Apple Likely to Be Only Major Device Maker With 3nm Chips in 2023 as Qualcomm Reportedly 'Caught In a Dilemma'

[quarkysg]

The manufacturing process is the most important and the most intellectually difficult part of chip development. Design is child's play and a rather insignificant part. Anyone can design a chip. It's not difficult asking your manager what their goals are and what your budget is.

It probably is not as simple as you make it out to be. You are essentially saying that an architect's and interior designer's role is not as important as that played by the contractor, or maybe the manufacturer's of the components used by the contractor, when it comes to building/renovation your home.

Using the world most advanced tools means nothing if you do not know what to do with those tools. A master artisan would create a masterpiece with just a pencil and a piece of paper, compared to me drawing with the best palletes of colours and canvas.

As usual, the sums of the parts should always be greater than it's individual components.

 

[MacsAre1]

The PowerPC G5 analogy is still bad.

It's not really intended to be an analogy. It's historical proof that Apple wasn't always able to stay in the lead with their "own" chips, since that era was in more similar to the ARM era than the Intel era. So I'm just saying that's not a good generalization to make. And as someone else pointed out, the G5 was way ahead of anything Intel had at that time... So the same tech they touted as an advantage was also an obstacle.

Yes, Apple has a lot more control this time. And that is more likely to help avoid the issues of the past. I believe the failure to get the G5 on a laptop was one of the driving factors to switch to Intel. But Macs are back to using chips with different instruction sets than the vast majority of other PCs, and the chip, and it's performance, is again one of the Mac's differentiating features.

 

[Fandroid killer]

Apple will always be ahead of the game when it comes to advanced chips.

You obviously never heard of intel nor of their arrogance that led to their failure in the chip space

 

[iManic]

Err no. Mediatek shipped a N4 Dimensity 9000 in May 2022. Apple didn't get to N4 until September A16.
MediaTek is on their 2nd N4 chip Dimensity 9200 now (and 8200 ) . Apple is still on their first.

Isn't that mostly if not all down to model specific release windows? Its not like Apple or Mediatek had a hand in developing the node change. Looks like it just fell favorably on Mediateks release window. Also from a quick look it looks like the Mediatek 9000 is on the A13/A14 performance level and the A16 looks way ahead?

 

[deconstruct60]

Isn't that mostly if not all down to model specific release windows?

Partially. There are two ways to move from N5 to N4. One library is a minimalistic change but doesn't full get all out of N4 than can be gotten. It is a faster path to market, but leaning more heavily on the common subset of "5nm family" design rules. . The second is a library that requires modest redesign work and gets a better 'grip' on the modest shrink possible with N4.

It looks like MediaTek did both. 9000 quickly. 9200 to optimize on N4. ( I couldn't find public sources on die size. If they are about the same that is probably what happened. If 9200 is much bigger then maybe they 'paniced' on a N3 abort. )

It is up in the air as to what Apple did. Some rumors peg it as they 'fumbled' the GPU design and have to quickly recover to a less ambitious , alternative design resuing maximum old 5nm work. That's is suggestive that they used that first option. The even bigger die size of A16 over A15 even though suppose to be using N4's modest shrink somewhat backs that up. ( Apple added more so it should get bigger , but there should have been a bit of an offset if optimally used N4 .)

Apple did add a fair amount of stuff though so perhaps N4 is helping control the size. ( so maybe they did use the second option and slightly more time.)

But a decent chance that Apple is only using one of those two paths. The only thing N4 is going to be the A16. (maybe a watch SoC die later if looking for lower costs on an 'old node. )

That said, different customers having different release windows is exactly why it would almost grossly irresponsible for TSMC to hide their advance tech from all but one customer. There shouldn't even be enough money Apple could pay to even try to bribe TSMC into doing something that dumb. Because it is just that dumb (way to run a fab with dozens of customers).

Before Huawei got excluded from access to TSMC advance nodes via the trade wars , they too were a visible early tech access partner with TSMC. If the trade war restrictions were not keeping them out they'd likely still be there.


[ Apple might pay enough money far in advance to get a substantive amount of minimal wafer starts on an new node. Help pay for the volume production base costs and get a slice of the action. But pay to block everyone one else out for a substantial amount of time or that it always lands in Q2 in time for iPhone preproduction.

That minimal wafer start volume access isn't access to the new tech. Those are really two different things. It is more so the breadth of the production line ; not that whole thing. ]

Its not like Apple or Mediatek had a hand in developing the node change.

Customers do have limited input into the node specifics. They don't completely run the show but there is influence. Back in pre-Covid 2019:


" ...
Therefore, it is not surprising to hear the annoucement that development of TSMC’s 3nm node is well underway, something the company publicly confirmed last week. As it appears, the manufacturing technology is out of its pathfinding mode and TSMC has already started engaging with early customers.

“On N3, the technology development progress is going well, and we are already engaging with the early customers on the technology definition,” said C.C. Wei, CEO and co-chairman of TSMC, in a conference call with investors and financial analysts. “We expect our 3-nanometer technology to further extend our leadership position well into the future.”

..."

TSMC: 3nm EUV Development Progress Going Well, Early Customers Engaged

www.anandtech.com

In 2021

"... said Mr. Wei. "Our N3 technology development is on track with good progress. We continue to see a much higher level of customer engagement for both HPC and smartphone applications at N3 as compared with N5 and N7."
..."
https://www.anandtech.com/show/16639/tsmc-update-2nm-in-development-3nm-4nm-on-track-for-2022



The fallacy that tends to get promoted on these macrumors forums is that Apple is exclusively the only ones providing input (or at least the 'relevant and significant' input ) and that TSMC is just some kind of puppet that Apple manipulates. Apple's money buys them TSMC repressing their other customers ability to do anything advanced.

Pretty good chance that if Apple is the only customer left standing on this "a bridge too far" version of N3 (N3B) is partially because Apple advocated for it. And many others didn't (and why they were quick to 'quit' and move on to something else) which is part of the reason N3E release is being pulled forward.

Looks like it just fell favorably on Mediateks release window. Also from a quick look it looks like the Mediatek 9000 is on the A13/A14 performance level and the A16 looks way ahead?

MediaTek is primarily just using the Arm baseline core library. It isn't going to jump radically. But it also is not all that bad either. It is affordable. If MediaTek built an expensive SoC that nobody bought how would that be a good outcome?

MediaTek is 'walk' up the ladder to higher priced SoC solutions. (helps that Huawei is in the penalty box and Samsung is stumbling. ). It isn't going to come in one big jump. Apple is selling A14 phones this year so it can't be that bad.

 

[High IQ Person]

And still, Apple’s design net a greater performance per watt than anything else TSMC manufactures. Better yet, Apple’s using an instruction set that other companies ALSO use, so there’s no special benefit there either. The biggest difference is indeed who designed the solution.

Maybe Apple’s the only company sending the order over to TSMC WITHOUT anchovies? Everyone else COULD leave off the anchovies, but for whatever reason, their designs still include anchovies. And mayonnaise.

This is only true when you're comparing SoCs from different nodes.
AMD just announced a 4nm process that's more efficient than the M1/M2.
Again, "design" is just a marketing gimmick. The fabbing process is more important.

 

[quarkysg]

This is only true when you're comparing SoCs from different nodes.
AMD just announced a 4nm process that's more efficient than the M1/M2.
Again, "design" is just a marketing gimmick. The fabbing process is more important.

I have to disagree.

Both should be equally important, or I would say a superior design may work around a less efficient manufacturing process.

We do not want to have a situation where CPU vendors rely solely on manufacturing processes to improve. That is why Intel's tick-tock process is good while it lasted.

x86's u-arch design (for both Intel and AMD) seems to rely on high clock speed to achieve their claimed performance, compared to Apple's AS u-arch. Of course the ISA have something to do with it, but if you compare Apple's ARM SoC to those from ARM SoCs, you will find that Apple's SoCs are better all around. So design is important as well.

Design thinking is a thing now for a reason. Design is to solve problems. TSMC's process solves the micro wiring problem. CPU designer solves the code execution problem, as efficient as they can make it. Better design will produce better product all around.

As usual, the sum should always be greater that it's individual parts.

 

[dgdosen]

If you don’t care about being first you can save a lot of R&D money.

If you're spending R&D money, you might want to work closely with your vendors in a mutually beneficial manner (within the law) and exploit that advantage. It also puts Apple in a position to grab 'fire sale' wafers as other companies (Intel) stumble in their efforts.

Apple was (probably) able to scoop up cheap 5nm/M1s because of Huawei. Maybe they can do the same in the 3nm/M3 space. I'm sure TSMC would rather ship to their backup customer compared to having new nodes go idle. Fortune favors the prepared.

 

[_Spinn_]

Apple is ahead in chips yet again. If this year turns out to be really bad financially for many people it will probably result in fewer phone sales. It might financially make sense for Qualcomm to hold off a year and see. Apple could use this as an opportunity to get further ahead.

 

[hugodrax]

Apple's decision to buy Intrinsity and PA Semi and bring processor design in-house was really smart. It's given them a huge advantage over the last decade and will probably continue to do so.

I heard the whole area suffered braindrain recently. Will be interesting to see what happens down the line.