Incidentally this means that Apple will no longer have a node advantage once Zen4 launches - both Zen4 and A15 will be on the same node, so we can make direct comparisons without people insisting that Apple's performance is solely due to node advantage/etc.
But yeah, that does go to show that 3nm is slow to launch in general - Apple would not willingly give up their node lead like this if there were anything ready for an upgrade. I don't think it's actually falling behind in the sense that it was delayed, but it seems even TSMC is feeling the heat and slowing down their node cadence a bit.
Also, as far as this:
> Second, the recourse to two M1 Max chips fused into a M1 Ultra means TSMC’s 5 nm process has reached its upper limit.
There is still Mac Pro to come, and presumably Apple would want an actual Pro product to offer something over the Studio besides expansion.
marcan42 thinks it's not likely that quad-die Mac Pros are coming based on the internal architecture (there's only IRQ facilities for connecting 2 dies) but that still doesn't rule out the possibility of a larger die that is then connected in pairs.
Also bigger/better 5nm stuff will almost certainly be coming with A15 on N5P later this year, so this isn't even "the best TSMC 5nm has to offer" in that light either.
I said quad-Jade Mac anythings aren't coming because that die is only designed to go in pairs (that's the M1 Max die). Everyone keeps rambling on about that idea because that Bloomberg reporter said it was coming and got it wrong. It won't happen.
Apple certainly can and probably will do quad dies at some point, it'll just be with a new die. The IRQ controller in Jade is only synthesized for two dies, but the architecture scales up to 8 with existing drivers (in our Linux driver too). I fully expect them to be planning a crazier design to use for Mac Pros.
That's a big change from Apple where they've historically put their newest processor in every single phone they launch (even the $430 iPhone SE announced last week has the A15 now).
I wonder if it's purely a cost cutting measure, or if they're not expecting good enough yields to supply them for every iPhone, or if they're holding some fab capacity back to have room for the higher end chips alongside the A16.
There's been rumors that they'll be skipping A15 cores for the upcoming M2 processors.
If they skipped over the best-selling iphone, that would give them a TON of extra space for M2 chips. This would allow them to put a little more ground between the new Air with the M1 and the pro iPads. It would also allow them to drop a new version of the macbook air and drive a lot of sales there. I know I'd gladly upgrade to a M2 model -- especially with a decent CPU bump and a 24/32GB RAM option.
Then again, they could just stick with what people expect. I wouldn't be surprised either way.
The iPhone 12 Pro was perhaps the least differentiated high-end model Apple has ever put out.
I think the 13 Pro has a few features that make it a bit more of a compelling buy:
- The new telephoto lens is a massive improvement (I wonder if your last experience was with the 12 or older? The new camera is actually worth something while the old one had mediocre quality compared to the main lens).
- ProMotion has no tangible benefit, but it makes every interaction with the screen look smoother. When you go back to old phones that don't have it, it's jarring. I can see why some of the Android-using tech enthusiasts have criticized Apple for not delivering high refresh rate for so long.
- The previous iPhone 12 had identical main/wide cameras with the Pro model unless you got the Max variant, which is no longer the case. The iPhone 13 Pro has different/better cameras all around over the 13.
- The GPU of the 13 Pro has an extra core over the 13, which was not the case for the iPhone 12 lineup. Anyone who does mobile gaming on graphically intense games should probably choose the Pro model over the regular one.
- Significantly better battery life over the non-Pro version, which was not the case for the 12 models, which had identical ratings.
I know the value, but I work in tech and spend tons of time digging into hardware as a hobby. Camera matters to some, but most of the rest are pretty bare features compared to the $200 (20%) increase in price.
When I list all the things I can buy with $200, where do these features rank in comparison to those other things? I'm blessed with a good job, so I can afford the luxury. I was poor when I was younger and I definitely wouldn't be spending that for those features. $220 out the door would be almost 20 hours of work at $15/hr (after taxes).
But, there are some other points to consider:
- It seems like most people in the USA who buy mid to high-end phones finance their phones from carriers, and pay 0% interest for it. So, what the consumer is really considering is "is the Pro model worth $5-8/month more to me?" or "Would I pay $200 extra over 2-3 years?" and I think that's an easier justification for many people.
- Carriers offer a number of financial incentives and discounts in exchange for loyalty (there aren't any contracts anymore, but there are "bill credits" that function the same way).
- You did use $15/hour as an example, which around the median US salary, but Pro models are not intended to be the top selling model for the median earner in the US. They're marketed at, I would guess, the top 20% of earners, which lines up with the Pro/Pro Max models only making up 20% of iPhone sales in 2020 [1]. That would mean that Apple would expect individuals buying the iPhone Pro models to make about $75,000/year or greater. About 10% of the population makes a 6-figure salary. [2]
- Smartphones are the primary communication and computing device for many if not most people. I think that there are many people who see the smartphone as the most valuable possession they own.
[1] https://www.knowyourmobile.com/phones/most-popular-iphone-mo...
[2] https://en.wikipedia.org/wiki/Personal_income_in_the_United_...
I find this attitude bizarre. It's not any cheaper! I guess it can make a difference if you have cash flow issues. But an iPhone Pro is decidedly a luxury, so if you have cash flow issues then you probably just shouldn't buy one?
It's not really about the total price paid, it's about whether the item is affordable on a monthly basis.
I agree with your philosophy where a lot of people are way too willing to let easy financing change their budget. I also assume that free financing would inflate prices (e.g.: cars and homes).
Would the iPhone and other flagship phones cost $700-1000 if financing wasn't so common? I don't think so, personally.
The other aspect of this is that, mathematically, $200 paid now is objectively worth more than $200 paid over time thanks to the time value of money. [1]
Let's say we're at a "normal" 2% rate of inflation. If I buy a $30,000 car today with 0% financing over 6 years, by the time I hit year 6 my monthly payment is representing a 12% lower value (inflation compounded yearly) than when I started.
Technically, even if you have $1000 in cash to buy your new iPhone, you "should" just finance it and invest/save/use the rest elsewhere.
[1] https://www.investopedia.com/terms/t/timevalueofmoney.asp
With my ADHD issues, the combination of
>The generally long battery of the ProMax in particular (I had issue with my phone dying midday, forgetting to charge)
>iOS 15 Focus mode
>UWB wireless trackers
>The most effective ANC in any wireless headphones
>Apple watch for notifications, timers, etc
Are all very helpful and the promax is just good for pragmatic and practical reasons. I would definitely say it's a high priced item but I wouldn’t frame it as a luxury. I fundamentally see my phone as assistive technology, it can’t die on me, Apple makes their longest lasting phone their most expensive.
Sure it does, particularly for a direct input device higher refresh rate screens improve input speed and accuracy.
Last year I upgraded from a 6s+ to a 12 and I can confirm this the other way around: my old phone did feel slow for a couple of key apps; turns out they are still slow on the new phone. They are just poorly written, and one is basically just a CRUD app with no excuse.
So my lesson is to likely keep this phone for a decade. It's not like I couldn't afford to upgrade but why bother?
[happy owner of a 6s here]
Another big difference is 5G. I can easily get 500+ MBit/s downstream while outside.
Oh, and about the cameras: the one update I especially like is the night photography, I can now even take photos of the starry night sky.
Besides that: it’s just a phone, it just has an absurdly powerful CPU in it, and now it does text detection in images. Screenshots and photos have selectable text in it.
Yes, the battery was kinda shot but I could have replaced it.
nm notation used to mean the width of the smallest feature that could be made. Even today there are processes such as atomic layer deposition (ALD) that allows singular atom thick features. The difference between nodes now are in shrinking macro features, you don't necessarily make them smaller, more important is density. This is currently done with 3d transistors (finfet) and perhaps in the future going full vertical. When all other optimization have been exhausted it's likely to see multiple layers of stacked transistors simulator to what they are doing with NAND memory chips. Eventually even that will hit a wall due to thermal limitations. Beyond that people have proposed using carbonnano tube transistors. That tech is very early but has been proved to function in labs. If we ever figure out how to manufacture carbon nanotubes chips, it will be truly revolutionary; you could expect at least another 50 years of semiconductor innovation.
That's the problem. We can't. All those technologies are in such a primordial state, if at all, that we don't even know if we will ever be able to use them efficiently 20 years from now.
Although if you just shrink the chips and keep the transistor count the same, then you have a more energy efficient chip. Which is especially useful for portable devices.
The real numbers have been around 20 nm for a decade. They decreased a bit with Intel's competitors achieving better lithography than them before them. And we are in the realm of tons of little tricks that improve density and performance - nothing really dramatic but there are still improvements here and there. The tens of billions of dollars thrown at research achieved them but it is not comparable to the good old days of the '80s, '90s and the '00s
I don't think that's fair. Density is still increasing fairly substantially. Just going off of TSMC's own numbers here:
16nm: 28 MTr/mm2
10nm: 52.5 MTr/mm2
7nm: 96.5 MTr/mm2
5nm: 173 MTr/mm2
Performance (read: clock speeds; but for transistors those are one & the same) are not really increasing, though, those have pretty much plateaued. And the density achieved in practice doesn't necessarily keep up, as the density numbers tend to be for the simplest layouts.Because it's all subjective now, companies went wild with marketing, because consumers know "lower nm => better". But, say, GF 14nm is much more comparable to Intel 22nm, and GF 12nm is still solidly behind late-gen 14++, probably more comparable to TSMC 16nm. Generally Intel has been the most faithful to the "original" ratings, while TSMC has stretched it a little, and GF/IBM and Samsung have been pretty deceptive with their namings. Intel finally threw in the towel a year or so ago and moved to align their names with TSMC, "10nm ESF" is now "Intel 7" (note: no nm) and is roughly comparable with TSMC 7nm (seems like higher clocks at the top/worse efficiency at the bottom but broadly similar), and they will maintain TSMC-comparable node names going forward.
Anyway, to answer OP's question directly though, "what comes after 1nm" is angstroms. You'll see node names like *90A or whatever, even though that continues to be completely ridiculous in terms of the actual node measurements.
future improvement is going to come from the same place it mostly comes from now: better design that unlocks better density and a revolutionary new litho process that as of yet doesn't exist
TL;DR: things get really murky after a notional 2.1nm generation. Past that we'll need a new generation of EUV sources, advancements in materials, etc, that AFAIK are still quite far from certain (but I am not an expert on this stuff by any means).
I personally think we're headed to a stall for a while where innovation will focus mostly on larger packaging/aggregation structures. Chiplets and related are definitely here to stay. DRAM is moving in package. Startups are playing around with ideas like wafer scale multiprocessors or ssds. I think clever combinations of engineering at this level will keep us with momentum for a while.
For GPUs, yield is less of an issue... the chips are manufactured with the expectation of a number of cores of the many thousand small (in terms of silicon area) ones being defective - overprovisioning, basically. That allows them to simply bin the sliced chips according to how many functional core units the individual chip has.
In contrast, even the Threadripper AMD CPUs have only 64 large cores which means the impact of defects is vastly bigger, and overprovisioning is not feasible.
Meanwhile if you're making 6 x 8 core chiplets and one of those cores is defective, well that chiplet can go into a 48 core or be a midrange consumer cpu or something, and you'll just pick one of your many many other 8 core chiplets to go with the rest for the 64 core.
>Qualcomm has decided to switch back to TSMC for the Snapdragon 8 Gen2 Mobile Platform. Samsung’s 4nm process node is plagued by a yield rate of as low as 35 percent.
https://www.techspot.com/news/93520-low-yield-samsung-4nm-pr...
https://www.theverge.com/22972996/apple-silicon-arm-double-s...
The author ran Apple Europe and then moved to the US and was an Apple VP for a long time. If anyone is allowed to have this kind of attitude then it's reasonable in Gassée.
In people in general, it's...weird.
Apple ultimately went with the NeXT / Steve Jobs combo, quite wisely, but for a long time there was a whole gang of BeOS fanboys lamenting that decision.
I find it hard to believe that this was a last-minute decision. Rather, I think this pattern of a new core design (using a new process if there is one) releases first for the smallest devices (iPhones), and the gradually moves its way up the lineup all the way up the Ultra before the cycle repeats with a new generation is likely Apple's new strategy going forwards.
My understanding is that this is pretty much what Intel and AMD do too (releasing their smaller dies on new processes first) and that this is a general strategy for dealing with poorer yield number on new process nodes. The idea that Apple would ever have considered releasing their biggest chip as the first chip on a new node seems far-fetched to me.