Apple Orders 4nm Chip Production for Next-Generation Macs

[cmaier]

Previous macrumors articles indicated that 4nm would be used for the A16 chip in iPhone 14. I’m glad this article sort of corrects that.

What seems most likely: 5nm enhanced for upcoming iPhone. Maybe 4nm (which is really just a minor update to 5nm since you can use the same design) for upcoming macs late 2021/early 2022. And 3nm for iPhone 14 in late 2022. The iPhone 14 will not be using 4nm unless something goes horribly wrong.

While it’s possible you can use the same design, I haven’t seen any evidence of that? If the metal layer thicknesses are different, or the doping of the semiconductor is different, you would have to change your design. Is there some information out there about the 4nm process?

 

[BornAgainMac]

I thought the jump would be from 5nm to 3nm from earlier rumors.

 

[kb923689]

Meanwhile Intel has been struggling with 14nm chips since 2015.

That's what happens when a company gets too comfortable. Apple better watch out. The iPhones are stale lately.

 

[joecomo]

How LOW can you go?
3nm?
2nm?
1nm?

You might Discover the GOD particle in all this.

once you use the word "GOD" particle your out of science - that's why your quoted video is fittingly named "beyond science"

 

[jonnyb098]

This is more important for the Mac , as nearly none of the available iOS apps have fully utilized the power of even the A10 from 4 years ago.

 

[adamw]

Great to see them going to 4nm so soon after the M1 5nm launch! Keep those chips cooler and with better performance than Intel's at a much lower power draw.

 

[trip1ex]

we getting close to the hard limit of silicon.

 

[cmaier]

we getting close to the hard limit of silicon.

No we aren’t.

 

[joep99]

sadly, intel seems happy to continue producing x86 legacy, while rest of industry accelerates towards RISC architectures, AI instruction sets, and price / watt

intel would do better at this point to manufacture potato chips

 

[averagenerd81]

How LOW can you go?
3nm?
2nm?
1nm?

You might Discover the GOD particle in all this.

Best part about the "GOD" particle is why they called it that in the first place.

 

[jz0309]

Assuming announcement/release of new Macs around WWDC, those chips are already out of the fab by now.
Also, there were articles a month or so ago in semiconductor news area that Apple secured chip supply from TSMC all the way through 2024 ...
Semiconductor/chips have a very long development/prototype/production timeline that spans several years depending on complexity ...

 

[rpmurray]

Soon means just that, soon. There is no hidden meaning to the word soon.

"Soon" spelled backwards is "Noos", which is Finnish for "sometime before the end of the century".

 

[vadimyuryev]

4nm M2 chip going into new $1099 MacBook Air and $1299 MacBook Pro (maybe base Mac Mini) in Q1 2022. Calling it early.

 

[eoblaed]

Going to have to start measuring in picometers (pm), soon.

 

[ArPe]

Digitimes 🤦‍♂️ nothing to see here.

Digitimes about as reliable as Epoch Times which was a front for the Falun Gong death cult but that FG is now finished and the modern FG is an intel op of the c c p.

 

[cmaier]

Going to have to start measuring in picometers (pm), soon.

Nope. Angstroms.

 

[MikeyUnix]

Aye but they will have their 7nm soon and Apple will be able to buy capacity there and make even more chips without relying on one supplier. And Intel 7nm != TSMC 7nm.

Soon? You must be joking. They are still only commenting about delivery of 7nm in 2022 or 2023 and the latter is more likely judging by previous targets.

By then TSMC will be established at 4nm and working towards 3nm

 

[MikeyUnix]

Based on my experience, they probably get 20% over Intel (much less over AMD) from better physical design, 20% over both for RISC vs. CISC, 20% over Intel from the process advantage, 20% from micro architecture, and an amount I can’t quite put my finger on from architecture (e.g. UMA).

We did many experiments back in the day and found that our physical designs put us at least 20% ahead of Intel’s physical designs (e.g. choosing where each transistor and wire goes, how big they are, what shape they have, etc.), and the people I know at Apple are well aware of those techniques. And RISC vs. CISC advantage has been clear for years. And each time we did a process shrink we got, on average, 20% speed/watt improvement.

Weirdly, everything is always either 2% or 20% when you look at these things.

Difficult to base projections on AMD as TSMC produce a lot of their chips and at present Ryzen still on x86-64 architecture, albeit a while since I've checked.

 

[Doomtomb]

Mac Pro chips

 

[MikeyUnix]

Apple has booked the initial production capacity of 4nm chips with long-time supplier TSMC for its next-generation Apple silicon, according to industry sources cited in a new report today from DigiTimes.

From today's report:
The latest Apple silicon, the M1 chip, is the first of its kind in the industry based on the 5nm process. The A14 Bionic chip in the iPad Air and iPhone 12 lineup is also based on the 5nm process. According to the report, Apple is already looking to the 4nm chip process for its next-generation Apple silicon.

A timeframe for when these new 4nm chips will debut isn't provided, but DigiTimes does report that TSMC will move to volume production of the new process in Q4 of 2021, ahead of the previously set 2022 timeframe. Additionally, Apple plans to use an enhanced version of the 5nm process for the A15 chip in the iPhone 13, with production set to get underway by the end of May.

The smaller process reduces the chips' actual footprint and provides better efficiency and performance. Apple's expected to launch multiple new Macs this year with more powerful Apple silicon chips; however, there's no indication that any will be based on the 4nm process.

Article Link: Apple Orders 4nm Chip Production for Next-Generation Macs

If correct, will this delay the expected iMac range to Autumn 2021? Or will it herald the base iMac range first , followed by next release of high end iMacs as and if chip designs and 4nm are ready.

It might not help their marketing if users know a 4nm chip is in the offing, depending of course on the specifications of the 4nm chip as yet we don't know for sure about any derivative of the M1 or M2 or whatever Apple will call it.

 

[dannyyankou]

Interesting. I would’ve thought 3nm was next. But not complaining, it’s still far ahead of where Intel is lol

 

[George Dawes]

One word ;

 

[oosamon]

System on an Atom(TM).

 

[name99]

Intel and TSMC use different terminology, so TSMC's 7nm is said to be roughly equivalent to Intel's 10nm. It's all very confusing. Best treat node names just as names, with very little practical significance behind the numbers.

Your statement is very misleading.
Multiple chips (Apple, Huawei, Qualcomm) shipped with densities close (~90%+) of the peak density TSMC claimed for 7nm.

5nm is less definitive. We don't have good numbers for a Huawei chip, and everything 5nm from QC so far is on SS.
The Apple density for A14 is much lower than expected, and some have claimed this shows a serious flaw in TSMC 5nm. To me that's far too early to say. The A14 (and M1) were clearly rush jobs where THE priority was getting an Apple Silicon mac chip working (so the changes necessary for x86 support, hypervisor support and suchlike). Everything else was subservient to this goal. The core design appears to be substantially the same as the A13, the GPU even more similar. Which suggests to me they likewise did not have time to engage in much density optimization...
Let's wait for the A15 (or a Mediatek chip, or some other data point)...

Nothing Intel ships (even Tremonts/Atoms) gets better than 50% of the peak density Intel claims, and the desktop cores get about 35%. You might reasonably then ask "why did Intel bend themselves into pretzels chasing a density that has no effect on their products and exists purely for marketing purposes?" Well, why indeed? Nothing Intel has done since about 2014 has made any sense to me -- one boneheaded decision after another.

 

[name99]

It will (should) give you lower currents needed to flip a bit and shorter signal paths allowing for lower voltages (and to reduce the pipeline while maintaining the same clock).

One could off course extrapolate the gains by going back to an A10 but that would also be guesswork.


The point I was trying to make is that none of the items can make up the difference on it's own.

Apple's design team being 4 times or "just" 2 times better then Intel/AMD just doesn't sound plausible so it must be a combination of the above and apart from the ARM vs AMD64 everything can be made up relatively quickly, and that doesn't even comes into play for other ARM based CPUs.

Have you closely examined the micro-architecture of the M1? I have.
Yes, it is COMPLETELY plausible that Apple's designers are 4x better than Intel/AMD. Every aspect of the CPU you examine is just packed with good clever ideas -- to the extent that it's difficult to figure out critical parameters of the design because it's hard to create probe kernels that actually stall the machine! Every time you think this code stream will cause this time of limit, you see something else that allows the code stream to keep going. Apple's team is constantly rethinking how to do things better, as you see in the patent stream where you can read a patent from 2012 and think "nice idea, dumb implementation" then see a patent in 2016 and think "ah, yes, that's the RIGHT way to do it, that's what I would have done" then see a patent from 2019 and think "DAMN! I would never have thought of that" -- same problem, just engineers keep thinking "how can we do it better" rather than "aggh, it works, good enough".

Meanwhile at Intel the last serious core redesign was Nehalem (or if you want to be charitable, Sandy Bridge). Since then it's just tweaks.
Apple is substantially redesigning their cores every year, with a teardown and total redesign every four years (7/8/9/10 | 11/12/13/14 -- should get another reset and a rethinking with 15 this year 🎉) while Intel is apparently terrified to ever change the guts of what they have.