit lower, then improves as manufacturing processes get refined and adjusted). I could see 3nm coming to a new Mac Pro and possibly the Mac Studio before it hits either the MBA or 13" MBP (which are Apple's best selling laptops).

As to which comes first, there's the tradeoff between lower volume possibly favoring the M3 Ultra Mac Pro, and lower complexity possibly favoring the M3 Air.

 

[steve123]

Rumors are that the 14"/16" MBP's won't get a chip update for about a year from when the M2's came out (in Jan)

M2 MBP's were delayed due to supply issues. M3 MBP's will be released in Nov 2023 (if all goes according to plan).

 

[Yebubbleman]

I think it depends on whether M3 is going to be an A16 derivative or an A17 derivative. The A-series SoCs have, so far, been first to introduce the new cores. But, I'm not sure it necessarily has to be this way. It's just happened that A14 came out before M1 (albeit they both came out within weeks of each other) and that A15 came out before M2.

 

[leman]

Ah, the next generation Apple M3 Silicon CPU architecture. I'm afraid most people wouldn't understand the sheer brilliance of my insights on this subject, but I shall attempt to elucidate some of the more advanced concepts for the mere mortals among us. I find it quite trivial to predict the intricate details of this upcoming marvel of engineering, despite the fact that I am not privy to any insider information.

You see, it's quite obvious that the Apple M3 will leverage a groundbreaking 3D-stacked die topology, employing a radical heterogeneous architecture that combines the unprecedented prowess of RISC-V cores and cutting-edge neuromorphic computing. All of this will be accomplished while maintaining the most stringent power efficiency standards, of course. It is only natural that the M3 will utilize a 2nm EUV lithography process, providing a level of miniaturization hitherto unseen in the industry.

The M3's cache hierarchy will be a thing of beauty, comprising a sublime fusion of traditional SRAM and emerging memristor technologies. This will enable an unheard-of capacity for on-chip memory storage, effectively eradicating the memory wall that has long plagued the industry. In tandem with the new cache hierarchy, an innovative predictive pre-fetching mechanism will be employed to ensure that data is readily available to the processor cores, as if they were psychic.

I would have considered becoming a CPU Silicon Timing Correlation Engineer, but I must confess that the work would have been as intellectually stimulating as a restaurant kids' meal placemat crayon maze to someone of my vast intellect. Alas, I am left to spend my days contemplating the more profound mysteries of the universe, while mere mortals like yourselves can merely marvel at the wonders of technology that I so effortlessly decipher.

BingAI, is that you?

 

[Mago]

radical heterogeneous architecture that combines the unprecedented prowess of RISC-V

🤣 Good as sarcasm

 

[Gudi]

A17 obviously comes with the iPhone 15 Pro in Fall. M3 could well come earlier with any new Mac. And the performance increase would actually make a difference on a true multi-tasking OS, like macOS. The iPhone benefits most from longer battery life.

 

[Populus]

A17 obviously comes with the iPhone 15 Pro in Fall. M3 could well come earlier with any new Mac. And the performance increase would actually make a difference on a true multi-tasking OS, like macOS. The iPhone benefits most from longer battery life.

Yep, my thoughts as well.

On the Mac platform, the M3 will represent a big jump in performance and capabilities, like you said, thanks to macOS being a more versatile operating system. Sadly, on the iPad platform, the increase in performance won’t be as evident. I have now a new M2 iPad Pro thanks to a sale, and honestly I’d say it is as capable as an M1 iPad Pro. Yes, the iPad is blazing fast, but so it is the M1 iPad Pro. And, indeed, the efficiency gains on the iPhone will be welcomed, and maybe better machine learning capabilities.

Back to the topic, I think we could see the first M3 products short after the A17 unveil, between October and November. I don’t think it is likely that we’ll see M3 before the A17 but who knows.

 

[sunny5]

Digitimes (unreliable) reports that 15" MBA is releasing in April/May. Mark Gurman thinks it's releasing in Summer.

Why M3 could be released in Summer at WWDC:

• Assuming that Apple is skipping A16 cores for the M series, then Apple would have had extra resources to create M3 using A17. This could explain why M3 using A17 can come out before the iPhone 15 Pro. In "normal years" (which we haven't had due to covid, supply chain issues, 3nm delay), I would assume that the base M always comes out 1-2 months after A series.
• At around April of each year, Apple should already be starting to mass produce the next A series that will be launched in September. So the A17 design has long been finished and the building blocks were long ready for M3 to consume.
• Apple would probably like to launch a new Mac computer with a new M series. Launching M3 with MBA 15" would garner a lot of interest.
• If 15" MBA uses M2 in May, and then a few months later, all the other laptops get M3, it'd make the 15" MBA look pretty bad. Apple wouldn't want to do this because the 15" laptop market is by far, the biggest, which means the 15" MBA will likely become its top seller.
• Mark Gurman did report that M2 will be a shortlived generation, which I agree based on the timing of 3nm mass production, A17, and M3. Even so, at WWDC 2023, it would be one year after M2 already.

My other speculation is that the M2 Ultra is also canceled along with M2 Extreme. I believe Apple would like to go all in on M3. It's also possible that the base M, M Pro, and M Max get updated once a year, but the Ultra and Extreme are updated every 2 years due to the effort, complexity, and niche market.

Yes, A16 sucks and that's why M3 should be A17 based. But I dont think M series will announce first ahead of iPhone announcement. Apple's biggest interest is iPhone, not Mac. Beside, we still didnt get M2 Ultra so far.

 

[senttoschool]

Yes, A16 sucks and that's why M3 should be A17 based. But I dont think M series will announce first ahead of iPhone announcement. Apple's biggest interest is iPhone, not Mac. Beside, we still didnt get M2 Ultra so far.

Has nothing to do with "it sucks". It has everything to do with the fact that A16 was designed from the ground up to target the 5nm family. It can't be re-used for 3nm unless Apple spends a lot of money and resources to redesign it which makes no sense given that A17 is available.

What do you think Apple has been manufacturing on 3nm all this time? TSMC started volume production of 3nm back in December and most reports suggest that Apple is the only 3nm customer at the moment.

 

[Jorbanead]

Let's suppose Apple does release an M3 Air on N3 sometime between now and WWDC. Rumors are that the 14"/16" MBP's won't get a chip update for about a year from when the M2's came out (in Jan). Wouldn't it ber a problem to have their flagship laptops linger so long without the newest tech?

It’s been noted that the M2 MBPs were likely ready in late fall of 22’ and got delayed to January for unknown reasons. I wouldn’t be surprised if we get M3 MBPs this fall.

My rough M3 prediction schedule is:

M3 releases this spring. Air, iMac, and 13” MBP get updated. This fall 14/16” MBP get updated with M3 Pro/Max. This winter Mac Pro get updated with M3 Ultra. But they may still announce the Mac Pro at WWDC this year as they’ve done with past Mac Pro, iMac Pro, and Pro Display - It just may not ship until winter though.

Then Apple can maybe even keep that cycle going, with baseline macs updated in the spring, pro notebooks in the fall, and highest-end macs in the winter.

 

[Boil]

Yes, A16 sucks and that's why M3 should be A17 based. But I dont think M series will announce first ahead of iPhone announcement. Apple's biggest interest is iPhone, not Mac. Beside, we still didnt get M2 Ultra so far.

With the assorted issues delaying the roll-out of Apple silicon, I think Apple will be skipping any M2-series SoCs beyond what we already have (M2 / M2 Pro / M2 Max)...

Has nothing to do with "it sucks". It has everything to do with the fact that A16 was designed from the ground up to target the 5nm family. It can't be re-used for 3nm unless Apple spends a lot of money and resources to redesign it which makes no sense given that A17 is available.

What do you think Apple has been manufacturing on 3nm all this time? TSMC started volume production of 3nm back in December and most reports suggest that Apple is the only 3nm customer at the moment.

^^^ This...

TSMC has been making N3B wafers for Apple for the last two+ months; N3B is the process no one really wanted, the process Intel dipped out of; seems perfect for Apple to burn thru those wafers for M3 Ultra & M3 Extreme SoCs (or whatever Apple has planned) for the ASi Mac Pro...

The A17 SoCs for the latest & greatest iPhone this Fall will be on the N3E process, which will probably start cranking up next month...

 

[Jorbanead]

I think Apple will be skipping any M2-series SoCs beyond what we already have (M2 / M2 Pro / M2 Max)

This is what I think too.

3nm is right around the corner and could even come out this spring based on rumors. Why make an M2 Ultra when the majority of users who need that power likely already bought an M1 Ultra last year.

Now if it was M3 Ultra, that may be a big enough leap forward to entice users still on intel Mac Pro’s to switch over. But the GPU performance is going to have to be competitive to the AMD Mac Pro GPUs.

 

[senttoschool]

This is what I think too.

3nm is right around the corner and could even come out this spring based on rumors. Why make an M2 Ultra when the majority of users who need that power likely already bought an M1 Ultra last year.

Now if it was M3 Ultra, that may be a big enough leap forward to entice users still on intel Mac Pro’s to switch over. But the GPU performance is going to have to be competitive to the AMD Mac Pro GPUs.

In my original post at the top of this thread, I suggested that the M2 Ultra is canceled or it was never planned.

It's especially fishy given that Apple has already been making chips on 3nm, which suggests that the M3 release is imminent (WWDC?), which also means releasing that M2 Ultra after releasing M3 is weird. Unless Apple surprises us with M2 Ultra at WWDC and delaying M3 until Fall. But what have they been making on 3nm if they aren't making M3?

I doubt M2 Ultra will make Apple much money. It's more important to get the M3 out sooner than to wait for the M2 Ultra.

 

[thenewperson]

It's especially fishy given that Apple has already been making chips on 3nm, which suggests that the M3 release is imminent (WWDC?), which also means releasing that M2 Ultra after releasing M3 is weird.

Well, March (or April) is an opportunity - that’s when they released the M1 Ultra after all. But it’d be kind of a weird time to release the Mac Pro (but then January was a weird time to release the MBPs 🤷). Hopefully all the noise about the M2 ultra in the MP is nonsense and it’s just a plain Mac Studio update.

 

[Serqetry]

Hopefully all the noise about the M2 ultra in the MP is nonsense and it’s just a plain Mac Studio update.

That's what I've been guessing and saying all along. It would make way more sense if the M2 Ultra rumors/"leaks" were because there is an M2 Ultra in production and it's just going in a Mac Studio update before the Mac Pro shows up later with an M3 derivative of some kind. I don't believe for a second that the Mac Studio is dead because Apple can't make a Mac Pro that significantly out-performs it... that was the dumbest rumor ever, and it if were true it would just mean Apple should scrap the Mac Pro in favor of the Studio.

 

[Jorbanead]

it if were true it would just mean Apple should scrap the Mac Pro in favor of the Studio.

One thing that goes overlooked is the timeline of the 2019 Mac Pro.

Does anyone remember the mini press event Apple had where they essentially admitted they made a mistake with the trashcan, and promised to be working on something new? The trashcan came out in 2013 and it took them SIX years to come up with a new solution. To me this timeline was always fishy.

One thing that always stuck out to me was how Apple said it was going to take them some time to figure out what to do. Why did they need that long? Yes it’s a beautiful machine and apples cordless design was genious, but I think it took them that long because they were planning for apple silicon down the road and needed to solidify chip plans before release. If they do chip design at least 3 years in advanced (as they have said in interviews), surely by 2019 they at least had some sort of plan for the Mac Pro with apple silicon - even just the basic idea.

My suspicion is that the 2019 Mac Pro design was done in part with apple silicon in mind. That’s why they’re not doing a new design for the new model.

I call BS that their plan is to just stick ONE ultra chip in the Mac and call it a day. I suspect they’re planning to put multipule Ultra chips. I think that’s why the Mac Pro cooling is so strong - imagine slapping in 2 or 3 M3 Ultra chips in one Mac Pro.

Now we’re talking CPU performance that will blow intel out of the water, and GPU performance that can actually compete with Nvidia.

 

[theorist9]

One thing that goes overlooked is the timeline of the 2019 Mac Pro.

Does anyone remember the mini press event Apple had where they essentially admitted they made a mistake with the trashcan, and promised to be working on something new? The trashcan came out in 2013 and it took them SIX years to come up with a new solution. To me this timeline was always fishy.

One thing that always stuck out to me was how Apple said it was going to take them some time to figure out what to do. I think they mentioned how the trashcan wasn’t able to keep up with new technology, so they were building it for the future.

Let’s also keep in mind chip design, according to Apple, happens at least 3 years before release. So in 2019 when the Mac Pro came out, we know they were heavily working on M1 macs. My suspicion is that the 2019 Mac Pro design was done in part with apple silicon in mind. That’s why they’re not doing a new design for the new model.

I call BS that their plan is to just stick ONE ultra chip in the Mac and call it a day. I suspect they’re planning to put multipule Ultra chips.

Maybe it took them three years to realize the Trashcan was a mistake and then, once they did this, three more years to design and release a new model. Slow, yes. Implausible? I'm not so sure.

But, in any case, the MP was very much an Intel/AMD-focused design. It accommodated power-hungry CPU/GPU combos and double-height plug-in graphics cards, and had lots of slots for upgradeable RAM. Given this, I don't think they designed it with AS in mind.

Of course, that's not to say this very nice case design can't also work well for AS. It will just need some internal reworking.

As to whether Apple will offer more than just a simple Ultra (e.g., plug-in GPU modules, expandable RAM), that remains to be seen.

 

[Serqetry]

My suspicion is that the 2019 Mac Pro design was done in part with apple silicon in mind. That’s why they’re not doing a new design for the new model.

I kind of doubt that. Also there's no evidence they aren't doing a new design, the claim that they aren't changing the case is just a Mark Gurman hallucination.

I call BS that their plan is to just stick ONE ultra chip in the Mac and call it a day. I suspect they’re planning to put multipule Ultra chips. I think that’s why the Mac Pro cooling is so strong - imagine slapping in 2 or 3 M3 Ultra chips in one Mac Pro.

On this I agree 100%. The idea that it's just going to have one M2 Ultra in it is another super dumb Gurman hallucination.

I doubt Apple is waiting this long to release a new Mac Pro just to be completely lazy about it. They already have the perfect machine to house a single M2 Ultra and it's the Mac Studio.

 

[leman]

On this I agree 100%. The idea that it's just going to have one M2 Ultra in it is another super dumb Gurman hallucination.

I doubt Apple is waiting this long to release a new Mac Pro just to be completely lazy about it. They already have the perfect machine to house a single M2 Ultra and it's the Mac Studio.

There is no evidence that Apple has anything larger than an M2 Ultra in the current pipeline, and the recently published patents (I mean like stuff published a month ago) only describe a two-die system, albeit with a new on-chip network. So I'm sceptical about the perspective of an Extreme system in the near future.

There was a suggestion that the Mac Pro will feature an M2 Ultra, but running a much higher clock, which will make it competitive agains Threadripper and new Xeon WS series. That could definitely work. Another possibility is horizontal scaling via pluggable SoC boards, but that's difficult to pull off technically and has major consequences for the software model.

 

[senttoschool]

There was a suggestion that the Mac Pro will feature an M2 Ultra, but running a much higher clock, which will make it competitive agains Threadripper and new Xeon WS series. That could definitely work. Another possibility is horizontal scaling via pluggable SoC boards, but that's difficult to pull off technically and has major consequences for the software model.

I know you've talked about horizontal scaling before but I just don't buy it. The Mac Pro isn't a server. Any Apple Silicon Mac Pro will be used to run software similar to what your Max or Ultra chip will run but users expect it to run much faster or run much bigger models. If you scale horizontally, software won't run faster unless they're rewritten from the ground up to take advantage of horizontal scaling - which very few software can make use of.

There is very little financial incentive for Apple to rearchitect most of macOS and rewrite most of their software to use horizontal scaling just for a niche machine that is already dying in relevance.

Hence, I'm still on the side of 4x Max dies glued together for a Mac Pro chip. This chip won't be cheap in R&D cost and it won't be cheap to manufacture. This is why I'm speculating that Apple will create an Apple Silicon Cloud to expand the market of any "Extreme" chip.

Between these options: cancel the Mac Pro, use only the Ultra, glue 4x max together, and horizontally scale Ultra chips, I think the horizontal scaling is the least likely option.