Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
The memory in the m1, m2 and m3 is all on-chip. It is not a separate module you can buy from a 3rd party.
If this is true, except for the yield rate being less, the cost of screening extra RAM onto the chip is about $0.00

What BS

I despise price descrimination
 
  • Like
Reactions: compwiz1202
Apple doesn’t make the memory modules. Yes they are on-chip, but for example on M1 Apple uses SK hynix LPDDR4X memory chips. They are not on the silicon layer - they are on the package.

In other words: Apple buys them from a 3rd party and puts them on their chip package.
OK never mind my previous post then. Still feeling ripped off, regardless
 
While TSMC and Apple are partnering on one level, they are now bargaining on another level to get Apple to use more of that capacity. We know Apple keeps future plans hidden, but this article is exposing information about TSMC. Apple could slap an M3 into more stuff or sell M3 devices more cheaply, easily using that capacity - but only if the price is right. TSMC on the other hand, has incredibly high fixed costs and needs (from an accounting perspective) to absorb those costs over as many wafers as possible. TSMC is probably in a situation where it makes sense to sell additional wafers to Apple at a short term loss, as it's a better long-term financial decision.

I wouldn't doubt Apple is leaking this information, because it looks like they have much of this power and want to know how low TSMC can go.
Or it could be the exact opposite of that.
 
The quote from Digitimes mentions that currently the utilization rate for the N3 isn't even at 50%. There is no way they are doing 30M chips in March. The capacity is about half empty and only gets to half empty by the end the month.

Also not very likely getting 600 usable chips out of the wafer either. Might get low-mid 500's out of something as small as an A-series, but the M3 is likely substantially bigger ( drop off in units to mid 400 range , or lower). Any Pro/Max sized die is even bigger drops in volume per wafer. ( both bigger sizes and drops in yield percentages. )


About 25K wafers per month is indicative of something not solely iPhone. Not solely plain M3 either.

Digitimes numbers about max N3 wafer capacity are flakey at best. A little over a week ago they pegged it at 45K . now it is 50K (and the capacity running at 50%. So why would they be cranking from 45K to 50K when only running less than 30K. ???? Or these max wafer runs numbers are extremely soft and hand wavy. )

Apple doesn't particularly need more than 250-30K/Waf/mo for iPhone Pro + Mac this time of year and if they are the only buyers there isn't much of a point.
"The foundry will also grow the process output to 50,000-55,000 wafers monthly in March, with Apple being the main customer"

The way I read the article is that the aim is to get to potentially producing 30M chips per month beginning in March.

The article does not indicate how many serviceable chips have been produced since manufacturing began on December 29th 2022 until the end of February 2023, if my memory serves me correctly.
 
While TSMC and Apple are partnering on one level, they are now bargaining on another level to get Apple to use more of that capacity. We know Apple keeps future plans hidden, but this article is exposing information about TSMC. Apple could slap an M3 into more stuff or sell M3 devices more cheaply
I’m not sure what else apple could put their chips into? All their products now use some form of TSMC chip in them, and those that need to will transition over to 3nm over the next year or so. They can’t just slap an M3 into an iPhone, but the iPhone will surly use the new 3nm process.

Im not sure what incentives Apple has to lower prices? If it’s to get more people to buy products - how is this any different than any other year? That’s always been an option.

Keep in mind Apple has been working with TSMC for a long time.
 
I’m not sure what else apple could put their chips into? All their products now use some form of TSMC chip in them, and those that need to will transition over to 3nm over the next year or so. They can’t just slap an M3 into an iPhone, but the iPhone will surly use the new 3nm process.
They could use them in future iPad Airs, minis, a renewed Macbook line, an iPhone Ultra. Maybe a refocused Apple TV? Maybe value-base-models like the mac mini - where they can charge you for upgrades but still be seen as providing customers a great entry-level value - you name it.

M3 will probably remain very competitive in terms of arm-based alternatives, but that gap may (will?) close over time. They have the advantage now and for the foreseeable short-term.
 
No surprise apple is going all-in on N3 given lackluster 14 sales. I skipped 14 for the incremental nature of the update.
You should get used to that. All updates for phones are going to be incremental from here on out. They already have been for a few years. I have a 12Pro and see absolutely no reason to upgrade yet, and doubt I will for a few more years still.
 
The memory in the m1, m2 and m3 is all on-chip. It is not a separate module you can buy from a 3rd party.
Yes, they are. They are on-package, not on-chip. They are still off the shelf memory modules, just included on the same package as the processor right next to it so it's easier to run it faster.
 
Well just a few more of these node upgrades and WHAM smashed in the face by physics and the end of Moore's Theorem. Pesky quantum-tunneling electrons and big ass atoms!
 


Apple supplier TSMC is making strides to improve its production capacity for chips based on its cutting-edge 3-nanometer process technology, according to industry sources, which is expected to debut in this year's iPhone 15 Pro and upcoming MacBook models.

tsmc_semiconductor_chip_inspection_678x452.jpg

DigiTimes reports that TSMC's 5nm fabrication capacity began to loosen in November 2022 as a result of reduced orders from Apple, amongst other partners, with orders for iPhone chips alone having been slashed by 30%. However, the Taiwanese manufacturer has apparently been able to keep its utilization rate at 70% or higher thanks to Apple's thirst for 3nm:
Apple's upcoming iPhone 15 Pro models are expected to feature the A17 Bionic processor, Apple's first iPhone chip based on TSMC's first-generation 3nm process, also known as N3E.

The first-gen 3nm process is said to deliver a 35% power efficiency improvement over TSMC's 5nm-based N4 fabrication process, which was used to make the A16 Bionic chip for the iPhone 14 Pro and Pro Max. The N3 technology will also offer significantly improved performance compared to current chips manufactured on 5nm.

Apple's next-generation 13-inch and 15-inch MacBook Air models are both expected to be equipped with an M3 chip, which is also likely to be manufactured on the 3nm process for further performance and power efficiency improvements. Apple is also reportedly planning to release an updated version of the 13-inch MacBook Pro with an M3 chip. The M2 chip and its higher-end Pro and Max variants are built on TSMC's second-generation 5nm process.

Orders for new AI processors from Nvidia and AMD, as well as Apple's new iPhone chip, are expected to help TSMC avoid further fab utilization declines in the second quarter, DigiTimes' sources said.

Article Link: Apple's Huge 3nm Chip Orders for iPhone 15 Pro and M3 Macs Helping to Maintain TSMC's Fab Utilization Rates
So, Apple will remove the Pro label from the current M2 14/16 MBPs when the M3 15 inch MacBoor Air comes out, and return the label once the 14/16 go M3? Or will they call the M3 Air „MacBook Air Pro“, and the M3 13 inch MBP „MacBook ProPro“?
 
I’m guessing that Apple is expecting a surge in product demands in the near future.
 
The memory in the m1, m2 and m3 is all on-chip. It is not a separate module you can buy from a 3rd party.

Misleading statement, in the context of what you replied to. The RAM is not "on die."

The RAM are separate 3rd party chips that are soldered onto the SOC package. So a RAM surplus (or shortage) in the industry would absolutely affect Apple.

Apple_M1.jpg
 
They could use them in future iPad Airs, minis, a renewed Macbook line, an iPhone Ultra. Maybe a refocused Apple TV? Maybe value-base-models like the mac mini - where they can charge you for upgrades but still be seen as providing customers a great entry-level value - you name it.

M3 will probably remain very competitive in terms of arm-based alternatives, but that gap may (will?) close over time. They have the advantage now and for the foreseeable short-term.

Mac mini - already doing that
iPad Air - already doing that
iPad mini - uses A series
Apple TV - maybe!

iPhone Ultra - they would never. You probably don’t understand the thermal envelope, size, and battery requirements of M-series chips if you’re suggesting that. A smartphone physically cannot handle that chip.

Im not totally understanding what your point is here.
 
Apple is ready with M3/A17. Looks like 15" Air might launch at WWDC with M3
 
Here are excerpts from TSMC's Jan 2023 Earnings Call with investors.

Essentially they're saying overall chip demand is expected to slow in 2023 ("our gross margin faces challenges from lower capacity utilization due to semiconductor cyclicality…"), but demand for N3, specifically, is expected to exceed supply ("As our customers' demand for N3 exceeds our ability to supply, we expect the N3 to be fully utilized in 2023.").

I don't know if things have changed in the subsequent two months:
1678429709448.png


1678431357056.png


1678429720852.png

1678429725978.png


And here's a 2022 revenue summary, by process node. Note N3 is not there yet:


1678430147262.png


1678429954025.png
 
Last edited:
Mac mini - already doing that
iPad Air - already doing that
iPad mini - uses A series
Apple TV - maybe!

iPhone Ultra - they would never. You probably don’t understand the thermal envelope, size, and battery requirements of M-series chips if you’re suggesting that. A smartphone physically cannot handle that chip.

Im not totally understanding what your point is here.
Apologies, my point was Apple could adopt more chips on that node, I'm talking both A17 and M3 which I'm assuming will be on that node. That means Apple could have the freedom to stick an A17 into the next iteration of an iPhone SE.
 
Here are excerpts from TSMC's Jan 2023 Earnings Call with investors.

Essentially they're saying overall chip demand is expected to slow in 2023 ("our gross margin faces challenges from lower capacity utilization due to semiconductor cyclicality…"), but demand for N3, specifically, is expected to exceed supply ("As our customers' demand for N3 exceeds our ability to supply, we expect the N3 to be fully utilized in 2023.").

I don't know if things have changed in the subsequent two months.
I think a caveat here is that any company talking about expectations for future performance are talking about goals, and those could be pure fiction. Maybe TSMC was banking on some other third party jumping on that N3B train. We'll find out soon enough how that worked out.
 
Apologies, my point was Apple could adopt more chips on that node, I'm talking both A17 and M3 which I'm assuming will be on that node. That means Apple could have the freedom to stick an A17 into the next iteration of an iPhone SE.
Okay. Thanks for the clarification.

I’m still confused though and maybe I’m missing something. Apple moving to 3nm on these products isn’t anything unusual? They move to the new node on these products every generation.
 
I think a caveat here is that any company talking about expectations for future performance are talking about goals, and those could be pure fiction. Maybe TSMC was banking on some other third party jumping on that N3B train. We'll find out soon enough how that worked out.
Yeah, that's why I said "I don't know if things have changed in the subsequent two months." The next investor call should be in early April, and we'll be able to see hard data then. They can inflate what they expect, but can't inflate what their actual numbers are (well they could, but lying to your investors is bad behavior and perhaps illegal).
 
"The foundry will also grow the process output to 50,000-55,000 wafers monthly in March, with Apple being the main customer"

The way I read the article is that the aim is to get to potentially producing 30M chips per month beginning in March.


The article also says that utilization ( actual capacity used) is below 50% now and only getting to 50% by end of March. The 'potential' capacity doesn't matter if there is no orders actually flowing through it.

I suspect the digitimes folks are feeding sloppy information that it trickling through a "telephone game" filter that injects noise. There are two 'N3' lines. N3B and N3E. N3B is running now and has Apple as only big customer. Somewhere around June - August the N3E line will start to ramp up. Pretty good chance some of that unused capacity that TSMC is putting into place (i.e., installing new equipment and validating it for worthiness) is going to eventually feed into N3E and be distinct from what Apple is doing (at least for the products they are making right now. N3E isn't in high volume production ( but there are customers running 'at risk' batches to do late stage development and verification of their products) . If Apple is churning out 15-20K wafers a month it is likely N3B not N3E being made. )

The machines and tools are not substantially different for N3E versus N3B. N3B does more passes through the EUV fabrication machines because it has 20+ layers and N3E has fewer (below 20 ). N3E is going to take less time to 'bake' and cost less. TMSC can have the same number of EUV machines and increase the 'wafers / mo ' output by just saying going to do N3E instead of N3B 30% more of the time. ( N5 even less time in a EUV machine) . It really should be wafers of what process to really lock down a solid estimate of chips per month.


The article does not indicate how many serviceable chips have been produced since manufacturing began on December 29th 2022 until the end of February 2023, if my memory serves me correctly.

How many chips would largely depend upon what chips were being made. All chips are not the same size. The number of M2 Max sized chips can get out a 300mm wafer is substantially different than how many A14 sized chips can get out of the same wafer. The yield percentage also change on size. ( bigger die , bigger probability get hit with a random defect. )
 
The article also says that utilization ( actual capacity used) is below 50% now and only getting to 50% by end of March. The 'potential' capacity doesn't matter if there is no orders actually flowing through it.
It's all very confusing. We've got a lot of conflicting statements/rumors about N3:

1) TSMC faced overcapacity, and had to scale back: Reports that Intel cancelled most of its 2023 N3 orders with TSMC, causing TSMC to scale back N3 production to avoid being over-capacity (https://www.electronicsweekly.com/n...-orders-tsmc-pares-back-n3-expansion-2022-08/).

2) TSMC is undercapacity: TSMC's January 2023 earnings call with investors, in which they said they expect to be under-capacity for N3 and 2023, and thus unable to meet all customer orders. ("As our customers' demand for N3 exceeds our ability to supply, we expect the N3 to be fully utilized in 2023.")

3) TSMC is overcapacity: Digitimes (which probably should never be taken seriously) saying that TSMC is significantly over-capacity on N3.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.