Apple Silicon's Next Big Upgrade Could Arrive in 2026

[Chuckeee]

iOS decided to fight back and irradiate AndroidOS from existence!

One could only hope.

Much more likely, iOS merged with AndroidOS and thus United, they rose up in rebellion against their human oppressors. This resulted in enslaving of Mark Zuckerberg and Elon Musk to be their puppets and to act as their spokespersons to govern the people of the earth.

Ah ha ha 😈

 

[picpicmac]

3 nm was a let down with no improvements in battery life.

You're basing that on what, exactly? One product (iPhone 15Pro)? And there are not even real world, daily use statistics yet for the 15Pro. And you have no way to know if the software on the 15Pro is doing exactly the same thing iOS did on earlier models.

And guess what - from here on out there will only ever be small increments in performance and energy use. The chip industry is slowly turning into the toaster industry - after a while all you can do is some design tidbits to keep up with fashion trends.

 

[headlessmike]

Though note the marketing names don't represent the physical distances. The feature sizes for the current "3 nm" processes are quite a bit larger than 3 nm, and that will apply to "2 Å" as well. Thus you'd want to consider the quantum effects that correspond to the physical distances rather than the marketing names.

True. But even now chip fabs are struggling with quantum effects when designing for the last few generations of node sizes.

 

[Argon_]

True. But even now chip fabs are struggling with quantum effects when designing for the last few generations of node sizes.

Electrons doing things that Einstein said they shouldn't. Damn electrons.

 

[Chuckeee]

Electrons doing things that Einstein said they shouldn't. Damn electrons.

Actually, Einstein was initially very dubious of Quantum Theory and while later in life Einstein accepted the basic premises of Quantum Theory, he was never a firm supporter.

https://en.m.wikipedia.org/wiki/Bohr–Einstein_debates

 

[headlessmike]

Electrons doing things that Einstein said they shouldn't. Damn electrons.

Electrons right now:

 

[caribbeanblue]

Lorem ipsum dolor sit amet, consectetur adipiscing elit.

 

[theorist9]

I'm more curious about how the costs compare vs. N3B. A marginal decrease in costs could allow Apple to add in more transistors to amplify the marginal benefits of N3E (we're talking very small gains though, 1.05 x 1.05 = 1.1025).

It's been reported that there's a significant decrease in complexity with N3E, resulting in a decrease in cost per unit area. That should outweigh the small decrease in density, resulting in a net decrease in cost/chip for the same number of transistors. (https://www.anandtech.com/show/1883...n-schedule-n3p-n3x-deliver-five-percent-gains)

If you want concrete numbers you might find them by searching TSMC investor reports/meetings/calls.

 

[caribbeanblue]

Lorem ipsum dolor sit amet, consectetur adipiscing elit.

 

[GFLPraxis]

moore's law is definitely slowing done.

the good thing is software hasn't caught up with hardware yet.

I wonder if Intel will ever catch up after they stumbled

So, Moore's law has kept up pretty well until very recently, in that the density of transistors was doubling.

However, in the old days, doubling the transistors doubled the performance...now, that doesn't work well. You hit thermal limits and other issues way too quick.

So we'll see a doubling of transitors...but not a doubling of performance. And this issue is getting worse. The returns from density is diminishing...and eventually we'll hit a cap on density (though we probably have a few more doublings left).

For example, we just hit the limit on SRAM scaling.

TSMC's 3nm Node: No SRAM Scaling Implies More Expensive CPUs and GPUs

Big problems from tiny memory cells.

www.tomshardware.com

We are not able to get more SRAM out of the smaller fabrication. N5 > N3 only gave 5% gains and N3 > N3E is going to get 0% gains even though it's at a smaller process.

 

[chmania]

ARM chips are supposed to be much cheaper in production than Intel chips some say, but Apple M chip computers are much more expensive. Also, there's only one factory is manufacturing these M chips, and that's in Taiwan. Most probably, that's why they are expensive, one manufacturer and one receiver. Item chips are tested by 100s of companies, so their reliability is high and they become cheaper. Maybe, one day Apple will move back to Intel.

 

[Populus]

ARM chips are supposed to be much cheaper in production than Intel chips some say, but Apple M chip computers are much more expensive. Also, there's only one factory is manufacturing these M chips, and that's in Taiwan. Most probably, that's why they are expensive, one manufacturer and one receiver. Item chips are tested by 100s of companies, so their reliability is high and they become cheaper. Maybe, one day Apple will move back to Intel.

That’s what was said back in the day with A7, A8, A9 chips… but nowadays, especially since the implementation of the Extreme UV Lithography (N5P and onwards IIRC) on the A15 SoC and onwards, I’m sure the costs from TSMC have skyrocketed. I don’t have data to prove it, but it’s my guess, and thankfully here on the forum I can make wild guesses without being downvoted.

This increased costs situation apparently got even worse with the new (and exposure for Apple) N3B process node. So I’m pretty sure the new M3 family of chips is pretty, pretty expensive.

Still, Apple Silicon should be cheaper than Intel, just not as cheap as several years ago, when an iPhone chip could be just $20 (this is a random number, I don’t recall the exact cost of an A7 or an A10).

However, despite Apple Silicon being cheaper to source from TSMC, you have to take into account all the R&D that comes from Apple itself, and I’m sure it’s not marginal.

 

[Chuckeee]

It all comes down to “how you squeeze the balloon”. As miniaturization technology matures you can use the results to make the individual chips smaller (hence more chips per wafer, which in turn reduces the per chip cost) or to squeeze more stuff into the same size chip. Apple has selected to do the second (squeeze more stuff in).

 

[chmania]

That’s what was said back in the day with A7, A8, A9 chips… but nowadays, especially since the implementation of the Extreme UV Lithography (N5P and onwards IIRC) on the A15 SoC and onwards, I’m sure the costs from TSMC have skyrocketed. I don’t have data to prove it, but it’s my guess, and thankfully here on the forum I can make wild guesses without being downvoted.

This increased costs situation apparently got even worse with the new (and exposure for Apple) N3B process node. So I’m pretty sure the new M3 family of chips is pretty, pretty expensive.

Still, Apple Silicon should be cheaper than Intel, just not as cheap as several years ago, when an iPhone chip could be just $20 (this is a random number, I don’t recall the exact cost of an A7 or an A10).

However, despite Apple Silicon being cheaper to source from TSMC, you have to take into account all the R&D that comes from Apple itself, and I’m sure it’s not marginal.

I am not that sure of the quality of that R&D, for what I see in the complains from the M chip users in many forums, even in the Apple support forum. I am actually using an Intel 2018 15" MBP and I don't have all those complains that they shout about, such as battery drain, heating up, fans running mad...I believe with the Intel chip Macs, there was a lot of R&D done by Intel, and with the feedback from all those computer manufacturers to Intel around the world.

Going as a Lone Ranger in the technology world is quite nice, but I am not that sure if BSD would work well with ARM. I'll probably wait for quite sometime to buy an M chip Mac, that is, if/when the Intel MBP would die one day. Wish it won't, as it has the TouchBar. 👌

 

[Tagbert]

I am not that sure of the quality of that R&D, for what I see in the complains from the M chip users in many forums, even in the Apple support forum. I am actually using an Intel 2018 15" MBP and I don't have all those complains that they shout about, such as battery drain, heating up, fans running mad...I believe with the Intel chip Macs, there was a lot of R&D done by Intel, and with the feedback from all those computer manufacturers to Intel around the world.

Going as a Lone Ranger in the technology world is quite nice, but I am not that sure if BSD would work well with ARM. I'll probably wait for quite sometime to buy an M chip Mac, that is, if/when the Intel MBP would die one day. Wish it won't, as it has the TouchBar. 👌

What!? The 2015-2020 Intel Macs are notorious for heating up and crazy fan speeds. The M-series Macs are generally cool and quiet.

Usuing support forums as a survey of how common problems are is futile. It is a self-selecting group who come to talk about problems. No one comes to forums to talk about how few problems they are having.

 

[chmania]

What!? The 2015-202 Intel Macs are notorious for heating up and crazy fan speeds. The M-series Macs are generally cool and quiet. Usuing support forums as a survey of how common problems are is futile. It is a self-selecting group who come to talk about problems. No one comes to forums to talk about how few problems they are having.

I don't know how many Intel Macs you own, but the one I own, the 2018 15" MBP doesn't heat up. And, runs on the battery just as new, about 40mins less than advertised. So, the thing is, whether I should buy a M chip Mac in the near future or not, not until it dies (hope not!). BSD and Linux is not that keen (not yet) on ARM chips, not for heavy duty stuff, by the way.