Discussion Samsung Exynos SoC thread

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Doug S

Diamond Member
Feb 8, 2020
3,083
5,316
136
The roadblock for SMIC is EUV.

But they can still pursue scaling technologies such as GAAFET and BSPD without EUV, right?

BSPDN yes, but not GAAFET.

Or rather, the main point to doing GAAFET is that FinFET starts to become leaky if you shrink them down too much - and shrinking them down too much is what is enabled by EUV. Without EUV they can't shrink their FinFETs enough that it is worth going to GAAFET, and there is little benefit to doing GAAFET except to fix the problem of leaky FinFETs.
 
Reactions: FlameTail

soresu

Diamond Member
Dec 19, 2014
3,689
3,026
136
and there is little benefit to doing GAAFET except to fix the problem of leaky FinFETs
GAAFET is just a classification of FETs, the upcoming Nanosheet/Ribbon FET/MBCFET being just one variant.

Nanosheet does give better performance over finFET, but loses some areal density at iso transistor pitch.

Forksheet (next before CFET) is a more advanced GAAFET variant which fixes the density issue at iso pitch, while giving even higher performance than Nanosheet.
 
Reactions: FlameTail

Gideon

Golden Member
Nov 27, 2007
1,983
4,907
136
GAAFET is just a classification of FETs, the upcoming Nanosheet/Ribbon FET/MBCFET being just one variant.

Nanosheet does give better performance over finFET, but loses some areal density at iso transistor pitch.

Forksheet (next before CFET) is a more advanced GAAFET variant which fixes the density issue at iso pitch, while giving even higher performance than Nanosheet.
Here is a nice illustration for that from this article


 

Thibsie

Golden Member
Apr 25, 2017
1,010
1,186
136
Weren't there rumours of deals between I tel and Samsung ? Intel might make a bit of money licensing their tech to Samsung.
 

Doug S

Diamond Member
Feb 8, 2020
3,083
5,316
136
Samsung spokesperson says Exynos 2600 is not cancelled, dispelling rumours that about it's cancellation due to the low yields of the Samsung 2nm node.

For it to make economic sense Samsung only has to have yields sufficient that the payment from the mobile division exceeds the cost of the raw materials (raw wafers, chemicals, etc.) plus keeping the SF2 fab line's lights on, staffed, etc. The development of the process, building/equipping the fab etc. is a sunk cost so while you sure want to make enough money per chip to amortize that, you at least don't want to be losing money against the variable per wafer cost. 10-20% yield is probably sufficient to cover those variable costs.

The problem is that if yields are that low they probably also have issues hitting target frequency and power. Let's say 1 in 100 chips was good enough to hit their targets, they can't set the bar that high and have an effective yield of only 1%. They'll probably have to lower their frequency target and/or increase their power target, so it may not even be as good as TSMC's N3 family even if on paper it is a better process.

I suppose if Samsung wanted to win a pissing contest they could sell those 1 in 100 at superior performance levels in extra high end phones (some sort of Ultra + or in their low volume premium folding phone line) so they might have a chance to win at benchmarks against the competition.
 

FlameTail

Diamond Member
Dec 15, 2021
4,384
2,756
106
From The Elec, via Jukanlosreve:

"Samsung Held a Closed-Door Meeting with Korean Journalists, Summarized by The Elec on YouTube

I’ll translate this for you! Please credit the secondary source.

News: Exynos 2500 to Power Galaxy Foldable Phones

There have already been speculations from various benchmark sites that Samsung was testing the Exynos 2500. The key question was: Where will the Exynos 2500 be deployed? Previous reports suggested it would not be used in the Galaxy S series. It has now been confirmed that it will be featured in Samsung’s foldable lineup, such as the Galaxy Z Flip 7.

(Note: A Samsung representative said “Flip,” but it’s unclear whether it refers to the Flip 7 or Flip FE. Based on my findings, the Flip 7 will use the Exynos 2500, while the Flip FE will feature the Exynos 2400.)

The Significance of Exynos Entering Galaxy Foldables

The journalist interpreted the deployment of the Exynos 2500 in foldables as having two key implications:

1. Breaking Qualcomm’s AP Monopoly
Samsung’s Exynos has entered a segment traditionally dominated by Qualcomm’s Application Processors (APs). This marks Samsung’s first move to increase its chipset adoption in the foldable market.

2. Stabilization of Samsung Foundry’s 3nm Process

Featuring the Exynos 2500 in foldables indicates that Samsung’s 2nd-generation 3nm process has reached a degree of stabilization. However, this still requires final verification through actual products.

Production Volume and Market Limitations

Despite this, Samsung’s production situation remains challenging:

• Samsung’s total smartphone production for next year is estimated at 229.4 million units.
• Of these, the Galaxy Z Flip 7 accounts for only 3 million units (approximately 1%).
• The Flip FE has an even smaller production volume of 900,000 units.

The exclusion of Exynos from flagship devices like the Galaxy S series and its limited deployment in foldables reflect lingering market doubts about Samsung Foundry’s yield rates and process competitiveness.

Challenges in the Foundry Market

Samsung Foundry continues to face difficulties in the global market:
• TrendForce Report: Samsung’s market share dropped below 10% for the first time, reaching 9.3%.
• Competitors:
• TSMC maintains its dominant position with a 64.9% market share.
• China’s SMIC is closely trailing Samsung with 6%.

To narrow the technological gap with TSMC, Samsung must successfully implement next-generation processes like GAA (Gate-All-Around). However, current challenges include yield issues and cost burdens.

Internal Conflicts and Accountability

Reports have surfaced about internal conflicts between Samsung’s design and manufacturing divisions. A high-ranking official acknowledged there were clashes but emphasized that the divisions have now agreed to collaborate to generate synergy.

There were no questions about outsourcing Exynos production to TSMC during the meeting. Analysts suggest that if such a move were made public, it would be seen as admitting Samsung Foundry’s lack of competitiveness—a scenario that would have significant internal repercussions. (Note: Sorry for spilling the beans, Samsung!)

Samsung’s 3nm Process and Future Outlook

Samsung aimed to strengthen its competitiveness in the 3nm market through GAA technology but has yet to deliver clear results.
• Current Status: Qualcomm continues to collaborate with TSMC for AP production, while Samsung’s GAA process remains under scrutiny for yield issues. (Note: The Elec’s editor described the joint next-generation AP development between Samsung and Qualcomm as “facing difficulties.”)
• Short-Term Strategy: Samsung is likely to focus on stabilizing its 4nm process, which has reportedly achieved improved yields and seen increased demand from clients.
• Long-Term Strategy: Samsung must secure competitiveness in 2nm and 3nm nodes to remain a viable player. (Note: The Elec’s editor mentioned Samsung’s 4nm yield rates have stabilized, with growing client demand.)"
 
Last edited:
Reactions: Jukanlosreve

DZero

Senior member
Jun 20, 2024
768
290
96
And... even TSMC wants Samsung to be out of bussiness...

Samsung is facing low 20% yields on its second-generation 3nm GAA process, far below its 70% target, while TSMC’s 2nm process has reportedly achieved a 60% yield, according to sources referenced by Icsmart.
 

Doug S

Diamond Member
Feb 8, 2020
3,083
5,316
136
And... even TSMC wants Samsung to be out of bussiness...

So what's the deal with this claim that TSMC doesn't want to fab Exynos chips because it is worried Samsung could "gain insight into TSMC's trade secrets, particularly how it maintains an acceptable yield".

How is fabbing chips for another company giving away secrets about how you increase yield? If it does, why aren't they worried about fabbing chips for Intel for the same reasons?

Really scratching my head on that one.
 

DZero

Senior member
Jun 20, 2024
768
290
96
So what's the deal with this claim that TSMC doesn't want to fab Exynos chips because it is worried Samsung could "gain insight into TSMC's trade secrets, particularly how it maintains an acceptable yield".

How is fabbing chips for another company giving away secrets about how you increase yield? If it does, why aren't they worried about fabbing chips for Intel for the same reasons?

Really scratching my head on that one.
I want to think that could be fake but seeing the current industry how this is going... anything might end be possible sadly.
 

soresu

Diamond Member
Dec 19, 2014
3,689
3,026
136
Plot twist: happens the opposite and SMIC manages to fully have a 7nm process that is cheap enough.
Dunno about cost, but I saw news that they had managed to wring a 5nm process out of DUV machines some months back, which prompted the US govmt to throw another hissy fit at ASML and ban China from new DUV litho machines too.
 

LightningZ71

Platinum Member
Mar 10, 2017
2,077
2,525
136
Late model DUV with multi patterning can get you roughly "5nm" specs. It just takes a lot of process steps.
 

Doug S

Diamond Member
Feb 8, 2020
3,083
5,316
136
Late model DUV with multi patterning can get you roughly "5nm" specs. It just takes a lot of process steps.

It is roughly the equivalent of TSMC's N7, but getting N5 due to the improved DUV. TSMC was able to get high yields and good cost (at least compared to newer EUV nodes) so there is nothing technical preventing SMIC from getting there.

Their volume is going to be limited by how many of those nicer DUV machines they were able to buy before the window closed.

I read somewhere that China is doing a lot of research into using free electron lasers for EUV light. It seems to me that might be a technically better/cheaper pathway than how ASML's EUV machines work, only because of how hideously complex and Rube Goldberg those are. But that "seems to me" is based on the end solution, without any clue how much effort would be involved in reaching that point.

My worry is that if does prove to be doable China will go from behind to ahead in short order because it would be by its nature better able to hit smaller wavelengths than ASML's technology. There needs to be some research in the west on this, but I doubt that's happening. At least not with anything like the resources that were thrown at EUV for ~20 years before ASML was able to bring their solution to market. Of course there are other technologies like self assembly and nanoimprint that may bear fruit and you don't want to count those out (at least there is a big company with the resources to make nanoimprint work trying to do so)
 

DZero

Senior member
Jun 20, 2024
768
290
96
It is roughly the equivalent of TSMC's N7, but getting N5 due to the improved DUV. TSMC was able to get high yields and good cost (at least compared to newer EUV nodes) so there is nothing technical preventing SMIC from getting there.

Their volume is going to be limited by how many of those nicer DUV machines they were able to buy before the window closed.

I read somewhere that China is doing a lot of research into using free electron lasers for EUV light. It seems to me that might be a technically better/cheaper pathway than how ASML's EUV machines work, only because of how hideously complex and Rube Goldberg those are. But that "seems to me" is based on the end solution, without any clue how much effort would be involved in reaching that point.

My worry is that if does prove to be doable China will go from behind to ahead in short order because it would be by its nature better able to hit smaller wavelengths than ASML's technology. There needs to be some research in the west on this, but I doubt that's happening. At least not with anything like the resources that were thrown at EUV for ~20 years before ASML was able to bring their solution to market. Of course there are other technologies like self assembly and nanoimprint that may bear fruit and you don't want to count those out (at least there is a big company with the resources to make nanoimprint work trying to do so)
Even if they reach Samsung's tier, is enough to cause problems to ASML big time.
 

moinmoin

Diamond Member
Jun 1, 2017
5,203
8,365
136
Chinese players would shift to those process to save money. Also other countries will see cheaper options and move on too.
For the record, as of the end of Q3 2024 their order backlog sits at €36 billion. (And let's see how much that changed since at the Q4 2024 earnings call on January 29th.) You are essentially asking ASML to reenact following meme

 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |