Discussion Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel) - [2020 - 2025]

[DisEnchantment]

TSMC's N7 EUV is now in its second year of production and N5 is contributing to revenue for TSMC this quarter. N3 is scheduled for 2022 and I believe they have a good chance to reach that target.


N7 performance is more or less understood.


This year and next year TSMC is mainly increasing capacity to meet demands.

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

www.anandtech.com

For Samsung the nodes are basically the same from 7LPP to 4 LPE, they just add incremental scaling boosters while the bulk of the tech is the same.

Samsung is already shipping 7LPP and will ship 6LPP in H2. Hopefully they fix any issues if at all.
They have two more intermediate nodes in between before going to 3GAE, most likely 5LPE will ship next year but for 4LPE it will probably be back to back with 3GAA since 3GAA is a parallel development with 7LPP enhancements.

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

www.anandtech.com

Samsung's 3GAA will go for HVM in 2022 most likely, similar timeframe to TSMC's N3.
There are major differences in how the transistor will be fabricated due to the GAA but density for sure Samsung will be behind N3.
But there might be advantages for Samsung with regards to power and performance, so it may be better suited for some applications.
But for now we don't know how much of this is true and we can only rely on the marketing material.

This year there should be a lot more available wafers due to lack of demand from Smartphone vendors and increased capacity from TSMC and Samsung.
Lots of SoCs which dont need to be top end will be fabbed with N7 or 7LPP/6LPP instead of N5, so there will be lots of wafers around.

Most of the current 7nm designs are far from the advertized density from TSMC and Samsung. There is still potential for density increase compared to currently shipping products.
N5 is going to be the leading foundry node for the next couple of years.

For a lot of fabless companies out there, the processes and capacity available are quite good.

---------------------------------------------------------------------------------------------------------------------------------------------------


FEEL FREE TO CREATE A NEW THREAD FOR 2025+ OUTLOOK, I WILL LINK IT HERE

 

[DZero]

Oh no you don't actually know how worse Samsung is in the node game rn 🤣

yeah, but seeing Intel, leaves Samsung situation on a better light.

And this is bad. Depending all in TSMC is not good since is something happens, the development halts.

And yeah, Trump effed the chinese initially but seems that they might be the only option that is working if Rapidus fails.

 

[jpiniero]

I assume China will do what they do and steal all of TSMC's IP.

 

[511]

yeah, but seeing Intel, leaves Samsung situation on a better light.

And this is bad. Depending all in TSMC is not good since is something happens, the development halts.

And yeah, Trump effed the chinese initially but seems that they might be the only option that is working if Rapidus fails.

LoL we are really believing Returs the same sources that said Qualcomm is buying Intel. Intel is in JV with TSMC and Intel's Yield is low and what not ?

We had all kinds of BS Rumor last year and we are still falling for it amazing job everyone.

 

[igor_kavinski]

We had all kinds of BS Rumor last year and we are still falling for it amazing job everyone.

Enough people believe the BS for long enough and it becomes reality.

Manifestation, baby!

 

[511]

Enough people believe the BS for long enough and it becomes reality.

Manifestation, baby!

Yeah for those people 🤣🤣

On a sidenote On Semiwiki Forum

The truth of the matter is that Intel 18A is a great process but it missed the market window for the big fabless customers. Rather than chase 3rd and 4rth tier customers Intel may focus on 14A and 1st tier customers which makes sense. It is a similar effort with much bigger results. Go Intel!

Exclusive: Intel's new CEO explores big shift in chip manufacturing business

SAN FRANCISCO (Reuters) -Intel's new chief executive is exploring a big change to its contract manufacturing business to win major customers, two people familiar with the matter told Reuters, in a potentially expensive shift from his predecessor's plans. If implemented, the new strategy for...

semiwiki.com

 

[igor_kavinski]

Exclusive: Intel's new CEO explores big shift in chip manufacturing business

SAN FRANCISCO (Reuters) -Intel's new chief executive is exploring a big change to its contract manufacturing business to win major customers, two people familiar with the matter told Reuters, in a potentially expensive shift from his predecessor's plans. If implemented, the new strategy for...

semiwiki.com

It can work if LBT micromanages the hell out of the fab teams with a stick and makes them work day and night. Like TSMC.

 

[511]

It can work if LBT micromanages the hell out of the fab teams with a stick and makes them work day and night. Like TSMC.

Micromanaging won't solve a damm thing

 

[igor_kavinski]

Micromanaging won't solve a damm thing

In Intel's case, there is explicit need for that. They've been getting away with lies for far too long. They need a very strict taskmaster.

 

[DZero]

Even if Intel goes fabless, it will crash and burn. The core is really damaged

 

[511]

In Intel's case, there is explicit need for that. They've been getting away with lies for far too long. They need a very strict taskmaster.

it will make it worse lol

Even if Intel goes fabless, it will crash and burn. The core is really damaged

Yeah kind of their core strengths were design and manufacturing both have been brain drained and lost it's edge

 

[Io Magnesso]

And what if is succesful?

Now seeing this...

If so, I'd like to ask you a question, why can you trust it?
In the first place, Japan has never dealt with cutting-edge processes... So to speak, I was abandoned from the competition
It's quite a risk to deal with 2nm node all of a sudden, right? You know, right?
Considering this, at least Intel and Samsung are more reliable

 

[Io Magnesso]

I assume China will do what they do and steal all of TSMC's IP.

Well, TSMC has also had a scandal in the past that HUAWEI made a request to manufacture a processor via a company disguised as a company… However, they won't make mistakes like leaking IP.
That's exactly what if there was an invasion of Taiwan In preparation for the robbery of FAB and IP, It's a last resort I think you are prepared to destroy your own equipment. It's just for example, but

 

[igor_kavinski]

Yeah kind of their core strengths were design and manufacturing both have been brain drained and lost it's edge

More like brain washed.

Investors and board members are STUPID.

They hear some catchphrase and then destroy the company in more ways than one trying to make that catchphrase a reality.

My company's board heard cloud computing was great.

They spent MILLIONS. Almost two years of stressful testing in which I was the sole performance tester and kept telling the COO that the performance advantage wasn't that great. We could get cheaper and much more economical computing power by staying on-premises with the latest servers or going to a hybrid model. He didn't listen and probably didn't want to admit that some techie knew better than everyone else involved. Well, they are now paying $300K per year where before our monthly data center hosting cost was barely $15000 per year. And the performance benefit is nowhere to be found because Microsoft limits us to simply 7000 IOPS max in Azure DB whereas we could've had several hundred thousand IOPS with our own servers. Going with the business critical tier of Azure DB will almost triple the DB running expenses which the management is balking at paying. See how dumb people ruined everything?

Similarly, Intel board latched onto this extremely stupid idea of going all in on tile design, even for desktop. It could've worked fine for mobile chips and when they had ironed out the kinks, they could've migrated desktop chips to tile design too. BUT NO. Gotta have tiles. AT ALL COSTS.

Honestly, I would self-detonate if I could convert my anger into thermal energy.

 

[Io Magnesso]

It can work if LBT micromanages the hell out of the fab teams with a stick and makes them work day and night. Like TSMC.

fab should be automated
In fact, Intel is doing a project to automate semiconductor(Post-process) manufacturing with Japanese companies.

半導体後工程自動化・標準化技術研究組合 (SATAS)

And You don't have to work in a way that hurts your body like an Asian does.

In the latter half of the 20th century, the Japanese looked at the pain of unreasonable work and reflected on them...

How many years later the word 996 was born...
Taiwan is in a similar situation...
I don't know much about the situation in Korea...
The word 996, which was originally born in mainland China, can be applied not only to China but also to Taiwan and South Korea…
They didn't learn Japanese who failed in the past, so they make the same mistakes...
Humans are stupid after all...

It seems that TSMC is also struggling to secure local human resources due to differences in corporate culture and national culture at factories outside of Asia.
Don't seek unreasonable labor... (commandment)

Intel seems to be better than TSMC in terms of securing local human resources through global expansion.
Well, TSMC has just started making FABs outside of Taiwan, so there are some things that can't be helped...

 

[Io Magnesso]

More like brain washed.

Investors and board members are STUPID.

Right, I agree
In the end, It is swayed by the convenience of these people
Even Pat... LBT... Even Bob Swan...

 

[igor_kavinski]

fab should be automated

I'm talking about the R&D part. You absolutely need brilliant people working their asses off to make a leading edge node and work all of its kinks out. It's not something you give to a bunch of guys and expect them to return when "it's done". You give them strict deadlines and that would put pressure on them to work round the clock. That's what TSMC does. I don't know about other fabs but I'm guessing TSMC is the only one with at least 22/7 operations?

 

[Io Magnesso]

I'm talking about the R&D part. You absolutely need brilliant people working their asses off to make a leading edge node and work all of its kinks out. It's not something you give to a bunch of guys and expect them to return when "it's done". You give them strict deadlines and that would put pressure on them to work round the clock. That's what TSMC does. I don't know about other fabs but I'm guessing TSMC is the only one with at least 22/7 operations?

No matter how talented the human resource is, it's the same person as us.
Unreasonable work can damage your body
The victim of overwork has nothing to do with excellence or inferior…
People can break, there's no such thing as not breaking
If it's an unfamiliar stress environment and unstable situation, I can't even stand on the arena for competition
That's right, R&D people have to be desperate... However, you don't have to imitate the way of working in Asia that doesn't fit the Western style of work at all.
It's not bad to take a different approach from the way TSMC works. You don't have to imitate them
Intel is Intel It's better to develop a better way of working by improving the good points and crushing the bad points.

 

[igor_kavinski]

Intel is Intel It's better to develop a better way of working by improving the good points and crushing the bad points.

They can do that once they have crossed the river of fire towards absolute process leadership. They sat on their asses for too long and now they gotta pay with blood and sweat or become obsolete and irrelevant like IBM.

 

[Io Magnesso]

They can do that once they have crossed the river of fire towards absolute process leadership. They sat on their asses for too long and now they gotta pay with blood and sweat or become obsolete and irrelevant like IBM.

I don't even have time to die
Anyway, should think about it and put it into action.
Destroy your employees Or Create a new way of working

 

[igor_kavinski]

Destroy your employees Or Create a new way of working

If a charismatic leader (hoping LBT is one) tells his men (and women):

We are in an absolute state of emergency. Our very existence is threatened. All the good we can do in the future will be nullified if we don't take action NOW!

Who's up for some real hard work? Who wants to taste the glory of success? Who wants to be remembered as the employee of the company that made a miraculous resurgence from an almost unstoppable fate of fading into obscurity?

WHO WANTS TO CHANGE HISTORY????

WHO WANTS TO MAKE INTEL GREAT AGAIN????

I mean, it could work. Depends on the speech and the leader's oratorship

 

[Io Magnesso]

I don't even have time to die
Anyway, should think about it and put it into action.
Destroy your employees Or Create a new way of working

Also, let me add that
TSMC will increase the number of non-Asians (Taiwanese) employees in the future global development… However, more than half of the East Asians are still occupied.
Unlike that, Intel has a variety of races, of course because they are different. I think it's unreasonable to ask everyone to work like TSMC…

 

[Fjodor2001]

Too much lol-ing in this thread.

It's not good for us consumers if we don't have any competition.

At least I'm hoping for Intel 18A(P) to be competitive and available on time, so it moves progress along and prices down.

 

[DrMrLordX]

It's not good for us consumers if we don't have any competition.

While that's true, it's also important to realize that many foundries are . . . foundering.

Remember when Samsung saved GF by licensing 14LPE and (I guess) LPP? GF wound up making processes that were arguably better than what Samsung licensed to them initially. What if Intel, Samsung, GF, and possibly others had a cross-licensing agreement and formed a multinational fab consortium? Everybody gets to look at everybody else's patents and trade secrets, and then they make the best nodes they can with what's available.

Survive together or die alone. Take your pick!

 

[Hulk]

Could it be that...

1. 18A yields are so low that they can't fab their own parts AND be a foundry.

and/or

2. 18A is not performing up to expectations as a foundry part but will suffice for Intel parts.

Either is another ominous eventuality in a long line of ominious happenings for Intel.

 

[Tigerick]

Hoho, finally CFO is telling the truth: It is almost 100% failure rate for external customer. I am sure IFS will give best deal on wafer prices, yet there is virtually no customer committed. The reason is simple: the yields are "horrible"; as I mentioned before regarding Intel Internal Business, IFS could ask Intel Design to flex the design to improve yield at the cost of die area. These "technique" won't work with external customers cause PPAC matter the most to external customers.

Hereby, I could say 18A is a failure process, there is no chance IFS could compete with TSMC, period. Any hearsay well is just a hearsay....

As I speculated, with zero customer and zero commitment, Intel 18A is a complete flop process. And if people still not able to get what CFO said about 100% failure rate, well now that Intel might totally scrapped the plan to offer 18A to external customer. Are you getting it yet?

One of Gelsinger’s mantras seems to be that unquestioned leadership products bring unquestioned leadership margins for those products, and for Intel to execute, it needs to return to its days of old.

If I am BOD, I will fire Pat also due to his BS execution. He was trying to hide that facts IFS cannot provide good yield by going chiplet solution. Yes, chiplet is the future solution only if the monolithics design cannot sustain PPA any further. As I compared upcoming PTL-H and AMD's Gorgon Point: AMD able to offer 24 threads and 16 CU, meanwhile Intel PTL is having 12+4 threads and 12XE (Thanks TSMC!). Do you think PTL has a chance to perform faster than Gorgon Point? Now that Gorgon Point's NPU rumored to hit 65TOPS; 15 TOPS faster than PTL. And what process node AMD is using???

Also if you think AMD is sitting idle and let Intel catch up a bit, hoho let's wait for AMD's Soundwave...but that is topic for another thread.