▲ 285 points
back
231 comments
M1 and M2 actually are not produced on the exact same process node. M1 is N5 and M2 is N5P, an optimized version of N5.
I think Kuo might be misinterpreting the statement from TSMC regarding revenue from N3. The key is that they said it wont "substantially" contribute to revenue until 2023. Of course processors like M2 Pro/Max/Ultra wont generate the same amount of numbers like something more high volume like an iPhone and in the grand scheme of things can't represent a substantial contribution to TSMC revenue.
The fact is TSMC said they'll start N3 HVM in September. So they are producing something and we know Apple is expected to be the first customer for this node. It's too early for the A17 so either it's the M2 Pro/Max/Ultra or something new like the VR headset chip. Can someone see another possibility?
Apple still btw has to replace the Mac Pro with an Apple Silicon based model and their own deadline (2 years from first M1) is running out. It could make sense that they want to bring this one with a "bang" and claim the performance crown just to stick it to Intel :)
I would expect that Apple would push its goldenchild (iphone) onto the node first. Its a small chip which they can use as a pipecleaner, makign sure they can get the yields up and optimise the process before pushing a larger die onto the node.
They easily could have been allocating risk production on the iphone for the past couple of months, ready for the launch. Apple being like yes we will take lower yields for less cost.
I do not expect any company to announce production 3N product, until apple has had one out for atleast 6-12 months. Look how long it took the rest of the industry to move to 5N. I swear part of that reason was an exclusivity agreement with Apple, and it massively paid off for their CPUs. Having a node advantage is always massive in terms of price / performance / power matrix.
Starting HVM in Sept does not mean you get Revenue in Sept. It takes months before volume reached, testing, packaging done and shipped. TSMC isn't unusual in stating it they wont get revenue from N3 until 2023.
> something new like the VR headset chip
My bet is on this one.
> Apple still btw has to replace the Mac Pro with an Apple Silicon based model
At about the same price, I think 10x Mac Studios Ultra count as replacing much more than a single maxed-out Mac Pro. Though I do expect an Apple Silicon Mac Pro is still coming, I do not see any need for Apple to rush to meet some marketing deadline from almost 2 years ago.
They don't ,,have to'' be ready in 2 years. The M2 numbers from the N5P process were underwhelming, I wouldn't replace my M1 MacBook pro without seeing significantly superior performance / watt numbers, and happy to wait for the N3 process to be in production whatever it takes.
The Mac Studio is the obvious Mac Pro replacement. It's hard to do a fixed system and also charge much more than $5k for it, and Apple has failed over and over at any kind of non-fixed Mac Pro.
They might just leave it as is. The studio is a very capable machine at the high end, and if given a beefier GPU/CPU and _maybe_ more DRAM it probably replaces the bulk of the professional use cases.
This is cool!
At the same time, not being part of the Apple ecosystem, should I be worried about the closed nature of this. I have been using Linux for over two decades now, and Intel seems to be falling behind.
(I do realize Linux runs on the M1. But it's a mostly hobby projects, the GPU is not well supported, and the M1/M2 will never(?) be available with open H/W.)
I'm running Asahi Linux on the M2 and it's great. Drivers are not all complete yet, but it's awesome.
Eventually non-Apple laptops will be sold with silicon from this process node. You just won’t be the first to use it, which is fine.
Also, Asahi is getting closer to “generally usable” at an astounding pace, so who knows.
AMD is not behind at all. Have you seen the latest benchmarks?
Me too. I really wish I could buy a Samsung Galaxy Book Go 360 which is ARM and has amazing battery life, and install Ubuntu on it, but I don't think there's a known possible way to do so.
I really want a competent, high-end ARM Ubuntu laptop to happen. The Pinebook Pro has shitty specs and looks like shit with the 90s-sized inset bezels and 1080p screen.
Apple isn't going to somehow make 64 bit ARM in to something proprietary. Sure, they have their own special instructions for stuff like high performance x86 emulation, but aarch64 on Apple is only going to mean more stuff is optimized for ARM, which is good not only for Linux, but for other open source OSes like the BSDs.
GPU progress looks pretty great, here is the latest post:
There are other ARM providers, and personally I expect to see some of them ramping up significantly in the next couple years.
Qualcomm's taking another shot, for example. https://www.bloomberg.com/news/articles/2022-08-18/qualcomm-...
Good question
https://aws.amazon.com/pm/ec2-graviton/ is an indication that Amazon cares about linux support for the arm64 architecture. So the question is how much variance there is to the M1 relative to that.
x86 processors will be produced on the same nodes. Many ARM SoCs require binary blobs or otherwise closed source software, so they are not the best choice to run Linux on if you're approaching it from a longevity and stability perspective.
There is... another
What would be your concern? M1/M2 is just arm which Linux has run on for decades?
According to Wikipedia:
> The term "3 nanometer" has no relation to any actual physical feature (such as gate length, metal pitch or gate pitch) of the transistors.
I thought it at least maps to something physical. But it's just a marketing term.
It used to be the gate length. But eventually improvements were made that didn't neatly map to the gate length but still doubled transistor density, so they transitioned to just dividing the number by sqrt(2) each generation (since chips are two-dimensional making everything sqrt(2) smaller in each dimension increases density by factor 2).
Until 14nm the relation to transistor density was kind of true, all 14nm processes have similar density that is roughly double that of 22nm, which is roughly double that of 32nm. But now even that's meaningless and all they're doing is steps of factor sqrt(2) out of habit.
If M2 Pro is similar to M1 Pro (two M1s duct-taped together with very fancy duct tape), this is interesting because usually chips need to be significantly reworked for a newer process and this implies an M2 core complex will be printable both at 5nm and 3nm. It would be interesting to know how much of this is fabrication becoming more standardized and how much is Apple's core designs being flexible. If this is the latter, then Apple has a significant advantage beyond just saturating the most recent process node.
The M1 Pro was not two M1s duct taped together. Their core configurations do not share the same proportions (8+2 vs 4+4).
You may be thinking of the GPUs? Each step in M1 -> M1 Pro -> M1 Max -> M1 Ultra represents a doubling of GPU cores.
Or you may be thinking of the M1 Max and Ultra. The Ultra is nearly just two Maxes.
Regarding your point about flexibility, it’s hardly unprecedented for the same core to be used on different processes.
Apple has at times contracted Samsung and TSMC for the same SoC. Qualcomm just recently ported their flagship SoC from Samsung to TSMC. Even Intel backported Sunny Cove to 14nm. And of course there’s ARM.
There was some speculation in another thread not too long ago that the M2 design was originally 3nm, and was backported to 5nm after the fact.
I mean, Intel used a "tick-tock" model for a decade. (New microarchitecture, then die shrink, then new arch...) https://en.wikipedia.org/wiki/Tick%E2%80%93tock_model
I wonder if they'll also bump the M2 machines to 3nm silently, if the efficiency bump is minor? Apple previously split the A9 between TSMC and Samsung at two different node sizes, so it wouldn't be completely crazy.
Or perhaps they're content to leave the M2 as 5nm for easy performance gains in the M3 next year. It also has the advantage of keeping the cheapest machines off of the best node size, which is surely more expensive and more limited than 5nm.
I think you're thinking of M1 Ultra - which is 2x M1 Max on an interconnect.
M1, M1 Pro and M1 Max are separate dies.
There's no reason to assume that a 3nm and 5nm M2 core is identical in that way. It's probably similar to the changes Intel used to do for die shrinks when they were doing tick-tock.
you are mixing M1 Ultra which is two M1 Max taped together with M1 Pro which is a weaker variant of M1 Max.
The heading is just blatant speculation. It was written before Ming-Chi Kuo tweeted about M2 Pro being on enhanced 5nm. Once he said that, the article was updated to include his tweet - adding an element of uncertainty. Ming-Chi Kuo is a credible news source; cultofmac is an aggregator - they don't care if they're wrong and they optimize for eyeballs.
This post reeks of fanboy-ish excitement; there's nothing to see here.
So we'll see at least 1-2 years of Apple Silicon being at least one node ahead of competition. I am curious for how long will be Apple able to pull this lead off, and what the perf/watt will look like when (if?) AMD has node parity with Apple in the near future. Or when perhaps Intel uses TSMC as well, and the same process node.
I think this was Apple's game for a LONG time. They have led in mobile chips to the point where they are sometimes 2 years ahead of the competition.
They do this using their monopsony power (they will buy all the fab capacity at TSMC and/or Samsung, and well before competition is aiming to do so either).
Yea this is what I am wondering as well. If nobody else ends up switching to ARM in the laptop/desktop space and eventually AMD and Intel are making 5 or 3nm chips then surely this massive lead in power efficiency is going to close. At the current levels the new apple computers seem awesome - but if they are only 10-20% more efficient?
Can anyone speculate what will happen to Apple if for any reason TSM abruptly stops supplying chips? Will their revenue just drop by 80%?
Everybody except Intel and Samsung is screwed if TSMC stops making chips.
Apple (and the rest of the mobile industry) would try to move to using Samsung's fabs and Intel would go back to being the undisputed king on desktops, laptops, and servers.
I think TSMC has like 2-3 times the fab capacity that Samsung does right now for modern chips, so there would be a huge chip shortage.
Apple's $200 billion cash pile would come in handy when trying to buy fab capacity from Samsung so they might come out ahead of less cash-rich competitors.
There would be a significant hit to processor performance. Samsung fabbed the recent Snapdragon 8 Gen 1, which has comparable single core performance to the iPhone 7's A10 fusion chip.
You mean TSMC.
Well, they would be seriously hurt. However, does that matter when almost every tech company (including Qualcomm, MediaTek, AMD, Apple, ARM, Broadcom, Marvell, Nvidia, Intel, so forth) would also be harmed?
TSMC going down is basically the MAD (Mutually Assured Destruction) of tech companies. Kind of a single point of failure. Intel would probably weather it best but would still be hurt because they need TSMC for some products. Plus, well, in the event of TSMC's destruction (most likely by a Chinese invasion), Intel might raise prices considerably or even stop sale to consumers especially as their chips would now have major strategic value for government operations. NVIDIA might also survive by reviving older products which can be manufactured by Samsung in Korea, but same situation about the strategic value there, and getting chips from Korea to US might be difficult in such a conflict.
I wonder how many people have to collaborate to get a 3nm semiconductor out the door. TSMC has 65,000 employees. ASML has 32,000 employees and 5000 suppliers. The complexity of it all is unimaginable!
It is possible it is for a Mac Pro, but that would be just for bragging rights to have a halo processor that beats everything AMD and Intel have in terms of power and single threaded and multithreaded workloads. It all depends on how good the yields are on 3nm.
As an aside, I am not sure who a Mac Pro computer would be for at this stage. I wonder if they would design it to be able to slide into a rack mount. How many Mac workloads need a computer more powerful than a Mac Studio at this point? It was once you needed a pro machine to properly use photoshop or illustrator and now a MacBook Air is fine most of the time and a Mac Studio is perfectly capable of doing even the most demanding video editing. The only thing that comes to mind would be for render workloads for animation houses and for video game developers. Will the chassis design be tailored for that? Will it be a good value for for those types of customers who are not going to care what it looks like because it is going to be in a rack or in a corner?
So it seems like the M2 is really an "M1+" or "M1X", whereas the M2 Pro/Max/Ultra are really the second-generation Apple Silicon.
That's fine, in my opinion. M1 is still an amazing chip, and if that product class (MacBook Air, entry iMac, etc.) gets even marginal yearly revisions, that's still better than life was on Intel.
I know I'm being irrational about this, but for some reason this makes me lean toward getting an M1 Air or 13-inch Pro rather than an M2: it's like with the M2 performance gains are being squeezed out of the same (or similar enough) M1 process rather than changing the process significantly, at the cost of efficiency.
I'm curious what actually unifies the "MX" for some X. There are different chips in the series, and apparently they can even be on different-sized processes and keep the name
Anybody know more detail?
Is the node process size comparable across architectures?
"HVM" = High-Volume Manufacturing
Used in the comments here... I had to think for a minute to get it.
HVM is full production status, as opposed to early risk or testing.
Has this analyst ever been correct?
hey anyone at Apple or able to influence there:
my use cases for machines expand when they can address 128gb RAM and above.
everything between 32gb and 96gb is the same use cases, for me. nice to haves but doesn't dramatically change anything.
just in case anyone needed to know
excellent. i've been waiting for an M2 macbook pro
What will the industry do when it reached 1nm?