Discussion Foundry Node advances (TSMC, Samsung Foundry)

DisEnchantment

Senior member
Mar 3, 2017
499
822
106
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.

1587737990547.png
N7 performance is more or less understood.
1587739093721.png

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

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.


1587739615344.png

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.
 

Ajay

Diamond Member
Jan 8, 2001
6,821
2,197
136
Are there any significant semiconductor companies fabbing on Samsung's latest nodes? They are spending allot of money, just don't know if they are attracting any volume.
 

DisEnchantment

Senior member
Mar 3, 2017
499
822
106
Are there any significant semiconductor companies fabbing on Samsung's latest nodes? They are spending allot of money, just don't know if they are attracting any volume.
It is a necessity, they need to diversify, can't rely only on DRAM and NAND business, extremely profitable though it may be.
IBM, NVIDIA, Qualcomm (765) ... that is what I know so far. Samsung LSI is guaranteed anyway and volumes are quite big especially now with Exynos Auto/Base Stations devices and others also contributing to revenue.
 
  • Like
Reactions: Tlh97 and Ajay

DisEnchantment

Senior member
Mar 3, 2017
499
822
106
From Samsung's Quarterly result
Samsung officially declared they are starting 5LPE HVM this quarter.
In the second quarter, the Company aims to expand EUV leadership, beginning with the start of mass production of 5nm products, while closely monitoring the uncertain market situation caused by COVID-19.
I have also read rumors there is a new reworked Exynos 992 fabbed on 6LPP for the Note 20 which is addressing many of the shortcomings of the 990 to be released before year end.

The 5LPE HVM start would probably take a couple of quarters which is usual for Samsung, and they usually refresh their SoCs by year end for the next devices on Feb every year.
But this is unexpected news for me, was expecting 5LPE next year.

On the same note rumors are floating around in korean forums about a next gen device with RDNA2 and I imagine this could be fabbed on 5LPE. AMD is not involved in this because they only provided IP and not physical design, routing etc.
 

DisEnchantment

Senior member
Mar 3, 2017
499
822
106
There is a article on SA about TSMC losing process tech leadership to Intel
Which was debunked by SemiWiki(Scotten Jones)
Interesting take away is below

For 2020, TSMC is the clear leader. N5 is shipping in volume in this quarter [https://www.tsmc.com/uploadfile/ir/quarterly/2020/1OLf2/E/TSMC 1Q20 transcript.pdf]


1588200977840.png
By 2021, Samsung will start HVM for 4LPE but that is not a significant node.
However by 2022, Intel is still going to lag behind TSMC again, with a massive 50% density advantage for TSMC.
1588200918412.png
In terms of capex and investment, TSMC's expects around 16 Billion USD for 2020, about the same like Intel if not more. And consider the fact that TSMC dedicates everything to Process, interconnect and packaging R&D unlike Intel.

With 3GAA/N3/Intel 7nm the density gains are massive, But with such density the wild card here is performance.
TSMC is being conservative, going with FinFET and they know what works and if anything, we can be sure it will not be a failure node.
Samsung's MBCFET could be risky, but since they have a tapeout already at this point, they have 2 years to work on the process. Their target is 2021 but most likely will slip to 2022.
Intel is unknown at this point.

However the most interesting to me is GAA and if Samsung is to be believed it is about as radical as the jump from Planar to FinFET. The density disadvantage could be mitigated with a chiplet design. Important would be if they have good packaging and interconnect portfolio.
Samsung's Semi twitter account is very aggressive with promoting this ad

Wikichip/David Schor is slightly more conservative with the densities but the ratio are similar.
 

maddie

Diamond Member
Jul 18, 2010
3,065
1,685
136
There is a article on SA about TSMC losing process tech leadership to Intel
Which was debunked by SemiWiki(Scotten Jones)
Interesting take away is below

For 2020, TSMC is the clear leader. N5 is shipping in volume in this quarter [https://www.tsmc.com/uploadfile/ir/quarterly/2020/1OLf2/E/TSMC 1Q20 transcript.pdf]


View attachment 20362
By 2021, Samsung will start HVM for 4LPE but that is not a significant node.
However by 2022, Intel is still going to lag behind TSMC again, with a massive 50% density advantage for TSMC.
View attachment 20360
In terms of capex and investment, TSMC's expects around 16 Billion USD for 2020, about the same like Intel if not more. And consider the fact that TSMC dedicates everything to Process, interconnect and packaging R&D unlike Intel.

With 3GAA/N3/Intel 7nm the density gains are massive, But with such density the wild card here is performance.
TSMC is being conservative, going with FinFET and they know what works and if anything, we can be sure it will not be a failure node.
Samsung's MBCFET could be risky, but since they have a tapeout already at this point, they have 2 years to work on the process. Their target is 2021 but most likely will slip to 2022.
Intel is unknown at this point.

However the most interesting to me is GAA and if Samsung is to be believed it is about as radical as the jump from Planar to FinFET. The density disadvantage could be mitigated with a chiplet design. Important would be if they have good packaging and interconnect portfolio.
Samsung's Semi twitter account is very aggressive with promoting this ad

Wikichip/David Schor is slightly more conservative with the densities but the ratio are similar.
Some of the comments in that SA piece were brutal, justifiably so, in my view.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
20,137
7,637
136
Some of the comments in that SA piece were brutal, justifiably so, in my view.
Regardless of the article, here are the CURRENT facts. Intel has lots of problems with their current fabs. TSMC is executing fine with a superior tech.

Do you believe in the current tech ? or an article about what MIGHT be ? Based on Intels problems for the past 2 years, I say I have no faith in what they MIGHT do.
 

OriAr

Member
Feb 1, 2019
63
35
51
Regardless of the article, here are the CURRENT facts. Intel has lots of problems with their current fabs. TSMC is executing fine with a superior tech.

Do you believe in the current tech ? or an article about what MIGHT be ? Based on Intels problems for the past 2 years, I say I have no faith in what they MIGHT do.
Intel's execution has been fine since Swan and Murthy cleaned house in the manufacturing group, that and the fact they actually seem to have learned from mistakes make me think they'll do an OK job of executing their manufacturing roadmap from now on. (At the very least no 3 year delays like 10nm)
The future? Who knows, both TSMC and Intel might very well slip behind their roadmaps, past 2022 I don't think anything can be taken for granted, it is not a secret that the challenges now are far more difficult and navigating around them might lead to delays, and that's without any COVID-19 hits.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
20,137
7,637
136
Intel's execution has been fine since Swan and Murthy cleaned house in the manufacturing group, that and the fact they actually seem to have learned from mistakes make me think they'll do an OK job of executing their manufacturing roadmap from now on. (At the very least no 3 year delays like 10nm)
The future? Who knows, both TSMC and Intel might very well slip behind their roadmaps, past 2022 I don't think anything can be taken for granted, it is not a secret that the challenges now are far more difficult and navigating around them might lead to delays, and that's without any COVID-19 hits.
How do you know their execution has been file ? what evidence do we have ? NONE
 

DrMrLordX

Lifer
Apr 27, 2000
15,494
4,281
136
@Markfw

Tiger Lake is the first possible evidence that Intel has finally cleaned up a bad 10nm process. They still have a lot to prove with 7nm.
 

ThatBuzzkiller

Senior member
Nov 14, 2014
906
108
106
It is a necessity, they need to diversify, can't rely only on DRAM and NAND business, extremely profitable though it may be.
IBM, NVIDIA, Qualcomm (765) ... that is what I know so far. Samsung LSI is guaranteed anyway and volumes are quite big especially now with Exynos Auto/Base Stations devices and others also contributing to revenue.
The days of NAND flash memory being a profitable business anymore are over with the imminent arrival YMTC because they're going to flood the market with relatively high-end and cheap NAND modules unless Samsung employees are willing to take a 50% pay cut ?

That just leaves them with their DRAM business which is pretty good all things considered since it'll stay as a defacto triopoly between Micron, Samsing, and SK Hynix for at least several more years since it's just that much harder for any Chinese startup to catch up to them ...
 

DisEnchantment

Senior member
Mar 3, 2017
499
822
106
The days of NAND flash memory being a profitable business anymore are over with the imminent arrival YMTC because they're going to flood the market with relatively high-end and cheap NAND modules unless Samsung employees are willing to take a 50% pay cut ?
I think as of today, it is difficult to assess with what capacity YMTC is planning to do this.

As of December 2019, Samsung had the most installed wafer capacity with 2.9 million 200mm-equivalent wafers per month. That represented 15.0% of the world’s total capacity and about two-thirds of it was used for the fabrication of DRAM and NAND flash memory devices. Major construction projects underway include large new fabs at its sites in Hwaseong and Pyeongtaek, Korea, and in Xi’an, China.
Also Samsung is playing the same game like YMTC in China, and Samsung's Xi’an Fab in China is a mega fab designed for churning out NAND at highly competitive prices. Chinese Govt. currently is facilitating Samsung Foundry Engineers to operate freely despite Covid 19. There is even a special charter plane to ferry Samsung Engineers across to Xi’an and Korea.
Just Pyongtaek 1 alone has the capacity for almost half a million 300mm wafers each month. It is very hard to top that. Pyongtaek 2 is also planned to be online later this year.
Just P1, P2 and Xi’an and excluding other Samsung Fabs should be able to process in excess of 1 million 300mm wafers each month for NAND/DRAM, by end of 2020 no one is going to be remotely close to topping that.
Anyway, it is clear NAND and DRAM is not sustainable as the main source of revenue and they keep saying this, also the statement from their 1st quarterly report of 2020
In the second half, to address current uncertainties, the Foundry Business will focus on diversifying applications beyond mobile to include areas such as consumer and computing applications.
Hwaseong and Pyongtaek 2 are specifically designed for Logic besides DRAM/NAND.

Just to add, I am not trying to defend Samsung or any other foundry, each of those 5-10 Billion dollars fabs cost big money and they need to recover the capital expenditure. Sometimes as a consumer it seems they are price gouging, but when you think of it from an engineering perspective you can't blame them for it.
 

LightningZ71

Senior member
Mar 10, 2017
303
190
86
While TSMC's N3 won't be unusable by most any standards, I still have very deep concerns about the final performance numbers from N3 with respect to thermals and switching speed. Sticking with FinFET should at least insure that they will have a working product, as FinFET's are quite well understood. I am just not highly confident that they will be able to extract enough heat from them per unit time to accommodate the required switching speeds at that node to be performance competitive. Granted, there are significant hurdles for all of the competition at that node, so my concerns there may not be market relevant. It may also be true that, given the inherent density gains from going to such a tiny node, they can relax their density goals and still achieve the needed performance without taking too much of a hit to productivity, as I also think that chips won't be able to shrink too much and still be able to achieve the I/O density that is needed. In other words, you still need to route thousands of pins into the chip. At some point, you can't shrink the chip any more and still get all the pins connected to it in a manner that keeps individual pin signal integrity to a sufficient level.
 

Gideon

Senior member
Nov 27, 2007
780
1,108
136
I have a feeling TSMC 3nm is something similar to their 20nm process which in itself was mostly useless (almost 0 perf/power improvement compared to 28nm but more expensive) but ended up being a good pipe cleaner for their highly succesful 16nm (which was the exact same node with same density but with finfets).

3nm looks to be something similar to 2nm (which I bet is similar but with GAA)
 

moinmoin

Golden Member
Jun 1, 2017
1,568
1,425
106
Some of the comments in that SA piece were brutal, justifiably so, in my view.
Absolutely.

The user "SemiWiki.com" mentioned an interesting detail in the comments:
"The other thing you should know is that if a customer is taping out leading edge designs at multiple foundries (NVIDIA, QCOM, Intel, etc...) they do not get early access to TSMC processes. AMD made a very wise choice in working only with TSMC at 5nm. When AMD 5nm chips come out you will see."
We already know AMD does a lot of silicon level optimization, so this is intriguing info if true.
 

maddie

Diamond Member
Jul 18, 2010
3,065
1,685
136
Absolutely.

The user "SemiWiki.com" mentioned an interesting detail in the comments:
"The other thing you should know is that if a customer is taping out leading edge designs at multiple foundries (NVIDIA, QCOM, Intel, etc...) they do not get early access to TSMC processes. AMD made a very wise choice in working only with TSMC at 5nm. When AMD 5nm chips come out you will see."
We already know AMD does a lot of silicon level optimization, so this is intriguing info if true.
Wondering how this applies to Nvidia's reputed use of both TSMC and Samsung fabs for next gen products.
 

DisEnchantment

Senior member
Mar 3, 2017
499
822
106
Wondering how this applies to Nvidia's reputed use of both TSMC and Samsung fabs for next gen products.
If this is true, and I believe it could be true, it is quite sad, because NVIDIA cannot get enough wafers even if they wanted to (they could be going to Samsung for cost but it is not like they could go all TSMC anyway).
As for AMD, to partner exclusively with TSMC they need to consider that TSMC cannot supply them enough, should there be demand. Not sure if AMD has the cash to tape out at multiple fabs notwithstanding.

In my opinion, one of the many reasons for the slow ramp up for AMD at various OEMs could be the perception with AMD, in that the the big operators are wary of AMD not having enough chips to supply and they don't want to be in a situation where they have an AMD ecosystem but AMD cannot supply enough chips when they need to scale. Same applies to to laptops and commercial OEMs.

TSMC is trying to address this issue no doubt. More capacity investment has been planned, but to enable the likes of AMD to make a proper dent in the marketplace, at least 60-80K wpm are needed for AMD alone. If AMD were to be in a leadership position in the CPU market, they are going to need 300K+ wpm. Intel for example has around 800K+ wpm

TSMC's and Samsung's successes are critical for proper competition in the Semiconductor industry. Unsurprisingly, long time ago, Piednoel calculated this, and he said even if AMD wins, it is not a critical threat to Intel because AMD can never have the capacity to supply the market. You can bet the sourcing departments and bean counters at various ODMs know this.
 

maddie

Diamond Member
Jul 18, 2010
3,065
1,685
136
If this is true, and I believe it could be true, it is quite sad, because NVIDIA cannot get enough wafers even if they wanted to (they could be going to Samsung for cost but it is not like they could go all TSMC anyway).
As for AMD, to partner exclusively with TSMC they need to consider that TSMC cannot supply them enough, should there be demand. Not sure if AMD has the cash to tape out at multiple fabs notwithstanding.

In my opinion, one of the many reasons for the slow ramp up for AMD at various OEMs could be the perception with AMD, in that the the big operators are wary of AMD not having enough chips to supply and they don't want to be in a situation where they have an AMD ecosystem but AMD cannot supply enough chips when they need to scale. Same applies to to laptops and commercial OEMs.

TSMC is trying to address this issue no doubt. More capacity investment has been planned, but to enable the likes of AMD to make a proper dent in the marketplace, at least 60-80K wpm are needed for AMD alone. If AMD were to be in a leadership position in the CPU market, they are going to need 300K+ wpm. Intel for example has around 800K+ wpm

TSMC's and Samsung's successes are critical for proper competition in the Semiconductor industry. Unsurprisingly, long time ago, Piednoel calculated this, and he said even if AMD wins, it is not a critical threat to Intel because AMD can never have the capacity to supply the market. You can bet the sourcing departments and bean counters at various ODMs know this.
Isn't TSMC pushing Capex aggressively. Will take some time, but with their present process lead, I'm sure they don't intend to be relegated to 2nd place again. Maybe the wafer capacity expansion will advance alongside AMD's expected growth in market share. Almost symbiotic really, assuming plans are realized.

In the X86 PC market in general, top to bottom, only AMD can give TSMC a huge increase in revenue. No other company can do this as Intel was so dominant with AMD being the tiny 2nd player and no others.
 

moinmoin

Golden Member
Jun 1, 2017
1,568
1,425
106
All those tables comparing the technical specs of each foundry processes with each other all miss one important detail, the resulting yield rate.
And the max achievable frequency (which directly correlates with performance like Intel shows itself with both 14nm and 10nm).
 

ASK THE COMMUNITY