Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/what-about-samsung.12823/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

What about Samsung?

prime007

Active member
I know the narrative lately has been about Intel...but what about Samsung? Their stated goal is to supersede TSMC in process technology by 2030. Samsung is investing $116 billion towards that goal, and South Korea has indicated that it will invest 1.5 trillion won ($1.34 billion USD) to assist Samsung (and SK Hynix) develop their next generation technology.

The tech media generally recognizes TSMC, Samsung and Intel as the only bleeding-edge foundries. With Intel currently slipping in process technology, it would seem like Samsung is the only viable competitor to TSMC for process technology leadership. However as Robert Maire pointed out, their previous 10nm and 7nm (yield rate is rumored to be about 30%) launches have been a disaster. Digitimes recently indicated that Samsung is struggling to improve their 5nm process yield. Moreover when Scotten Jones compared TSMC, Intel and Samsung...it seems the state of Samsung's process technology to be more or less equivalent to where Intel is now.

Samsung's decision to skip 4nm and go directly to 3nm GAAFET feels like a "bet the farm" move to me. It seems like a high risk bet for Samsung to bring a new technology into the market on-time AND with good yield. During a recent earnings call, TSMC chairman (Mark Liu) indicated that Samsung's 3nm GAAFET was only roughly equivalent to TSMC's current 5nm process node. TSMC's future (superior?) 3nm process is scheduled to be introduced around the same time as Samsung's 3nm. The only upside that I can see is that Samsung's engineers become well-versed in GAAFET that their 2nd-generation product will be superior to TSMC's future products going forward.

But what ultimately maybe more worrying for Samsung (and SK Hynix), is China's recent entry into the DRAM/NAND market. It's been speculated that China's state-sponsored/state-supported firms will cause a memory glut and result in lower profit margins and perhaps even a collapse of a current supplier. If this happens, Samsung may soon no longer be able to use their profits from their memory business to subsidize their foundry...and perhaps their only means to beat TSMC.

Those are my thoughts on Samsung....would appreciate hearing what others think regarding Samsung's future.
 
I liked the headline" Samsung skipping 4nm to Compete with TSMC". Shortly thereafter TSMC started talking about 4nm. Funny stuff.

Samsung skipping 4nm means that no customers signed up for it. Simple as that. Samsung is also having problems with 5nm yield so skipping 4nm is the right thing to do but let's not spin it into a good thing. It's never good when a foundry fails to deliver.

Here is Samsung's foundry problem in a nut shell: Samsung does not have the ecosystem loyalty that TSMC does and without a loyal ecosystem a foundry cannot compete with TSMC. Period. Without silicon proven EDA tools and IP customers cannot tape out with confidence. Without the trust that you can deliver wafers at a given yield and time frame you cannot compete with TSMC. That goes for Samsung, UMC, SMIC, and Intel.

The TSMC ecosystem of customers, EDA, IP, Services, and Equipment manufacturers is a force of nature and there is no stopping it. There will always be a need for a second foundry source and at the leading edge that is Samsung and Intel. But, until they can deliver on the expectations a foundry must set, Samsung and Intel has zero chances of being a first source provider, absolutely.
 
The market is too big for a single company. I just read that some ereaders were cancelled because of production problems.
 
Samsung made a statement yesterday: https://www.samsungfoundry.com/foun...yOut=homepageLayout&menuIndex=0504&blogId=280

Foundry Blog

[Statment] Regarding negative news coverage on our 5/4nm process development

2020-08-03

Recent media coverages on the delay of Samsung Electronics’s 5/4nm process development are inaccurate and have no factual basis. Please find our official statement regarding the state of our 5/4nm process technologies below.
  • Samsung 5nm process node
- Samsung has already started mass production of its 5nm EUV process in Q2 2020 and plans to ramp up volume production in 2H 2020 with an expanded customer base. The yield rates of 5nm process are being improved as planned.
  • Samsung 4nm process node
- 1st generation process development is on track. Samsung is also accelerating the development of 2nd generation 4nm process technology in support of a diverse set of advanced applications while at the same time strengthening our process portfolio.

- Additionally PPA (Power, Performance, Area) improvements in this 2nd generation of 4nm process, will further expand Samsung’s leadership in advanced process technology offerings.
 
Samsung made a statement yesterday: https://www.samsungfoundry.com/foun...yOut=homepageLayout&menuIndex=0504&blogId=280

Foundry Blog

[Statment] Regarding negative news coverage on our 5/4nm process development

2020-08-03

Recent media coverages on the delay of Samsung Electronics’s 5/4nm process development are inaccurate and have no factual basis. Please find our official statement regarding the state of our 5/4nm process technologies below.
  • Samsung 5nm process node
- Samsung has already started mass production of its 5nm EUV process in Q2 2020 and plans to ramp up volume production in 2H 2020 with an expanded customer base. The yield rates of 5nm process are being improved as planned.
  • Samsung 4nm process node
- 1st generation process development is on track. Samsung is also accelerating the development of 2nd generation 4nm process technology in support of a diverse set of advanced applications while at the same time strengthening our process portfolio.

- Additionally PPA (Power, Performance, Area) improvements in this 2nd generation of 4nm process, will further expand Samsung’s leadership in advanced process technology offerings.

Basically 5nm still needs improvement (even though it has same design rules as 7nm) and 4nm, while not planned to be skipped, is also not yet here.

Samsung's layout style tends to use higher pupil fill, so it will lose more yield from stochastics.
 
Last edited:
Can someone please tell me the difference between "mass production" and "volume production"? And when you say "started" and "plans to ramp up" prior to those statements what does that really mean?

And I will bet if you go back and look at Samsung's previous statements on 10nm and 7nm the difference between what they said in the press to what actually what happened is not equivalent. Ergo my previous statement on being a "Trusted Foundry".


Samsung 5nm process node
- Samsung has already started mass production of its 5nm EUV process in Q2 2020 and plans to ramp up volume production in 2H 2020 with an expanded customer base. The yield rates of 5nm process are being improved as planned.


"being improved as planned" is another one of those wiggle room statements. Very little trust in wiggle room statements.

- Additionally PPA (Power, Performance, Area) improvements in this 2nd generation of 4nm process, will further expand Samsung’s leadership in advanced process technology offerings.

Lot's of wiggle room here!
 
Last edited:
Can someone please tell me the difference between "mass production" and "volume production"? And when you say "started" and "plans to ramp up" prior to those statements what does that really mean?

Almost missed that doublespeak - what does it mean, indeed!

Regarding direct to 3nm, it was reported that the covid-19 delayed some equipment so that it would have to be 2022 instead of 2021, leaving room for 4nm next year:

https://www.digitimes.com/news/a20200406VL201.html:
Samsung unlikely to move 3nm to volume production until 2022: Samsung Foundry, the foundry operations of Samsung Electronics, has set its goal of moving its advanced 3nm process technology to volume production as early as 2021. But the Korean firm will probably have to reschedule it to 2022, as the coronavirus pandemic's impacts on logistics and transportation services are causing delays to deliveries of EUV and other critical production equipment, according to industry sources.
 
Here is another Samsung mystery. Why is Samsung VLSI counted as part of Samsung Foundry revenue? That would be like counting Intel chip revenue as Intel Foundry revenue making them the #1 foundry in the world!

Global Foundry Revenue 2H2020.jpg
 
A rumor has just popped up saying Samsung has lost its Qualcomm Snapdragon X60 modems and 875 processor orders due to 5nm production difficulties.
Source: GizmoChina
 
This started at 10nm when yield was single digit and QCOM had a serious wafer shortage. Samsung responded by selling QCOM good die rather than wafers to cover the bad yield. QCOM moved products to TSMC as a result and has been playing both sides of the foundry fence ever since. Samsung has better pricing but TSMC delivers wafers on time and within yield expectations, absolutely. Thus the whole TSMC trusted foundry theme.

Samsung Foundry had a great run on 14nm but has been struggling with yield ever since. The Samsung goal it seems is to be the first with leading edge technologies where as TSMC puts customers first. Just my opinion of course.


A rumor has just popped up saying Samsung has lost its Qualcomm Snapdragon X60 modems and 875 processor orders due to 5nm production difficulties.
Source: GizmoChina
 
Here is another Samsung mystery. Why is Samsung VLSI counted as part of Samsung Foundry revenue? That would be like counting Intel chip revenue as Intel Foundry revenue making them the #1 foundry in the world!

I read that as they are counting revenues for chip sales from Samsung Foundry to Samsung VLSI. (that makes sense to me) I did not read it as counting Samsung VLSI revenue as part of Samsung Foundry revenue.
 
https://www.digitimes.com/news/a20200406VL201.html:
Samsung unlikely to move 3nm to volume production until 2022: Samsung Foundry, the foundry operations of Samsung Electronics, has set its goal of moving its advanced 3nm process technology to volume production as early as 2021. But the Korean firm will probably have to reschedule it to 2022, as the coronavirus pandemic's impacts on logistics and transportation services are causing delays to deliveries of EUV and other critical production equipment, according to industry sources.

There was a a report from South Korea that Samsung will be selling parts (and design services and backend services) to Google & Cisco. (hard to the know the extent -- I'd guess limited engagement from each to start) Samsung is also doing business with Qualcomm, Nvidia, IBM.

So Samsung has an opportunity to become a credible 2nd place foundry to TSMC. But they have to yield & deliver.
 
My understanding is that Samsung counts internal logic customers as foundry customers. All logic fabs are organized under the Samsung Foundry Group so anything they manufacture is considered foundry revenue.

The tricky part for internal customers is how much is actually paid per wafer? Are they getting them at manufacturing cost? Or is there an average price? Again, lots of wiggle room here.


I read that as they are counting revenues for chip sales from Samsung Foundry to Samsung VLSI. (that makes sense to me) I did not read it as counting Samsung VLSI revenue as part of Samsung Foundry revenue.
 
Samsung just announced their Galaxy Note 20 models today. It was previously reported by ZDNet Korea that these models would feature their 5nm Exynos 992 processor. This supposed chip was to offer an 20% improvement on power efficiency and graphics processing.

Instead according to Ars Technica, Samsung will ship models featuring the Qualcomm Snapdragon 865+ to the US, China, Japan and Latin America. The models featuring the "disappointing" Exynos 990 will ship to Europe, India and the rest of the world. It's been noted that Samsung's Exynos chips "are far behind" what Qualcomm produces. Samsung fans have even started a petition titled "Stop selling us inferior Exynos phones!" Samsung, itself, appears none to happy with the performance of its Exynos processor. It pulled the Exynos version of the S20 from South Korea earlier this year which reportedly "humiliated" the Exynos team.
 
Of the two foundries, who is the the most advanced at SOFI using small chips? I feel that is going to be a significant product in the future. Any thoughts or comments appreciated.
 
Here is another Samsung mystery. Why is Samsung VLSI counted as part of Samsung Foundry revenue? That would be like counting Intel chip revenue as Intel Foundry revenue making them the #1 foundry in the world!

View attachment 198
Some interesting info:
1. From Samsung 2020Q2 PR news:
a. "System LSI reported lower earnings amid weakness in the mobile market while the Foundry Business saw a recovery in customer demand."
b." The Foundry Business achieved record quarterly and half-yearly revenue as customers’ inventory build-ups increased. The Company began mass production of 5nm products and is developing 4nm process technology. The Company also strengthened the foundation for expansion by starting investments in production lines in the Pyeongtaek complex. "
Source: https://news.samsung.com/global/samsung-electronics-announces-second-quarter-2020-results
2. From Samsung 2020Q2 financial result(attached), we can see 2020Q2 revenue of Semiconductor segment decreased( Q2 KRW 3.6t vs. Q1 KRW
4.5t). If comparing to Taiwanese Research data(Jun. 2020), it becomes ~20% short( in US$) of estimation.
3. If it is true that Qualcomm switched his 5nm foundry to TSM due to low yield in 5nm, then it will impact revenue in Q3 and after.
Referring to System LSI+ Foundry's revenue decrease but Foundry made record high, and it is due to "customers' inventory build-ups", what will be the message mean? I would say, this data can not be trusted. Samsung's semiconductor revenue might decrease in next quarter if without effective actions. Wondering who will be the "customers" to build-up inventory in foundry now?
Samsung 2020Q1 EC1.JPG
Samsung 2020Q2 EC1.JPG
 
Last edited by a moderator:
The Moore's Law is Dead YouTuber stated that he was told that Samsung's current yield rate on 7nm is about 80% (at approximately the 7:40 minute mark).

Quick notes:
1. Daniel also recently appeared on his YouTube video. I didn't see anyone mentioning it but you can view it here:
2. I didn't realize it until recently but SemiWiki also has a YouTube channel! You can visit it here: https://www.youtube.com/channel/UCHdjD8mzAU_nV3OJSCZY9pw or click the YouTube icon in the upper right-hand corner of the site.
 
Back
Top