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

[mschmalenbach]

Everybody hypothetically spends Apple’s money, so I’m going to throw my hat in:
This is the company they should purchase.
No other company going forward will be so integral to Apple’s plans or success than the one manufacturing it’s chips. Which, in the not so distant future, could be for all of its devices.
-Of course, I realize it’s a more complex issue than what I laid out (spreading risk, geopolitics, etc) than just spending imaginary $.

I can't see the regulators in any country of consequence allowing that kind of acquisition to go through...

And if it did - how many existing clients of TSMC would go elsewhere, just in case? Not that Apple would care about the lost business... BUT... how many of the brainiacs at TSMC would stay?

The money aspects is typically the LEAST complex aspect of any acquisition.
[doublepost=1550891379][/doublepost]

Exciting to finally see EUV being used in chip production.

TSMC is truly leaving Intel behind. Still waiting for their 10nm chips, meanwhile TSMC is in full scale production.

A 10nm TSMC chip isn't the same as a 10nm Intel chip. It's closer in the bits that matter to an Intel 14nm chip.

That said, it's still impressive.

Getting a profitable yield is the truly impressive achievement though.

Think about how much a 10nm fab costs - about $10-$12 BILLION. Throw in the development costs of the chip itself, and mask sets costing around $250 million each,and you're looking at another $2.5-$3 BILLION.... you need to make that money back in 3-5 years or less. That's going to need pretty good yield levels...

 

[BaltimoreMediaBlog]

INTEL appropriately realized that Moore's Law has been invalidated and already slowed down on this "nm Fight", so I don't see how Apple's Fool's Gold investment is going to pay off. Didn't we go through this before with PowerPC?

The only advantage here is thinner smaller, more breakable cheaper & key word, disposable device!
Isn't that what Taiwan is best known for? HAHA!

So much for Apple being so environmentally friendly.

 

[Jakewilk]

The moment you realise that most people don't benefit or need any more CPU power. There are a lot of ways the iPhone can be improved (as can any phone), I don't think people are craving for more performance though.

Tired of hearing every year of how the brand "new" phone has 40% more CPU/GPU power. You could have lowered it by 20% and I wouldn't have been able to tell, that is the difference.

- Improve Siri
- Make it so that the camera bump no longer exists, it should be flush with the phone
- Innovative ways to hide the notch
- Greatly improve battery life
- USB-C port
- Maybe the headphone jack back?

But a 5nm chip? Don't care.

Thinner chips means better battery life, also saying the iPhone doesn’t need to be faster is incredibly short-sighted. Just because A12 is good now doesn’t mean it’s going to be good for a 2020 phone
[doublepost=1550931464][/doublepost]

INTEL appropriately realized that Moore's Law has been invalidated and already slowed down on this "nm Fight", so I don't see how Apple's Fool's Gold investment is going to pay off. Didn't we go through this before with PowerPC?

The only advantage here is thinner smaller, more breakable cheaper & key word, disposable device!
Isn't that what Taiwan is best known for? HAHA!

So much for Apple being so environmentally friendly.

Thinner chips necessarily means faster and more power efficient. It’s physics, not moore’s law

 

[cmaier]

Thinner chips means better battery life, also saying the iPhone doesn’t need to be faster is incredibly short-sighted. Just because A12 is good now doesn’t mean it’s going to be good for a 2020 phone
[doublepost=1550931464][/doublepost]

Thinner chips necessarily means faster and more power efficient. It’s physics, not moore’s law

The chips aren’t getting thinner.

 

[GuruZac]

It will be interesting to see what they do to get around the fact that a single silicon atom is 0.2nm!

Reminds me of this article I just read. A lab out in UC Berkley just used non-contact Atomic Force Microscopy to image the reactions taking place to produce graphene molecules. I spent two semesters drawing theses structures back in Organic Chemistry 1 & 2 and turns out they were pretty accurate!

 

[Jakewilk]

The chips aren’t getting thinner.

5nm is thinner than 7nm, unless they changed how counting works

 

[cmaier]

5nm is thinner than 7nm, unless they changed how counting works

Except the figures you are citing do not refer to the thickness of the chip.

 

[827538]

I can't see the regulators in any country of consequence allowing that kind of acquisition to go through...

And if it did - how many existing clients of TSMC would go elsewhere, just in case? Not that Apple would care about the lost business... BUT... how many of the brainiacs at TSMC would stay?

The money aspects is typically the LEAST complex aspect of any acquisition.
[doublepost=1550891379][/doublepost]

A 10nm TSMC chip isn't the same as a 10nm Intel chip. It's closer in the bits that matter to an Intel 14nm chip.

That said, it's still impressive.

Getting a profitable yield is the truly impressive achievement though.

Think about how much a 10nm fab costs - about $10-$12 BILLION. Throw in the development costs of the chip itself, and mask sets costing around $250 million each,and you're looking at another $2.5-$3 BILLION.... you need to make that money back in 3-5 years or less. That's going to need pretty good yield levels...

As an Electronic and Electrical engineer I fully understand chip technology naming schemes. Going by the gate/fin/spacing pitches Intel's 10nm is similar to TSMC's 7nm - but at least TSMC is in full scale production. 5nm EUV will put TSMC firmly in the lead when it comes to transistor lithography.

I agree with you that Apple should not buy TSMC, it doesn't even make much sense. It would cost Apple probably around 1/4 trillion dollars all in and what 'extra' do they get for that? I'd say the direction is more for companies to design their own chips and have whoever has the best price/perf fab for them. Also by doing things the way they do currently the development costs of new fabs and processes is shared amongst many customers and not just one. Easier for Apple to just buy first dibs on a new process. And yeh, fully agree, no regulator in their right mind would allow Apple to buy TSMC. It would just be an all round bad move for everyone.

 

[341328]

Chips are going to be invisible soon at 0 nanometer

 

[827538]

Chips are going to be invisible soon at 0 nanometer

The transistors in these are more like 40nm in size. The 14/10/7/5nm are really just marketing terms and pretty much nothing to do with the naming schemes used.

 

[chucker23n1]

Next stops, 4000-3000-2000 pm chips

For those wondering, we hit 3000 nm (or 3 µm) in 1977, and 1000 nm in 1985.

 

[Jakewilk]

Except the figures you are citing do not refer to the thickness of the chip.

Thank you, Captain Semantics

 

[cmaier]

Thank you, Captain Semantics

How is that semantics? Even colloquially, nobody refers to reducing feature size as making "thinner chips."

 

[Jakewilk]

How is that semantics? Even colloquially, nobody refers to reducing feature size as making "thinner chips."

Yes they do

 

[cmaier]

Yes they do

Having written about chips extensively, and having designed a lot of microprocessors, I never heard anyone say it once. Because it's nonsense. Process node has absolutely nothing to do with how thick or thin anything is, let alone the chip. The fact is somebody thought that 10nm chips are actually thinner than 14nm chips, and is now trying to hide that fact by pretending he knew all along and I am just being pedantic.

 

[Jakewilk]

Having written about chips extensively, and having designed a lot of microprocessors, I never heard anyone say it once. Because it's nonsense. Process node has absolutely nothing to do with how thick or thin anything is, let alone the chip. The fact is somebody thought that 10nm chips are actually thinner than 14nm chips, and is now trying to hide that fact by pretending he knew all along and I am just being pedantic.

It doesn’t take an expert to realize that there are no 7nm thick components in any devices anywhere. Just think about that for a a second. But anyway, no one cares about how thick the actual chip is, it’s all about the tech inside. It’s unfortunate that your expertise has made you unaware of how normal people talk about processors

 

[cmaier]

It doesn’t take an expert to realize that there are no 7nm thick components in any devices anywhere. Just think about that for a a second. But anyway, no one cares about how thick the actual chip is, it’s all about the tech inside. It’s unfortunate that your expertise has made you unaware of how normal people talk about processors

There actually may be 7nm components on a 7nm processor. W/L = 1 is rare, but not unheard of, particularly in analog circuits like PLLs. So I don’t know what you think “experts realize” but you aren’t making any sense.

By the way, chip thickness DOES matter, and a lot of work goes into, for example, SOI and thinning the back side. But whatever.