TSMC catching up with Intel on leading edge process nodes

[Arachnotronic]

Intel's been getting jumps here and there and keeping ahead of the pack by being first to market with new technologies. If they fail to continue that trend, the rest of the pack will inevitably catch up. They need to go to 450 mm already.

450mm is more about cost (getting far more chips per wafer) than about technological prowess. Also, 450mm is such a big thing, no company can go at it alone; it will need to be a whole industry thing.

Intel's density advantage isn't really relevant if they can't keep their fab utilization (and yields) up. Likewise, TSMC and Samsung have to do the same if they want to keep up.

Fab utilization is merely a function of how good you were at planning out your capacity builds several years in advance

 

[III-V]

450mm is more about cost (getting far more chips per wafer) than about technological prowess. Also, 450mm is such a big thing, no company can go at it alone; it will need to be a whole industry thing.

Right, but they go hand in hand, really. Frankly, cost is of higher importance, at least over the shorter term.

Also, it doesn't really matter if it's industry wide or not -- Intel has the most to gain from it.

Fab utilization is merely a function of how good you were at planning out your capacity builds several years in advance

As well as executing internally, both with your own product development, and with attracting new revenue streams.

Frankly, for a company that used to revolve around the idea of "only the paranoid survive," Intel has gotten incredibly fat and lazy. I guess the good news is that they can weather basically any storm, but they're opening themselves up to attack.

 

[Arachnotronic]

Frankly, for a company that used to revolve around the idea of "only the paranoid survive," Intel has gotten incredibly fat and lazy. I guess the good news is that they can weather basically any storm, but they're opening themselves up to attack.

I don't think it's a question of laziness...

 

[III-V]

I don't think it's a question of laziness...

I don't either, but that wasn't my point anyway.

 

[Fjodor2001]

It's possible if the economy falls into a recession.

How would that be any different for Intel? They are not immune to recession and would be affected in the same way. So maybe 10 nm will never materialize at Intel either due to a recession. Is that what you're saying?

The design costs at "10" presumably are going to be even more expensive at that lower node. If you don't have the demand to justify it, it won't be done.

Are you suggesting Apple/Samsung/Qualcomm/... does not have the R&D money to design chips on 10 nm!? Got any source backing up that claim?

Apple has shown that they will pay for access to smaller nodes... within reason of course.

Yes, as has many others.

Also, you "forgot" to answer these questions:

You have to remember that nobody is going to use "10" nm unless it's comparable or cheaper than "14"/"16" nm except for maybe Apple.

So Samsung would just skip 10 nm (e.g. for Galaxy series) despite being able to produce such chips? Why would they do that!? They would just let their 10 nm capable factories sit idle for a couple of years? Please explain how that makes sense.

So maybe the 2018 iPhone will use it but it'd be 2019 or 2020 before you saw any products from nVidia or AMD for instance.

Source? Or just speculation? Why don't you speculate about Intel not releasing 10 nm products until 2020 too? What makes Intel different compared to TSMC/Samsung/GF? Are you expecting the latter to just let their 10 nm capable factories sit idle for several years, not producing anything!?

 

[jpiniero]

How would that be any different for Intel? They are not immune to recession and would be affected in the same way. So maybe 10 nm will never materialize at Intel either due to a recession. Is that what you're saying?

Intel is not immune. But they still have the volume and the server sales to make their 10 nm work. Below that is still very much a question.

Are you suggesting Apple/Samsung/Qualcomm/... does not have the R&D money to design chips on 10 nm!? Got any source backing up that claim?

If the numbers don't justify it, it won't happen.

Also, you "forgot" to answer these questions:

I don't need to answer all of your little questions.

 

[Fjodor2001]

Intel is not immune. But they still have the volume and the server sales to make their 10 nm work. Below that is still very much a question.

Any sources to back up the claim that Intel is different in the regard compared to all other companies?

If the numbers don't justify it, it won't happen.

Again, what are your sources for that the other companies mentioned do not have sufficient R&D capital to design chips on 10 nm? Sounds like your just doing the usual incorrect FUD speculation without anything to back up your claims.

I don't need to answer all of your little questions.

Of course. But I think your lack of answer is an answer in itself...

You simply skip answering because you know you are wrong and don't have anything to back up your claims. So far we've just seen tons of controversial claims from you, but no sources to back it up.

 

[Fjodor2001]

Intel 10nm = 0.0312µm²F / 0,0272µm²F (with assumed 1.84x scaling like 22->14 high density)

Do we have any source confirming the 1.84x scaling from Intel 14->10 nm, or is that just a guess?

Since Intel had severe yield issues on 14 nm, couldn't it be that they want to play it safe on 10 nm and thus go for a less dense 10 nm process than scaling it 1.84x compared to 14 nm?

@witeken: Did you have any source for this?

 

[ShintaiDK]

Samsungs 10nm SRAM is 0.040um2 for HD and 0.049um2 for HC. 0.63x from 14nm.

Intels SRAM is 0.049um2 with HD. So Samsungs 10nm is 0.82x of that.

http://www.eetimes.com/document.asp?doc_id=1328272

 

[Fjodor2001]

Samsungs 10nm SRAM is 0.040um2 for HD and 0.049um2 for HC. 0.63x from 14nm.

Intels SRAM is 0.049um2 with HD. So Samsungs 10nm is 0.82x of that.

http://www.eetimes.com/document.asp?doc_id=1328272

Yes, but the 0.049µm² you mention is for Intel's 14 nm process. I was looking for a source that Intel's 10 nm will be "0.0312µm² / 0,0272µm² (1.84x scaling)" as witeken said. So far I've not seen any source for that.

Also, as I wrote before:

"Since Intel had severe yield issues on 14 nm, couldn't it be that they want to play it safe on 10 nm and thus go for a less dense 10 nm process than scaling it 1.84x compared to 14 nm?"

I.e. I'm not sure we can just assume Intel's 14->10 nm transition will scale as much as 22->14 nm did. Hence I'd like to see a source confirming it.

 

[jpiniero]

Any sources to back up the claim that Intel is different in the regard compared to all other companies?

The revenue that Intel makes?

Again, what are your sources for that the other companies mentioned do not have sufficient R&D capital to design chips on 10 nm? Sounds like your just doing the usual incorrect FUD speculation without anything to back up your claims.

They do have the R&D capital. But it won't make sense financially to do it if their products don't have the volume to justify the higher expense of doing a "10" nm node compared to "14"/"16". You only have to look at the 20 nm planar nodes that virtually everyone ignored except for Apple - they couldn't justify the expense and had to wait until they added FinFets to it to get the cost at least comparable to the 28 nm nodes.

 

[ShintaiDK]

14/16nm node vs 28nm is a battle of cost. It will kill of companies due to cost structure. Both NVidia and AMD for example is doing it against all their (financial) will. But at the same time they cant afford to stay at 28nm. Its pretty much a lose/lose case.

 

[ShintaiDK]

Without Apple, its hard t see who can pay for new nodes at TSC/Samsung. Apples financials are coming up soon. And if their outlook reflect what the production chain reports with a 30% cut in Q1. Then the outlook for new nodes from TSMC and Samsung suddenly looks very weak.

At the same time for Intel, the constant shrinking PC market will do its course there.

If the current consumerism/growth/debt/zerorates crisis becomes a new Japanese disease copy as everything points to. Then we can forget new nodes for good in the very near future.

 

[ehume]

I'm surprised that most people assume that all TSMC claims will be transformed into reality, but the history tells us the opposite. . . . We are still waiting for dedicated 16nm GPUs from Nvidia/AMD.

I bought a 28nm GPU for the gamer in the house. Now she tells me she has gotten a game that won't play on it. I was initially intending to wait for the 20nm GPU's to buy a new, quieter card. But that didn't happen. Now I'm waiting for a 16nm card. Still waiting. Still waiting.

 

[raghu78]

14/16nm node vs 28nm is a battle of cost. It will kill of companies due to cost structure. Both NVidia and AMD for example is doing it against all their (financial) will. But at the same time they cant afford to stay at 28nm. Its pretty much a lose/lose case.

What a load of crap. The fabless industry will benefit hugely from TSMC and other foundries push to future nodes like 10nm, 7nm and 5nm. In fact the fabless companies are going to get a golden opportunity to compete against Intel using competitive process nodes. This was not the case in the past. TSMC 7nm and Intel 10nm products will face off in 2018 and 2019 and we will see the results.

 

[ShintaiDK]

What a load of crap. The fabless industry will benefit hugely from TSMC and other foundries push to future nodes like 10nm, 7nm and 5nm. In fact the fabless companies are going to get a golden opportunity to compete against Intel using competitive process nodes. This was not the case in the past. TSMC 7nm and Intel 10nm products will face off in 2018 and 2019 and we will see the results.

Good luck with that. Specially in a shrinking market and a crisis.

 

[Vesku]

What a load of crap. The fabless industry will benefit hugely from TSMC and other foundries push to future nodes like 10nm, 7nm and 5nm. In fact the fabless companies are going to get a golden opportunity to compete against Intel using competitive process nodes. This was not the case in the past. TSMC 7nm and Intel 10nm products will face off in 2018 and 2019 and we will see the results.

Nvidia, a high margin company was already grumbling at the increased wafer cost going from 40nm to 28nm. It's going to be rough going for the fabless industry to pass the rising costs along to their customers or attempt to keep costs similar which means providing much lower than expected performance improvements for each new generation of product.

 

[DooKey]

What a load of crap. The fabless industry will benefit hugely from TSMC and other foundries push to future nodes like 10nm, 7nm and 5nm. In fact the fabless companies are going to get a golden opportunity to compete against Intel using competitive process nodes. This was not the case in the past. TSMC 7nm and Intel 10nm products will face off in 2018 and 2019 and we will see the results.

Saved for later. Much like your fiji predictions I expect this one to be equally as bad.

 

[Arachnotronic]

What a load of crap. The fabless industry will benefit hugely from TSMC and other foundries push to future nodes like 10nm, 7nm and 5nm. In fact the fabless companies are going to get a golden opportunity to compete against Intel using competitive process nodes. This was not the case in the past. TSMC 7nm and Intel 10nm products will face off in 2018 and 2019 and we will see the results.

IMO, only the richest/highest volume customers will move at that pace. 14/16nm foundry has been in production for quite some time but it won't be until mid-2016 that we see the first 14nm GPUs.

 

[frozentundra123456]

That may or may not be true, but even if it is, aren't those richest/highest volume companies the ones intel ultimately has to compete with in mobile? That said, lets see if TSMC can actually meet that shall we say....optimistic roadmap.

 

[JDG1980]

Nvidia, a high margin company was already grumbling at the increased wafer cost going from 40nm to 28nm. It's going to be rough going for the fabless industry to pass the rising costs along to their customers or attempt to keep costs similar which means providing much lower than expected performance improvements for each new generation of product.

Of course Nvidia was complaining about costs. All companies try to negotiate their suppliers to a lower price point. There's nothing unusual about that. It doesn't prove anything.

And this is why competition in the GPU industry is a good thing. If AMD was out of business, Nvidia could decide to stagnate on 28nm indefinitely (or at least until Intel caught up with iGPUs). With AMD in the picture, Nvidia has no choice but to move to FinFET or be outcompeted, just as they had to swallow the cost of 28nm or else cede leadership in the GPU sector.

 

[JDG1980]

They do have the R&D capital. But it won't make sense financially to do it if their products don't have the volume to justify the higher expense of doing a "10" nm node compared to "14"/"16".

You're assuming the costs will be substantially higher, and will remain that way. My understanding was that the jump from planar to FinFET was the big hurdle for the foundries, and once that was crossed, the next die-shrink shouldn't be as agonizing a wait as this one was. Of course, there will be another wall further down the line, but 16/14nm->10nm should be a fairly routine die shrink as these things go. Will TSMC and Samsung want to recoup their investment? Of course. Will smartphone SoCs be the first products before GPUs come along? Sure. But does that mean we'll never see 10nm GPUs? No, it doesn't - assuming both AMD and Nvidia are still in the business, they will both have to go there, or risk the other one doing so and being stuck behind with an obsolete and uncompetitive lineup.

You only have to look at the 20 nm planar nodes that virtually everyone ignored except for Apple - they couldn't justify the expense and had to wait until they added FinFets to it to get the cost at least comparable to the 28 nm nodes.

The TSMC 20nm planar process was a mess - there were reasons other than cost that it wasn't ever used for GPUs and was skipped even by most smartphone SoC manufacturers. For instance, it was reported that in addition to poor yields, power characteristics made the process completely unsuited for GPUs. This article is informative:

The essential difficulty of the 20 nm planar node appears to be a lack of power scaling to match the increased transistor density. TSMC and others have successfully packed in more transistors into every square mm as compared to 28 nm, but the electrical characteristics did not scale proportionally well. Yes, there are improvements there per transistor, but when designers pack in all those transistors into a large design, TDP and voltage issues start to arise. As TDP increases, it takes more power to drive the processor, which then leads to more heat. The GPU guys probably looked at this and figured out that while they can achieve a higher transistor density and a wider design, they will have to downclock the entire GPU to hit reasonable TDP levels. When adding these concerns to yields and bins for the new process, the advantages of going to 20 nm would be slim to none at the end of the day.

We know that AMD tested 20nm GPU designs (I think it even got as far as tapeout), but they never made it to market. If AMD already paid the big cost of design and tapeout up-front, then it would make no sense for them not to release the products if they were at all viable. Just having higher wafer costs wouldn't have changed this. Don't you think they would rather have charged premium prices for new, smaller chips while Nvidia was still on 28nm? Don't you think Apple would have liked something better than Cape Verde (a 2012 GPU) to put in their 2015 MacBook Pro? Do you think the 300-series across-the-board rebrands for 2015 were Plan A?

20nm was a dud of a process. AMD and Nvidia wanted to use it for GPUs, but simply couldn't.

 

[AtenRa]

Nvidia, a high margin company was already grumbling at the increased wafer cost going from 40nm to 28nm. It's going to be rough going for the fabless industry to pass the rising costs along to their customers or attempt to keep costs similar which means providing much lower than expected performance improvements for each new generation of product.

That may be true for the CPU consumer market but not for the dGPU market. Both AMD and NVIDIA changed their policy at 28nm and raised prices significantly over 40nm and yet the market (gamers mostly) continued to pay for new hardware at much higher prices.

As for the CPU Server segment, they need more 14nm high perf/watt products the next 2-3 years for data centers and cloud and they will pay for them.

One last thing, 14nm may be more expensive than 28nm but selling a faster product with a lot smaller die size and more than half the power consumption over 28nm will still bring you profit. This apply for both the CPU and GPU markets. We may see a 14/16nm 250mm2 GPU in 2016 outperform a 550mm2 28nm GPU (by 10-20%). Selling it at $750 will still make you a nice profit close to what you had at 28nm.

 

[prtskg]

IMO, only the richest/highest volume customers will move at that pace. 14/16nm foundry has been in production for quite some time but it won't be until mid-2016 that we see the first 14nm GPUs.

Since Samsung's 14nm Lpp reached volume production few days ago, I don't see Q2/Q3 2016 as bad point for new gpu, especially when GF will trail Samsung and that is what AMD is using.

 

[Fjodor2001]

The revenue that Intel makes?

Which proves what exactly? And how?

They do have the R&D capital. But it won't make sense financially to do it if their products don't have the volume to justify the higher expense of doing a "10" nm node compared to "14"/"16". You only have to look at the 20 nm planar nodes that virtually everyone ignored except for Apple - they couldn't justify the expense and had to wait until they added FinFets to it to get the cost at least comparable to the 28 nm nodes.

Here's a reality check for you:

http://www.techtimes.com/articles/1...017-with-8-gb-ram-support-and-more-tricks.htm

Qualcomm Snapdragon 830 To Come In Early 2017

The latest leak reveals that the new processor, dubbed Snapdragon 830, will support a massive 8 GB of RAM and is expected to come with an enhanced Kryo custom architecture. It will also be built based on Samsung's 10nm process.

Users can expect to see the first set of devices to come equipped with the new Snapdragon 830 in 2017.

Intel is 4-5 years ahead of the competition on 10 nm? No other company in the world will have the R&D budget to design 10 nm chips? No other company will release 10 nm chips, ever, because they will be hit with a recession that only Intel is capable of dealing with? If any company does that regardless, they won't have products on the market until 2020 anyway?

Are you stills sticking with those claims? Really...!?