TSMC on Track to Start 3nm Chip Production in Second Half of This Year

[HobeSoundDarryl]

they released the mac studio for more power, so i think the mini m1 (m2 in the future) is where the Mac Mini is at.

I Might be wrong, but I do not see the mini going beyond the base cpu.

You may just as likely be right.

I had the intention of buying a "loaded" Mac Mini M1 MAX as rumored ahead of the event and ended up buying Studio Ultra as a "Fat Mini" for about $1000 more than I intended to spend. However, if the intent was for Studio to "pull up" all buyers wanting something a bit more than base Mac Mini, it seems Intel Mini would have been discontinued then... much like iMac 27"

Instead, Intel Mini survived. I assume something will replace it. If not Mac Mini with Mx PRO then perhaps a junior version of Mac Studio with Mx PRO. If I had to bet, I'd bet on Mac Mini with M2 PRO, starting at about $500 below the "starting at" price of Studio: M2 Mac Mini, M2 Mac Mini with another core or two turned on, M2 PRO Mini filling the pricing gap, M2 Studio, etc.

Or maybe M2 turns out to basically be M1 PRO (or close enough anyway) and Apple keeps up to both M1 Minis around and rolls out M2 Mini up into that higher "starting at" price area of $1299-$1499 or so? If so, maybe PRO branding just goes away entirely: M2, MAX, ULTRA and whatever will go into Mac Pro: EXTREME? QUAD?

 

[munpip214]

Apple chipmaking partner TSMC says it will be ready to move its 3nm chip process to volume production in the second half of this year, putting it on track to supply Apple with the next-generation technology in 2023

Article title is misleading. Makes it sound like people will get 3nm this year which is clearly not the case.

 

[cgsnipinva]

I also ordered a Mac Studio Max with maxed RAM and 2 Terabytes to replace my Mac Mini. It is probably more power than I need for my use case, but given what I saw on performance, it should last me a while. I have a Mac Mini (2014) version that still works well.

What I am happy to see is that Apple has developed a structured and methodical approach towards their silicon development. The start with an M1 design that is broken out to a Pro, Max, and Ultra Version. This approach will most likely be repeated by the M2 design. Eventually all Macs will be on Apple Silicon.

I am really interested into seeing how my Studio performs.

 

[macduke]

Please be for the Mac and not the iPhone! I mean surely it can't be at that low of a scale to start. Maybe not even for the iPad unless it's the iPad Pro which has lower volume. If they put this in the MacBook Pro when they move to M2 Pro/Max this autumn I will be very tempted to upgrade then instead of waiting for M3 like I had originally planned. Didn't think we would get any 3nm this year and Apple's premium price can soak up a lot of early orders for it. 3nm will give even crazier battery life and performance than we've already seen and will make for a device I won't have to upgrade for the rest of the decade, except for maybe a battery replacement before it goes vintage. I just hope they make the notch smaller. Makes no sense to have it as big as it is now.

 

[Kpjoslee]

2nm on their own fabs? Yeah, I don’t see that happening and, historically, Intel has always overhyped when they would bring new nodes online. After they’ve appeased shareholders and gotten the hype train going, they come out six months later saying and announce that they are delaying rollout. And the analysts give them a pass. I’m not sure why anyone still believes anything Intel says.

And because technology isn’t static, I’m not sure why “node superiority” is important to anyone, other than to spec chasers. Perhaps it’s important to boosters after all the complete debacle that “10nm” has been for Intel. If Intel hits their targets, great for them. I don’t miss them one bit though.

Intel actually moved up schedule for 18A by 6 months from 1H 2025 to 2H 2024 so it seems like things are going extremely well for them, since it is first time ever they moved up the schedule instead of delay.

Intel having a very competitive node against TSMC is actually a really good news for Apple, since Apple couldn’t sold enough iPhone 13s due to chip shortages. Apple will not shy away from using Intel if they can get more volume.

 

[magbarn]

2nm on their own fabs? Yeah, I don’t see that happening and, historically, Intel has always overhyped when they would bring new nodes online. After they’ve appeased shareholders and gotten the hype train going, they come out six months later saying and announce that they are delaying rollout. And the analysts give them a pass. I’m not sure why anyone still believes anything Intel says.

And because technology isn’t static, I’m not sure why “node superiority” is important to anyone, other than to spec chasers. Perhaps it’s important to boosters after all the complete debacle that “10nm” has been for Intel. If Intel hits their targets, great for them. I don’t miss them one bit though.

Besides Intel, who else does leading edge nodes other than Samsung which with their bad power/performance ratio of Geforce 3XXX cards? If anything we should be cheering on the competition as I don't like that the whole world is basically dependent on TSMC for everything smaller than 10nm. Even Intel is going to use TSMC for some of their CPU's and all their GPU's. I don't like TSMC's monopoly at this point as since part of the semiconductor shortage can be blamed on everyone relying on them.

 

[ChrisA]

...allowing up to a 40-core CPU. In comparison, Apple's M1 chip has an 8-core CPU and the M1 Pro and M1 Max chips have 10-core CPUs.

Let's say you are a typical Mac user. At what point does adding more CPU cores stop making the computer faster?

For example a dual-core is better than a single core, 4 works better and 8 even better but does anyone notice a difference between 8 and 40?

Some tasks can be parallelized. Rendering video and many other media creation tasks are compute-bound, and so are many server workloads. But most users are media consumers

 

[Yebubbleman]

M3 in 2023 when we still haven't seen M2 in 2022? Yeah, seems unlikely.

 

[StrongArmmed]

So what happens after 1nm is reached? We get into sub-zero nm? Oh my

That is a great question...what is the next frontier in chip architecture??

 

[Mr. Dee]

So what happens after 1nm is reached? We get into sub-zero nm? Oh my

I suspect once they reach 2 NM it’s gonna be about optimization and efficiency. I wouldn’t be surprised if you start seeing like a 3 NMa 3 NMb, 3 NMc etc…

 

[StupidOpinion]

Let's say you are a typical Mac user. At what point does adding more CPU cores stop making the computer faster?

For example a dual-core is better than a single core, 4 works better and 8 even better but does anyone notice a difference between 8 and 40?

Some tasks can be parallelized. Rendering video and many other media creation tasks are compute-bound, and so are many server workloads. But most users are media consumers

majority of people on this site don't need anything more than the m1 basic, yet, they somehow justify spending several grand on the 16" macbook pro m1 max.

 

[_Spinn_]

Very exciting. I really wonder how much smaller we can actually go with these chips.

 

[criticasm]

TRANSISTOR DENSITY. That is how you describe chips on a level field. "Nanometres" & "Angstroms" are misleading marketing and utterly meaningless terms these days.

 

[Marx55]

If past promises are taken into account, we cannot trust Intel any more. Delays, delays and delays. And then not delivering to promises.

 

[Boil]

It sounds nice and exciting but the reality is that not even all Apple product lineup has the M1 chip.

The mac mini is still selling with an intel chip not to mention the mac pro. If they haven't been able to supply M chip on all their product yet I expect the M3 at the end of 2025. I mean mac products not ipads/iphones.

BTO MBP with the M1 Pro SoC is still on a two month wait, so the demand for M1 Pro SoCs seems to be too high to allow for using the M1 Pro SoC in the Mac mini at this time...

My guess: a Mac Mini with the M1 or M2 PRO chip will replace that Intel Mini within the next year, starting at about $1299-$1499.

Way too high of a price, the slightest upgrade would drive the price up enough where the M1 Max Mac Studio would be the better choice...

Apple needs to offer a full-die Mn Pro Mac mini for the current entry-price of the 2018 Intel Mac mini, $1099...

• M1 Pro SoC
• 10-core CPU (8P/2E)
• 16-core GPU
• 16-core Neural Engine
• 16GB LPDDR5 SDRAM
• 200GB/s UMA
• 512GB NVMe SSD
• Gigabit Ethernet (RJ45) port
• WiFi 6 / Bluetooth 5.0
• (4) Thunderbolt 4 / USB 4 (USB-C) ports (up to 40Gb/s)
• (2) USB 3.1 Gen 2 (USB-A) ports (up to 5Gb/s)
• HDMI 2.0 port
• 3.5mm audio out jack

 

[HobeSoundDarryl]

I appreciate the "too high price" perception but look at the Intel Mac Mini now: is it 3+ year old guts starting at $1099 with 8GB of memory? Is it hard to rationalize "latest & greatest" M2 PRO "starting at $1299-$1499" if that one is $1099 now? What is its price configured to 16GB RAM? $1299... for old Intel tech (8th gen!) that buyers know will be on the way to full deprecation not very far down the road.

Even if Mac mini with M2 PRO started at my highest guess- $1499, that's still a whopping $500 LESS than the starter price of Studio. There are plenty of computer buyers who see $500 as a HUGE amount of difference.

I myself bought a pretty loaded Studio Ultra for > $6K when expecting to a buy a loaded Mac Mini M1 MAX for around $5K... but, for many, breaking the 16GB RAM barrier and the strength of even a PRO chip would be PLENTY of Mac power. Some relative savings vs. a MAX Studio otherwise similarly equipped could buy them all of an alt-brand monitor or some of Apples monitor.

However, I LOVE your specs/pricing. I'm just a realist about modern Apple pricing... which seems to stress corporate profitability maximization much more than ever before. If Apple suspected $1299-$1499 was too close to $1999 Studio, I suspect they would jack Mac M2 Studio "starting at" to create more perceived value space for this Mac Mini Pro. I actually would NOT be surprised at M2 Studio "starting at" $2299-$2499 the way Apple seems to be going.

But I actually think $500-$700 IS a BIG gap to the average Joe and a Mini M2 Pro "starting at" even $1499 would sell well there. If anything, I think it would TAKE FROM Studio minimum config sales as a 'good enough' Mac option for some/many. For nearly 2 years, there have been countless passionate cases made for existing M1 Mini being plenty of computer for the everyman. If that's true, a PRO upgrade option would apparently be for those who think they want a bit more power than what is "plenty." That there is also something called MAX and ULTRA and EXTREME/QUAD may not matter to that type... much like an offer of small-medium-large tends to win a lot of medium orders.

 

[coolfactor]

TSMC is truly a marvel to behold.

 

[coolfactor]

I think the 3nm iPhone 15 will be my next iPhone. When will it run macOS lite? These chips are processing monsters!

iPadOS is already macOS Lite.

 

[McDaveH]

If true, it would be the second time in recent years that Apple has debuted new chip technology in an ‌iPad‌ before using it in its flagship smartphones. Apple first debuted the A14 Bionic chip, based on 5nm technology, in 2020's fourth-generation iPad Air.

Whether or not it goes down the same adoption route, Apple is expected to release the majority of its devices with 3nm chips fabricated by TSMC in 2023, including Macs with M3 chips and iPhone 15 models with A17 chips.

Releasing flagship silicon on higher end devices always made more sense to me but an M3 next year? M-series appears to be on a 2-year cycle. Even if they managed to get M2 on the new node later this year M2 Pro/Max/Ultra would be 2023. If anything this hunts at the M2 being delayed until next year.
Poor Apple, by trying to break free from the grip of a silicon vendor, they locked themselves into the grip of a silicon fabricator.

 

[name99]

"If true, it would be the second time in recent years that Apple has debuted new chip technology in an ‌iPad‌ before using it in its flagship smartphones. Apple first debuted the A14 Bionic chip, based on 5nm technology, in 2020's fourth-generation iPad Air."

Depends on what you consider "recent years" but the A5 debuted in the second generation iPad in early 2011, six months before it appeared in the iPhone 4S.

 

[Boil]

I appreciate the "too high price" perception but look at the Intel Mac Mini now: is it 3+ year old guts starting at $1099 with 8GB of memory? Is it hard to rationalize "latest & greatest" M2 PRO "starting at $1299-$1499" if that one is $1099 now? What is its price configured to 16GB RAM? $1299... for old Intel tech (8th gen!) that buyers know will be on the way to full deprecation not very far down the road.

Even if Mac mini with M2 PRO started at my highest guess- $1499, that's still a whopping $500 LESS than the starter price of Studio. There are plenty of computer buyers who see $500 as a HUGE amount of difference.

But then you upgrade that $1499 Mn Pro Mac mini to 32GB RAM & suddenly the base Mn Max Mac Studio is only $100 away, with better specs...

 

[name99]

TSMC is expecting products using its 2nm process to hit the market in 2026. Intel is claiming that its 20A and 18A (2nm and 1.8nm) nodes will enter volume production in 2024 (probably some 18A products will launch in 2025).

If intel does in fact stick to its schedule without delays (unlike the 14nm->10nm debacle), might Intel actually have node superiority in 2024-25?

TSMC seems to be slowing down. Moving to a 3 year cadence vs 2 years during 7->5->3.

TSMC runs its schedule based on engineering reality.
Intel runs its schedule based on marketing.
I know which one I would bet on...

"Node superiority" is a meaningless concept.
Even apart from the slipperiness of the dates (TSMC's dates will be based on when Apple ships Apple-volumes of chips, Intel's date will be, like Canon Lake, like Lakefield, now like Arc Alchemist, based on shipping a few thousand of some specialty product that no-one ever gets to see) what defines "superior"?

Intel think it's based on having transistors that give you high GHz (damn the power). Apple (and most of TSMC's customers) care a LOT about power.
Intel thinks its about having a density of transistors they can boast about; TSMC thinks it's about having a density of transistors in shipping products. (Intel's shipping products have an actual density that is one half to one third what the marketing claims. This still has not changed as of Alder Lake.)
etc etc

TSMC isn't even "really" slowing down. It's more that you associate TSMC's dates with Apple dates. But Apple dates are locked into September every year (at least so far, one day maybe that will change). And so if TSMC's "baseline" new nodes come out every 2.3 to 2.5yrs or so, that looks like an occasional missing a year -- because of how Apple's dates work, not because of anything on TSMC's side.

 

[name99]

Article title is misleading. Makes it sound like people will get 3nm this year which is clearly not the case.

They may. Many things are possible.
The A16 may ship on N4, while M2 ships on N3.
Hell, we may even see that M2 ships on N4, but M2 Pro, Max (which need separate masks anyway) ship on N3.
Apple have done things in the past equivalent to these sorts of splits.

It all depends on what sort of volumes TSMC believe they can hit on what date. They have announced volume in H2 2022; if that means December 2022, then iPhone can't make it, but there's no obvious reason M2, or just M2 Max+Pro can't make it.

Hell, maybe Apple would *like* to have Max and Pro operating at a slightly higher level than M2? Right now the general feel is that M1 is so good that you should only buy Max or Pro if you have really specialized, mostly GPU, needs (cf buying a Xeon is a specialized purchase). Apple might like to change the perception to something more like "M2 is good, but mid-range, like an i5, and if you want something in the same league but with more vroom, get a pro or max as more like an i7 or i9".

 

[name99]

Let's say you are a typical Mac user. At what point does adding more CPU cores stop making the computer faster?

For example a dual-core is better than a single core, 4 works better and 8 even better but does anyone notice a difference between 8 and 40?

Some tasks can be parallelized. Rendering video and many other media creation tasks are compute-bound, and so are many server workloads. But most users are media consumers

Nothing happens without programming. For 15 years or so, programming (by developers who care about and can use performance) has basically targeted 4 cores because that's what Intel, in their infinite wisdom, made the standard.
We're just at the edge of moving beyond that, and it will take another 10 yrs or so for improvements to really happen.

So yes, in a way if you cannot afford more than 4 cores and don't do much demanding, you will not miss much.
But it's only by flooding the market with more cores than most people "need" that developers (and youngsters who will become developers) think about performance in terms of *many* cores. This sort of short-sightedness is just one aspect of Intel's catastrophic effect on computing over the past 15 years or so.

 

[strawbale]

So what happens after 1nm is reached? We get into sub-zero nm? Oh my

0.7 nm by 2030: https://techbriefly.com/2021/10/01/asml-predicts-1-nm-processors-to-be-manufactured-by-2030/