TSMC on Course to Secure 5nm Chip Orders for 2020 iPhones

[iReality85]

Can someone explain to me, what will happen when it reaches 1 nm? What's next? I am clueless on this subject!

Picometers follow nanometers, so 1nm is 1000pm.

When we get to pm, we're getting into really, really small measurements. Like, atomic measurements. Atomic nuclei can be several dozen to a few hundred pm wide.

Practically speaking though, we may hit the limits of physics around 1nm, since at that size you start to encounter weird aspects of quantum mechanics like random positioning. Random positioning is very bad when it comes to things like CPUs that depend on the precise location of atoms passing through gates, which in turn is what generates a 0 or 1 value. When you have very tight gates (like at 1nm), it increases the chance that an atom might return 0 and 1 because of quantum mechanics. It will be interesting to see how the industry deals with this.

 

[TheFluffyDuck]

A transition away from intel will be a nail in the coffin for power users. I am sure some corporate type with more money than sense will love his razor thin rose-gold MacBook with only one USB-C. But for those of us who need grunt, we are gone.

 

[Freida]

This is pretty good news for the near future and beyond. At some point the nm is "good enough" just like at some point Ghz was good enough. The next thing is 5G everywhere on every device and accessory. I still would like to see backward software compatibility with Win.Unix.OSX apps and peripherals. Smallness is also good enough.

What is needed is form factors that suit the task. Desktop, mini headless, mini, mid, large laptops, mini, mid, large pads, mid and large phones. Wearables, injectables (bone fone, wrist, etc).

Its never going to be enough.
Ghz was limit so they needed new approach so they came up with cores. Now everything is shrinking etc. but eventually it will hit a limit also when they will have to come up with new architecture or totally new approach. We will always need faster chips, its just nature of the beast. At my work we have thousands of computers calculating a single frame and it still take hours. Some shots take days to render so imagine if we had a chip that would do that in real time. Wow, the amount of work we would be able to do in such time would be amazing.
Anyway, good on TSCM for leading (for now) the market. Intel is so behind (even though the technologies are apparently not the same so 10nm is like 7nm from Intel - correct?)
I personally think that Intel is done. Its just slowly collapsing

 

[now i see it]

The iPhones are currently plenty fast enough. What a 5nm chip will bring is lower power consumption so that  can reduce the battery size to make the phone thinner

 

[unfrostedpoptar]

This is pretty good news for the near future and beyond. At some point the nm is "good enough" just like at some point Ghz was good enough.

The GHz wasn't good enough. CPUs stopped getting faster because they generated too much heat. That's what started the move to multiple cores. No one wanted that since software wasn't designed to deal with splitting up execution but there was no choice. It's much better now - after many years of compiler improvements - but life would be so much easier if we could have kept going and have 50GHz, single-core CPUs now.

And, it will be the same with chip feature size. Things will hit a wall. And each shrink has greatly increased the complexity and cost of making chips. Lots of chips are still made on 28nm since you need iPhone level quantities to make 7nm worth the cost. Current RAM (and processors?) are already using some workarounds like 2 1/2 D designs and finFETs and lots of work in memories to stack vertically.

Also, on everyone bagging on Intel; some is deserved, but, as other articles have explained, each fab means a different things when they say "10nm" or "7nm" so you need a lot more info to do an apples-to-apples comparison of density.

 

[BvizioN]

Picometers follow nanometers, so 1nm is 1000pm.

When we get to pm, we're getting into really, really small measurements. Like, atomic measurements. Atomic nuclei can be several dozen to a few hundred pm wide.

Practically speaking though, we may hit the limits of physics around 1nm, since at that size you start to encounter weird aspects of quantum mechanics like random positioning. Random positioning is very bad when it comes to things like CPUs that depend on the precise location of atoms passing through gates, which in turn is what generates a 0 or 1 value. When you have very tight gates (like at 1nm), it increases the chance that an atom might return 0 and 1 because of quantum mechanics. It will be interesting to see how the industry deals with this.

Interesting! Thanks for the explanation. I assume they would be looking at using different materials from the ones they are using no, silicon! It is absolutely mind-boggling how fast technology is advancing! With this speed, I think we may reach the singularity very soon.

 

[Salty Pirate]

Slept late, you got the word in - superb!

Q1: Do we need "quantum" progression with each model? Asking because it keeps the cost high.

Q2: Such attention to detail to designing at the nanoscopic level Apple, great job. Now, how about that at the macro level? That design team stinks and is costing you sales and repair dollars.

Buzzword Bingo...

 

[Mr. Dee]

How much more power does one need to comment on Macrumors?

 

[anshuvorty]

Apple's silicon team is definitely going to hit a brick wall come 3nm, 2nm, and even 1nm. There is going to be a plateau at some point. This ever reliable pattern of shrinking die size will stop and when that happens, are these same Apple fanboys going to complain, like with what is happening with Intel?

 

[ph001bi]

Can someone explain to me, what will happen when it reaches 1 nm? What's next? I am clueless on this subject!

The end of life as we know it.

 

[MoreRumors?]

Nice but let's continue focusing on getting that modem moving along.

 

[iReality85]

Interesting! Thanks for the explanation. I assume they would be looking at using different materials from the ones they are using no, silicon! It is absolutely mind-boggling how fast technology is advancing! With this speed, I think we may reach the singularity very soon.

It is really interesting! Quantum mechanics is always a fun read, and it will have very important ramifications on CPUs and electronics in the next decade or so as we get closer to 1nm.

If you're familiar with Schrödinger's Cat, then you're familiar with how separate (but simultaneous) paths of the same object can lead to separate outcomes based on observation -- a main tenant of quantum mechanics. The cat in the box is both dead and alive, and is only one or the other once it is observed. Electrons can behave the same way when it comes to position, and this is a problem for current CPU gate-based architecture. Historically, it hasn't been an issue because gates have been (relatively) large and roomy for electrons to pass through. However, since electrons can be in two positions at once due to the nature of quantum mechanics, then small gates may not be able to accurately control electrons, thus throwing a wrench into CPU calculations.

Some interesting short reads:

https://computer.howstuffworks.com/small-cpu2.htm

https://phys.org/news/2015-01-atoms.html

 

[techwhiz]

The iPhones are currently plenty fast enough. What a 5nm chip will bring is lower power consumption so that  can reduce the battery size to make the phone thinner

Maybe.
Leakage will be an issue.

 

[mavericks7913]

Are these going into Macs as well? They're going to be absolute beasts at 5nm compared to whatever Intel puts out. Probably at least 6500 per core on Geekbench.

Maybe MacBook only.

 

[cmaier]

It is really interesting! Quantum mechanics is always a fun read, and it will have very important ramifications on CPUs and electronics in the next decade or so as we get closer to 1nm.

If you're familiar with Schrödinger's Cat, then you're familiar with how separate (but simultaneous) paths of the same object can lead to separate outcomes based on observation -- a main tenant of quantum mechanics. The cat in the box is both dead and alive, and is only one or the other once it is observed. Electrons can behave the same way when it comes to position, and this is a problem for current CPU gate-based architecture. Historically, it hasn't been an issue because gates have been (relatively) large and roomy for electrons to pass through. However, since electrons can be in two positions at once due to the nature of quantum mechanics, then small gates may not be able to accurately control electrons, thus throwing a wrench into CPU calculations.

Some interesting short reads:

https://computer.howstuffworks.com/small-cpu2.htm

https://phys.org/news/2015-01-atoms.html

This is why gate width is important, not just gate length. Like water running through a hose, the bigger your foot, the harder you can step on it and stop leakage. But both dimensions of your foot matter. Finfets are mainly beneficial because they allow bigger gate surface area without taking bigger lateral space.

 

[mavericks7913]

Microsoft is also working on adding Windows 10 to ARM-based Surface like Apple adding their own chips to MacBook.

 

[deebinem]

Did TSMC manufacture chips in the Apple Watch as well?

 

[Cosmosent]

Regardless as to what Synthetic Benchmarks may tell Users, the A10 is AAPL's Highest-Perf mobile processor when it comes to Real-World Applications Performance !

The issue is the Performance Controller, which is the Scheduler for the cores, + the low-level OS software that works with it.

The A10 probably has a basic Perf Ctrl, & it works beautifully !

The A12 has a "fancy" Perf Ctrl, that is, from my perspective of a Seasoned iOS App Dev with 20K+ R&D hours, very-likely broken.

My guess is AAPL simply didn't test Ramp-Up OR Switching Performance.

The XR is, IMO, the first "Lemon" AAPL has produced, because of this issue AND others (e.g., starting with iOS 12, the low-level camera frameworks are excluding up-to 432 MB of memory for NO "functional" purpose ... that's precious memory that apps could be using to provide a better / improved User Experience).

ALL of this is related to "silicon," so hopefully I don't get dinged again (by the MR editorial staff) for going Rogue / Off Topic / Trump

 

[KPandian1]

The A10 probably has a basic Perf Ctrl, & it works beautifully !

The A12 has a "fancy" Perf Ctrl, that is, from my perspective of a Seasoned iOS App Dev with 20K+ R&D hours, very-likely broken.

My guess is AAPL simply didn't test Ramp-Up OR Switching Performance.

ALL of this is related to "silicon," so hopefully I don't get dinged again (by the MR editorial staff) for going Rogue / Off Topic / Trump

I think you are on topic - we are comparing existing CPU dimensions and benefits with the possible new one.

 

[BvizioN]

The end of life as we know it.

Ironically, that's not entirely impossible!

 

[OverChris]

Such a powerful chip might be used in augmented reality glasses. I hope they plan that for 2021 or 2022. Apples focus on AR in the iPhone might lead to something useful after all.

 

[PKoz]

I don’t know if it is lower screen resolution combined with the power of the A12, but my XR absolutely screams through everything. It is so fast, I really can’t imagine apps opening faster, gestures getting any faster. But they will I’m sure. The A12 is going to be comfortably fast for several years.

I have a 12" MacBook with the 1.3 GHz CPU, which feels like a total slug compared to the latest 12.9" iPad Pro. Geek Bench shows the iPad Pro 12.9" latest, blows away the 12" MacBook. Better CPU and GPU performance. NOw all we need is to see OSX compiled for Apple's next-gen ARM processor. I would love a 12" MacBook ARM running OSX.

 

[cmaier]

Such a powerful chip might be used in augmented reality glasses. I hope they plan that for 2021 or 2022. Apples focus on AR in the iPhone might lead to something useful after all.

I think their first glasses will have a weak processor and rely on the iPhone for most of the work, much like the original Apple Watch.

 

[velocityg4]

I really hope in 2020 A14 5nm to come i to 12” Macbook to see how well it works,how powerful is conpared to previous 5w intel and how much battery can deliver,probably from 9-10h to 13-14h

I’d be curious how these scale up to desktops. Ones with 8/16/32 cores and the clock rates cranked up for 65w, 95w and 130w TDP.

 

[neutrino23]

Can someone explain to me, what will happen when it reaches 1 nm? What's next? I am clueless on this subject!

First of all, there is not a standard way to define this dimension. Basically it should refer to the smallest dimension in the chip. What is really amazing about this is that the atoms used in ICs have dimensions of roughly 0.2 or 0.3 nm. So a 5 nm width is only about 10 to 15 atoms across. In some cases these are not thin lines like wires but sheets of some sort so they are larger in two dimensions but very thin in the third.

All sorts of crazy things happen when dimensions get this small. I am amazed that these things even work.

A second problem is producing them. The basic idea is to use a cutout of the desired shape then shine light through that to cast shadows on the wafer. The wafter is first coated with a material that is light sensitive. This transfers the pattern to the wafer and unwanted areas of material are washed away. The problem is that using lenses you might expect that the sharpness of the shadow you would cast would be about the wavelength of the light. Green light has a wavelength of about 500nm. This is 100x the dimensions we are talking about here so you can imagine the level of technology they need to get down to the 5nm love. The article talks about using EUV which is Extreme UltraViolet. According to Wiki they are getting to 13.5nm or 92eV. This is incredible. In my field we call these soft x-rays. Li emits a 55eV x-ray and Be emits a 115eV x-ray. The Al L-line is about 87eV. Photons at these energies are incredibly difficult to work with as they are absorbed by almost everything. I can't even imagine how they make the masks and deal with focusing EUV through the mask.