Looking forward to to scrolling through Twitter on my 3nm iPhone!
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
TSMC on Track to Start 3nm Chip Production in Second Half of This Year
- Thread starter MacRumors
- Start date
- Sort by reaction score
"Apple first debuted the A14 Bionic chip, based on 5nm technology, in 2020's fourth-generation iPad Air."
Yes... the A14 iPads came out before the iPhones.. by a week or two. That hardly amounts to a precedent for debuting new processors on the iPad.
Yes... the A14 iPads came out before the iPhones.. by a week or two. That hardly amounts to a precedent for debuting new processors on the iPad.
If Elon buys all of Twitter, you might have to get a Tesla phone, which can only be charged in a Tesla automobile or from a Tesla Powerwall...! ;^p
HobeSoundDarryl
macrumors G5
I understand. But base model is $500 LESS at $1499. "starting at" vs. "starting at" resonates with average Joe. And Joes who can be more influenced on specs get upsold to giving Apple another $100+ for a bit more power to lots more power. I offer myself as an example: very strong expectations of paying about $5K for a loaded Mac Mini Max... ended up upselling myself to more than $6K for a not-quite loaded Mac Studio Ultra.
And while I suspect $1499, maybe Apple DOES surprise and make it $1399 or possibly even $1299. I just don't think so by doing some simple math around 14" MBpro pricing.
If Apple worries about that 32GB config getting too close, I would expect M2 MAX Mac Studio "starting at" $2299-$2499 instead of hanging at the introductory "low(?)" price of $1999. Why would I think that? Again, let 14" MBpro pricing be a little guide:
- MBpro 14 configured like baseline Studio is $2899. Difference (or value for the parts NOT included with Studio: screen, keyboard, trackpad, etc): -$900
- MBpro 14 configured for this hypothetical Mac Mini Pro 16GB: $1999, subtracting the rest at -$900 would imply $1099, but I just don't think Apple would -$900 for the rest down at that level.
It would not be hard to convince me that $1299 could be real but I have near zero expectation of less than that, given that old Intel Mac mini at 8GB is already $1099 in Apple pricing. Give that Mac 16GB at Apple RAM pricing and land at $1299 right now. I strongly doubt Apple makes a special M2 PRO with only 8GB for this ONE Mac, so to align existing Intel and this hypothetical Mac at 16GB RAM each seems to demand "starting at $1299" at the low. I just don't see Apple offering "more" & "latest & greatest" M-power for less than 3+ years old Intel with the same core specs.
However, I'll hope right with anyone interested. I'd love to see $999 Mac Mini with M2 PRO base. Maybe Apple will show less "because we can" greed or simply opt to be more aggressive with this one for some reason. My guess: M1 8GB Mac mini stays at $699, M2 16GB based Mac mini at $899-$999, M2 PRO 16GB Mac mini at $1299-$1499. However, I neither work for Apple, nor stayed at a Holiday Inn last night. One could make a great case that there will NEVER be a M2 PRO in Mac Mini.
Last edited:
Simple math around the 14" MBP can be used right now ; any M1 Max 14" MBP config will be $900 more than a Mac Studio with equal specs (CPU/GPU/RAM/SSD)...
By that logic, and using the 14" MBP pricing (with the $900 differential), a double-binned M1 Pro with 16GB RAM & a 512GB SSD, in a Mac Studio chassis, that would be $1099...
Maybe the Mac mini does hold as the entry-level Mac, Mn SoCs only; and the Mac Studio expands its low-end to include Mn Pro SoCs...?
HobeSoundDarryl
macrumors G5
Yes, but Apple isn't married to that as a rule. Thus, my simple math adapts as pricing gets to the low. Why? because it has to be reconciled with old Intel (tech) Mac mini with 8GB at $1099 right now... so new Mac Mini PRO with 16GB is probably NOT going to be that low at $1099.
Again, I'll hope right with you. I'd love to see Mac Mini M2 PRO for $999. I simply have 0% expectation of that, near 0% expectation of $1099 or $1199 given Intel Mac Mini 8GB RAM is $1099. Upgrade (old 8th gen) Intel Mac Mini to 16GB RAM and it's $1299. Thus, my guess starts at that- $1299- as the LOW.
You could very well be right. Or I could. Or neither of us. I'll hope you are more right than me on this. If someone wants to throw in their bid at $899 for Mini M2 PRO, I'll hope THEY are right. $799? $799? Anyone?
The new Mac mini is almost certainly coming
I have been going over the numbers for a Mn Pro Mac mini, and I will finally admit defeat...! ;^p At this point it seems the only logical solution might be for the Mac mini to remain the domain of the base Mn SoC, and the Mn Pro SoC is added as a low-end option for the Mac Studio chassis...? A...
forums.macrumors.com
Intel is claiming performance per watt leadership on its 1.8 nm node. Lunar Lake and beyond.
The guys at Semi Wiki who have been heavily anti-Intel in recent years and who know what they’re talking about when it comes to semiconductors have this to say:
“we believe Intel has been able to significantly accelerate their process development at a time when the foundries are struggling. Although we don’t expect Intel to regain the density lead over the time period studied [2022-2026], we do believe they could retake the performance lead. We should get another good read on progress by the end of 2022 when we see whether Intel 4nm comes out on time.”
Can Intel Catch TSMC in 2025? - Semiwiki
At the ISS conference held from April 4th through 6th I presented on who I thought would have the leading logic technology in 2025. The following is a write up of that presentation. ISS was a virtual conference in 2021 and I presented on who currently had logic leadership and declared TSMC the...
semiwiki.com
Not sure whether iPhone 14 Pro will end up having 3nm processor. Either way there will be some improvements.
There is an implicit assumption there that the "processor" is a CPU processor. Apple should be getting ready to ramp a cellular modem by end of the year also. Testing out modems in iPad Airs would be lower risk than testing them out in a super duper high premium iPhone. Especially if there are quirks to work out in the field.
At least for the "digital compute" part of the modem chipset , Apple is probably going to need to out compete Qualcomm on process node to beat them on overall performance and efficiency. And if not carrying around whole baggage of a fully flushed out SoC then it can be a small die that fit quite well on early volume TSMC N3.
At some point , Apple's. $1B buy of a celluar modem needs to ship. It has been years since sank the $1B in. Sunk costs are even higher now after producing little to no revenue.
Probably not bootcamp Windows per se.
The Intel Mini 2018 is certified for VMWare.
Update on ESXi on Apple Mac Mini 2018 & Mac Pro 2019
Although there has not been any news in some time regarding the support for ESXi on the latest Apple Mac Mini 2018 and the recently released Apple Mac Pro 2019, there has definitely been work happe…
williamlam.com
There are many 10's of thousands mini's deployed as service servers on the internet. (e.g., Minicoloc/MacStadium, In-house set ups. Amazon has some , etc. More than decent chance XCode Cloud has a sizable number also although not using VMWare. ) . "racking them up on their sides " is a use case explicitly mentions at the 2018 product introduction. It was already a substantive market worth mentioning. It just got bigger since 2018.
The M-series mainly tries to push folks into Apple's hypervisor/virtualization and everyone wants to do that.
Similarly, some minis are used in embedded contexts where it is stability that folks will pay extra for (not high rates of change).
[ Right up until the very end MacStadium was buying pallets of MP 2013 systems when most other Mac Pro fans were throwing buckets of 'hate' at the MP 2013. Some smaller set of folks were probably still buying those shamefully old, ancient MBA processor powered, non-Retina 21.5" iMacs up until the end also. ]
An update to the current Mini form factor with a M2 Pro would probably work extremely well the "rack them up" crowd isn't stuck on a type 1 (bare metal) hypervisor solution. If there is 8P cores that is better than a mix for Continuous Integration/Build servers. The classic Mini case would have higher rack density than a Studio flipped up on its side. ( almost a 3:1 ratio of classic Mini versus Studio).
Apple could do a "thinned out " Mini more consumer focused M2 only , but being able to seamless slot into any one of those 10's of thousands of rack slots out there is an upside. ( Studios aren't that far off and probably contributing reason got that incrementally taller Mini solution. )
iPad Pro's. A10X. (on 10nm) in June 2017 . A12 (on 10nm). in Sept. 2017 .
The iPad Pro chips after that point skipped odd-numbered A-series and only more forward on process node updates.
The iPads stuck on being "cheaper" ? No. But that isn't whole iPad line up the last several years.
An iPad modem could be on a high acceleration path the TSMC N3 . At least in 2022 or 1H 2023.
Probably not. If High volume production starts in Q4 2022 that doesn't mean the chips come out the "baking" pipeline in a couple of hours or days. Dies have to cut, tested, validated , packaged ... etc.
Pretty good chance TSMC is talking about wafer starts into high production and not the back end packaging of completed product. If wafer start ramp in November ... still not going to make substantive end user delivery volume in 2022.
N3 is too late for any substantive volume end user delivery for this Fall.
N3 would be handy for a Max worth of stuff placed into a smaller die. That comes has upsides iin making a 4 midsize-large die package more tractable to make. The current Max+UtlraConnector die isn't not reticle busting large , but it is pretty far away from being small also. If multiply by four for package size calculations even dropping back by 50mm^2 is going to save 200mm^2 of space. (and for two die packages 100mm^2 .. still bigger than most A-series iPhone chips. )
If Apple sold enough Ultra's could collapse that offering into a monolithic die ( not much bigger than a Nvidia 3090 is now ... and don't have to pay for 3D packaging and get higher performance/watt. ).
It is rat hole to get into the the ' 3 / 5 / 7 ' hocus pocus marketing mindset that Intel created and AMD roughly copied.
Apple isn't using product numbers for performance binning. Few if any customers are going to confuse a plain Mn price to a Mn Ultra price. The combination of Apple's adjectives and prices will segment the vast majority of customers just fine. Apple doesn't have over a dozen SKUs to pick from.
Really hinges on whether think Intel is seriously addressing being a Fab vendor or not .
When the Fab was mainly just for the CPU products the marketing was more mixed up with the end products. ( The fab business really didn't have a variety of customers so there was not much of real deliberate sales partnerships to do because had a captive market. )
The other major difference is that if have more than one customer than different tweaks on fab process can be a different value proposition to different customers.
For example Intel's fab roadmap isn't necessarily the CPU package product road map going forward.
So early Intel 20A that doesn't need super high volumes before the High-NA EUV tools come on line could roll out for a customer with wafer start needs that matched what Intel could put out. So the initial , nominal 20A and 18A are modest volumes and a 20A-Plus and 18A-Plus roll out later at higher max volume on. High-NA EUV tools.
TSMC does variations also. N7 has a family of variants ( N6 7P , etc.). N5 has a family of variants ( N5P , N4 , etc.)
There is a bit of a convention difference. Intel's timeline is more about what TSMC would more likey laobel as initial "at risk' availability and not highest volume. But TSMC does talk about "at risk" windows also.
The. "internal risk reduction node"
If Intel finds a fab customer that is a better match for their needs they could run that for revenue later. That will basically lead Intel into broader 'pathfinding' on incremental changes. They can settle down to less "swing for the fence mega home runs" and more predictable process.
TSMC's dates are not solely based on Apple's dates. That is huge disconnect. There are subset of dates TSMC talks about, but Apple isn't their only customer that matters.
Arc Alchemist are fabbed on TSMC. That has nothing to do with Intel fab's timelines at all. The delay is more of a driver issue than a GPU chip production issue. [ reportedly Intel is stock piling chips from TSMC. ] If turn off some of the "overly nifty" features it goes faster.
https://videocardz.com/newz/intel-a...games-with-dynamic-tuning-technology-disabled
[ And be real.... Apple's initial MBP 16" intel system wasn't bug free either. The Studio Display firmware is borked. etc. ]
Can't really measure TSMC sliding roadmap dates against just Apple. Other folks have run into date slippage.
https://seekingalpha.com/article/4451575-tsmc-confirms-3nm-delay
Intel probably should be more conservative until after getting Intel 4 (EUV) out in volume in 2022-23 than claiming future distant leadership in 18A . The "Lunar Lake" package probably isn't going to be a homogeneous die package either. That could be thrown off by some non-Intel process.
Similarly, decent chance Intel CPU cores won't be the first 18A customer that ships for revenue.
The guys at Semi Wiki who have been heavily anti-Intel in recent years and who know what they’re talking about when it comes to semiconductors have this to say:
“we believe Intel has been able to significantly accelerate their process development at a time when the foundries are struggling. Although we don’t expect Intel to regain the density lead over the time period studied [2022-2026], we do believe they could retake the performance lead. We should get another good read on progress by the end of 2022 when we see whether Intel 4nm comes out on time.”
Can Intel Catch TSMC in 2025? - Semiwiki
At the ISS conference held from April 4th through 6th I presented on who I thought would have the leading logic technology in 2025. The following is a write up of that presentation. ISS was a virtual conference in 2021 and I presented on who currently had logic leadership and declared TSMC the...
Even if Intel gets their processes to work , they still have capacity issues.
https://semiwiki.com/semiconductor-services/ic-knowledge/311036-intel-and-the-euv-shortage/
If Intel has to multipattern on 'regular' EUV tools to get 18A results before get enough volume of High-NA EUV tools then even if process is working they will have issues rolling it out broadly across their very high volume product lines.
Intel's Gen 14 (Meteor Lake) is suppose to have a TSMC N3 iGPU tile on it. According to the schedules that is suppose to ship by mid 2023. About when Apple would be doing a ramp of something. Intel really can bring all the EUV level work inhouse even if had a working EUV process of their own. They don't have enough EUV tools to get it all done in 2023-2024. [ Previous CEO/board seriously under invested in buying EUV tooling several years ago. Went out and bought substantive more DUV tools but really should have done (pre-allocated money to ) both. More effort put into keeping the stock price inflated than in doing what they needed to do. ]
"Performance per watt" is a BS statistic. It will mean whatever INTC marketing wants it to mean.Intel is claiming performance per watt leadership on its 1.8 nm node. Lunar Lake and beyond.
The guys at Semi Wiki who have been heavily anti-Intel in recent years and who know what they’re talking about when it comes to semiconductors have this to say:
“we believe Intel has been able to significantly accelerate their process development at a time when the foundries are struggling. Although we don’t expect Intel to regain the density lead over the time period studied [2022-2026], we do believe they could retake the performance lead. We should get another good read on progress by the end of 2022 when we see whether Intel 4nm comes out on time.”
Can Intel Catch TSMC in 2025? - Semiwiki
At the ISS conference held from April 4th through 6th I presented on who I thought would have the leading logic technology in 2025. The following is a write up of that presentation. ISS was a virtual conference in 2021 and I presented on who currently had logic leadership and declared TSMC the...
Does it mean that INTC E cores achieve better perf/watt than Apple E cores? Or the same for P-cores? Or does it mean INTC GPUs achieve more "performance" (defined how?) than Apple GPUs?
Or is it going to be micro-managed down to some weird technicality that no-one can actually validate ("our transistors switch at x pJ/transition when running at 5GHz"?)
I can promise you that when it comes to the actual performance vs the actual wattage of relevance to customers, there is no way Intel will be leading. They may be ahead of AMD (and that may be their technical out) but there is no way they will be ahead of Apple. This is just common sense based on knowing where Apple is technically strong vs where Intel is weak, and the speed at which things do (and don't) change in this industry.
(a) The issue is not Intel fab vs TSMC fab. The issue is that Intel (AS AN ORGANIZATION) is still pathologically committed to claiming shipping dates that it cannot make. They learned nothing from Cannon Lake.
You cannot blame this Arc delay on Old Intel, that's PG's watch and he made the deliberate decision to go along with fake dates, even after he has made so many claims that, "oh no, this is New Intel; New Intel can be trusted with dates".
So, you want to trust the new process dates, be my guest. But don't be surprised when it ends in tears.
(b) You do realize that seekingalpha article (like everything that comes from Arne) is crazy nonsense, right? It is NOT an article claiming "TSMC's dates slipped, here's the proof", it's about "the internet" thought TSMC would deliver on usch and such a date and TSMC refused to oblige.
You do realize there's a paper trail for all this stuff, right? TSMC's announced dates in 2020 are given here:
TSMC's current dates are given here:
3nm Technology - Taiwan Semiconductor Manufacturing Company Limited
TSMC’s 3nm technology (N3) will be another full node stride from our 5nm technology (N5), and offer the most advanced foundry technology in both PPA and transistor technology when it is introduced. N3 technology will offer up to 70% logic density gain, up to 15% speed improvement at the same...
Note the sameness of the dates.
Witeken's whole shpiel (in 2021, between those two references) is about how TSMC has kept up with some supposed timeline that TSMC NEVER COMMITTED TO.
um what?
Hope it helps...
| Prefix | Measurement | Scientific Notation |
|---|---|---|
| Milli- | 0.001m | 1 x 10–3 m |
| Micro- | 0.000001 m | 1 x 10–6 m |
| Nano- | 0.000000001 m | 1 x 10–9 m |
| Pico– | 0.000000000001 m | 1 x 10–12 m |
Hard to say. We should never count a technology company out, especially one investing billions of dollars in EUV machines.
My vote is to watch the technology wars that are coming and to profit (as a consumer) off of the cool tech that apple, amd, Nvidia, and intel are about to unleash on the ecosystem. Competition is great for us consumers. And it’s about to really heat up in the cpu space.
True. It's a big if, and Intel has stumbled before. I only raise this article to show that if Intel can deliver on time (unlike the experience with 14->10 nm), it has a chance of performance leadership. That's something that was laughable years ago, but Alder Lake performs very well (even if not very power efficient when under an all-core load). Intel could be coming back.
Either way, competition amongst the brands (Intel, AMD, Apple, Nvidia, and possibly Qualcomm) will be very interesting to watch in the next 2-3 years. I believe we consumers will benefit tremendously from the designs that are coming.
A major contributing problem with the 14->10nm transition was that Intel tried to "make up lost ground" by compressing more stuff into a shorter timeline. 14nm was a bit late so stuff an extra dose of density bump into 10nm to catch up. Trying to do all of the catch up in one big jump was a mistake.
I think folks are not fully appreciating that by shifting to Angstroms Intel gets to take smaller increments. Going from 2nm to 1nm is a jump of 10 Angstroms. Going from 20A to 18A is just 2. It is an approach where they are taking smaller increments. So pulling forward a 2A process increment by 3-6 months is different, smaller hurdle to jump over than pulling back 10A increment by 6 months.
The bigger jump that Intel is talking about doing is the Intel 7 to Intel 4 jump. 4 to 3 will likely be more a refinement than a huge jump.
Samsung has had problems getting gate-all-around to work. (essentially what Intel calls RibbonFet) TSMC hasn't been doing peculiarly better either ( they have avoided the transition away from FinFet. ) . So one issue is that Intel's two competitors can't necessarily quickly extend their lead either. There are two major inflection points coming ( gate-all-around and High-NA EUV tools) . What Intel mainly has to do is not shoot themselves in the foot and 'win' one step at a time.
However, if Samsung has a "Eureka" moment after diligently working away at the problem , then Intel will have issue ( that they don't control). Samsung open a window for Qualcomm interest in 20A.
Intel’s foundry roadmap lays out the post-nanometer “Angstrom” era
Could the “Intel 20A” process node (that’s 20 Angstroms) rival TSMC and Samsung?
arstechnica.com
Qualcomm has a variety of chips they need to make. they don't need to dump everything onto INtel 20A , but if 20A is tuned for what they need for a subset product then it could be a better choice. And Intel could lean on Intel 3 longer to get to 18A for a broader set of their products.
Similarly. not all customers are interested in ultra-hyper density.
"... TSMC’s N3E node was explicitly designed to improve the process window to speed up time to yield, increase yields, boost performance, and lower power. The report says this would all be accomplished at the cost of lower transistor density compared to the original N3. Initially, TSMC expected to initiate high volume manufacturing (HVM) using N3E about a year after N3 HVM start, sometimes in Q3 2023. But since test production yields of N3E are already high, TSMC wants to start using it commercially earlier, sometimes in Q2 2023. ..."
TSMC’s N3E Will Reportedly Enter Production Ahead of Schedule
But do not expect many 3 nm designs shortly.
TSMC seems to be forking off more supplemental variants on each generation now. N5 , N5P , N4 , N4P , N4HPC ( maybe N4N(vidia) on Nvidia slides). . There is a decent chance they will get bogged down a bit on that at N3. TSMC trying to make the widest set of variants makes them move vulnerable to a vendor who limits the focus to a more narrowly defined subset ( to cherry pick off a subset of clients. ) .
If the super duper highest transitory density has wafer costs that are 2x TSMC N3 wafer costs some customers are going to pass.
It wasn't laughable that AMD could catch up 5-6 years ago either. Again... narrow focus and stop shooting yourself in the foot and the odds aren't as bad. ( and willy nilly chunking out your fab business was one of those shot yourself in the foot moves made by AMD. A fab business that is completely addicted to 'lock in' wafer order contracts isn't really competitive. ) . Asking the Intel foundry services business to be competitive actually would help the whole ecosystem of Intel products.
Intel isn't guaranteed a win . but not guaranteed a loss either. Few were picking the Cincinnati Bengals to be in the Super Bowl back in October 2021. The Bengals didn't win the Super Bowl but just getting there was very good outcome. ( And if 3 or 4 plays had lucky bounces had gone the other way .. they would have won. )
I think we will probably see lower overlap between the processor package products from those five over the next 2-3 years rather than more. (or everyone trying to match Intel's make "everything for everybody" processor package approach. )
that isn't necessarily going to be bring more competition inside of specific , narrow product segmentations. (e.g., low end laptop or extremely high end workstation. )
Consumers in the same product category will probably be paying just as much (if not more). The factor of some consumers being able to move a lower tier and still get their needs met ( e.g., desktop -> laptop with no performance problems. ) has been in play over last 40+ years.