TSMC's 3nm chips won't be ready in time for 2022 MacBook Pros, Kuo says
Well-connected analyst Ming-Chi Kuo is doubling down on his forecast that Apple's upcoming MacBook Pro and iPad Pro models won't sport processors built on a 3nm chipmaking process.
-xl.jpg)
14-inch and 16-inch MacBook Pros
In a tweet on Friday, Kuo reiterated his prediction that upcoming 14-inch MacBook Pro and 16-inch MacBook Pro models would sport new chips based on TSMC's existing 5nm production process. That runs counter to other reports suggesting that the models would have 3nm chips.
According to Kuo, components must be purchased by October at the latest for products will enter mass production in the fourth quarter of 2022. He says that TSMC's 3nm chips won't be available until January 2023 at the earliest.
As far as evidence backing up his claim, the analyst says that TSMC's 3nm chip production cycle is about "4 months." So, despite the fact that TSMC is kicking off mass production of 3nm chips in September, they won't be ready to ship until January.
In a tweet from earlier in August, Kuo also noted that TSMC's own financial guidance only includes contributions from 3nm chips starting in the first quarter of 2023.
Kuo's prediction here runs counter to other reports from the likes of DigiTimes and other publications.
However, some reports, such as a piece in the Commercial Times, don't outline a specific start date for 3nm chips. Earlier in August, the Commercial Times said that Apple would be the first company to get its hands on 3nm chips, and that production of 3nm chips would start in September. Both of those claims can still line up with Kuo's forecast.
Apple is rumored to be working on new 14-inch and 16-inch MacBook Pro models for release in the second half of 2022. Those models are expected to be incremental upgrades with updated "M2 Pro" and "M2 Pro Max" processors.
Read on AppleInsider
14-inch and 16-inch MacBook Pros
In a tweet on Friday, Kuo reiterated his prediction that upcoming 14-inch MacBook Pro and 16-inch MacBook Pro models would sport new chips based on TSMC's existing 5nm production process. That runs counter to other reports suggesting that the models would have 3nm chips.
According to Kuo, components must be purchased by October at the latest for products will enter mass production in the fourth quarter of 2022. He says that TSMC's 3nm chips won't be available until January 2023 at the earliest.
As far as evidence backing up his claim, the analyst says that TSMC's 3nm chip production cycle is about "4 months." So, despite the fact that TSMC is kicking off mass production of 3nm chips in September, they won't be ready to ship until January.
In a tweet from earlier in August, Kuo also noted that TSMC's own financial guidance only includes contributions from 3nm chips starting in the first quarter of 2023.
Kuo's prediction here runs counter to other reports from the likes of DigiTimes and other publications.
However, some reports, such as a piece in the Commercial Times, don't outline a specific start date for 3nm chips. Earlier in August, the Commercial Times said that Apple would be the first company to get its hands on 3nm chips, and that production of 3nm chips would start in September. Both of those claims can still line up with Kuo's forecast.
Apple is rumored to be working on new 14-inch and 16-inch MacBook Pro models for release in the second half of 2022. Those models are expected to be incremental upgrades with updated "M2 Pro" and "M2 Pro Max" processors.
Read on AppleInsider
Comments
Apple could just delay the release of M2 Pro/Max.
But again, I can envision M2 Ultra using 3nm which would allow Apple to position it as a power workstation that makes a huge difference.
I would imagine that's what they would like to do (wait for 3nm for the Pro/Max/Ultra... versions). I am wondering though if they will make their projected 2 yr transition window with the MacPro and high end Mini before the end of this calendar yr. Or if they will just say "sod it" lets wait for the 3nm to be ready for those too.
https://ctee.com.tw/news/tech/702617.html
This is around the same as the highest 2019 Mac Pro. The highest performance Mac Studio is $5k. Doubling the chip would be under $10k, likely somewhere around $8-9k and would match a $24k 2019 Mac Pro for standard tasks and outperform it for video encoding.
NVidia will be launching the 4090 GPUs in September and this is rumored to be as high as 80TFLOPs. People will be able to fit two in a standard PC so 160TFLOPs of computer power.
I think 60TFLOPs would be enough for a Mac Pro launch and would sit between a 3090 and 4090 and it would be a small form factor chassis.
If the manufacturing cycle takes 3-4 months and starts in September, it makes more sense to use N5P because they can do an N3 refresh in October 2023.
I don't think there is much trust in the Mac Pro customer base for Apple. It's been 9 years since the 2013 Mac Pro. Apple's commitment to this set of customers is weird. I really don't get it as it can't be really that hard. It's basically to the point that whoever or whatever process that drives these decisions needs to be demolished, and they need to start over, again. It is likely a function of money as the Mac Pro seems more of a branding project, than a product line that tries to meet the needs of its market, so they leave it untouched for years-on-end.
I would hazard a guess that not many people will be putting 2 4090 GPUs in a box. The power consumption is rumored to be 400 to 500 Watts, and the Ti card could by 800. There is basically a "limit" at 1800 W. With next gen x86 desktop CPUs hitting 300 W, putting a high end CPU and 2 GPUs sounds pretty close to needing a dedicated circuit for one computer in your house. It's only the server GPU compute workflows that can drive sales of such a card. With crypto crashing and driving down the need for useless computations, perhaps the power usage in GPU cards will finally fall back to levels that aren't the equivalent of running a hair dryer in your room for a few hours. Hmm... have to think about selling my NVDA stock, but server-grade GPU compute is an increasing market.
What is left of real interest in M2 world is what Apple does with the Mac Pro. Tht’s makes good points above, but for the Mac Pro at least the emphasis on power management in the SOC should not need to apply. And should it be a SOC? And we are only hearing crickets.
It's always assumed that the latest Mac Pro or iMac Pro design is the one that works best for the target market because it hasn't been discontinued yet. Only Apple knows what unit volume they sell and which models. After the Mac Studio launch, I could see a lot of the towers being swapped out, especially for things like audio production.
It is hard to make a one-size-fits-all model for high-end use cases because the needs are varied. Audio production, VFX, film editing, scientific computation have different requirements. There's even software developers who buy these machines. A Unity developer here describes why they went for the 2013 model:
https://www.youtube.com/watch?v=OAZ16_rpcsg
Reusing the 2019 chassis directly isn't a good option, it needs a redesign for M2. Maintaining MPX module support would make people who invested in them happy but I doubt they'd bother with external GPU support. This would mean they'd have to commit to driver support for GPUs that would be used by a fraction of 1% of the Mac userbase and it doesn't seem necessary. They could easily make an external MPX box for the few that need them, maybe have a company like Sonnet make it and plug it into the Mac Pro via PCIe. Most other users would plug in an optical IO card for use in a networked environment and not need any other slots.
I think most people will use a single GPU too and 60TFLOPs dual Ultra M2 would be suitable for higher-end use cases. A 3090 is already very powerful for compute tasks:
https://www.youtube.com/watch?v=Q3UUkNxIdlU
It will be interesting to see what route they go with the Mac Pro. Building a 4-way UltraFusion for such a small portion of the Mac userbase seems excessive. M2 Ultra MPX modules might work but still feels over-engineered. I reckon 2 Ultras with a lower bandwidth interconnect would be the easiest way and they'd offer 128GB or 256GB models. It will be hard to differentiate the entry Mac Pro from the top Mac Studio.
I know we try to think like Apple product marketing would think - designing a machine for bulk use case in a niche - and it could be successful if Apple actually updated the machines. They haven't demonstrated any commitment. They didn't update the 2013 Mac Pro, 2017 iMac Pro and 2019 Mac Pro. They didn't have to go through this churn if they actually updated the machines. If the 2013 model was actually updated, it could have been successful. These restarts really don't help.
So, I definitely think the least worst path for them today is to continue to use the 2019 Mac Pro case, keep the 8 PCIe slots, with Apple Silicon. They should offer their own GPU card for people who want more. It could just be an M1 Max or an M1 Ultra in a clustered network mode, or perhaps they develop memory protocols to unify all the memory on the boot-up SoC and the SoCs in the cards. If customers want something that is cheaper than $6000, there is a lot of room below that to offer a model. Say, a M2 Max model for $3000. After 3 or 4 years, a lot of these models will become viable in the $1000 second hand market.
Companies like HP update their desktops and workstations frequently and they give a good indication of the industry trend.
https://d18rn0p25nwr6d.cloudfront.net/CIK-0000047217/0302cd18-964e-4bee-b427-d313202a7dd9.pdf
Desktops and workstations are trending down, laptops are trending up. HP is one of the biggest workstation manufacturers in the world and their annual revenue was $1.6b. Apple's Mac revenue last year was $35b.
There's nothing Apple can do to make a financial success in a dying market. The best they can do is to make the target audience happy with the product and their products work for the majority of the target audience.
Although they haven't updated the product lines directly, the models they released are suitable replacements. The 2017 iMac Pro was a suitable replacement for the 2013 Mac Pro and the Mac Studio is a suitable replacement for both of those. The Mac Studio is clearly the successor to the 2013 Mac Pro despite having a different name.
The industry is at a point where the spec isn't appealing in the same way. People aren't going to react to 100TFLOPs the same way they did to 10TFLOPs because significantly fewer people can make 100TFLOPs do something worthwhile. Faster is always better but it stops being meaningful when people don't feel that their workflows are slow.
The next Mac Pro just has to have some versatility while offering a worthwhile performance upgrade. If it's over-engineered, they will lose money on it because the cost of setting up the manufacturing will outweigh the profit from the unit volume.
I wouldn't characterize the workstation market as dying per se. It's jobs and a desktop's jobs are being taken over by laptops and SFF desktops (Mac mini sized PCs), and units and revenue are going down, yes. This has been happening for the better part of 2 decades, but it will eventually settle out to some sustained level. Apple knows this. They also know they can't abandon it, otherwise that could have axed workstations some time over the past 10 years, and they did a major reset in 2017 that culminated in the 2019 Mac Pro.
The software developed for use in workstations and servers, hardware too, trickle down to the premium and mainstream laptop level. It is what supports sales of Macbook Pros and Mac Studios. Without these workflows, sales of higher end Macs will diminish. These workflows are typically first developed using high end hardware like on a workstation. Performance in laptops will increase to the point that they will run these workflows well. The cycle continuously repeats because our problems and solution spaces are unending.
Hence, I think Apple decided that have to do a Mac Pro rather than abandoning this market. We are not in disagreement much other than Apple taking a rather odd path to support this market niche. Too much boutique. Not enough updates. This applies to software too. Don't you find it strange that FCP users wrote an open letter to Apple to get its shit together regarding FCP? This is Apple's software that is a big driver for its high end hardware. The hardware offerings themselves has been complained about ad nauseum.