10nm製程以材料的角度來看TSMC贏過三星

NQQegg wrote:
挑戰英特爾添火力,...(恕刪)


Samsung Plans 3nm Gate-All-Around FETs in 2021

https://www.eetimes.com/document.asp?doc_id=1333318

SANTA CLARA, Calif. — Samsung Electronics laid out plans to bring to mass production in 2021 the architectural successor to FinFETS, gate-all-around (GAA) transistors, at the 3nm node. The South Korean giant also reaffirmed plans to begin 7nm production using extreme ultraviolet (EUV) lithography in the second half of this year at its annual foundry technology forum here Tuesday (May 22).

GAA technology has been under development since the early 2000s by Samsung and other firms. GAA transistors are field-effect transistors (FET) that feature a gate on all four sides of the channel to overcome the physical scaling and performance limitations of FinFETs, including supply voltage.

Samsung's proprietary GAA technology, known as multi-bridge-channel FET (MBCFET), has been in development since 2002, according to Ryan Sanghyun Lee, vice president of market for Samsung Foundry. MCBFET uses a nano-sheet device to enhance gate control, significantly improving the performance of the transistor, according to the company.

Samsung said last year that it planned to use GAA transistors at the 4nm node starting in 2020. However, industry watchers expected GAA to be in production no earlier than 2022.

Samuel Wang, a vice president in foundry research at Gartner, said he had expected Samsung to have GAA transistors production ready some time in 2022. "It looks like they are moving faster than that," Wang said.

Kevin Krewell
Kevin Krewell
"The Samsung roadmap was aggressive," said Kevin Krewell, I already knew they were moving fast on EUV, but this also sets a high bar."

But, Krewell added, "It's still a ways out and schedules can slip."

Last June, IBM and its research alliance partners Samsung and Globalfoundries described the process they had developed for making 5nm GAA transistors based on stacked nanosheets at the 2017 Symposia on VLSI Technology and Circuits conference in Kyoto, Japan. Other chipmakers, including Intel and TSMC, are known to be developing their own versions of next-generation transistors beyond FinFET similar to GAA FETs.

SEM image of 5nm transistor with gate-all-around technology built by IBM and partners Samsung and Globalfoundries. Source: IBM


SEM image of 5nm transistor with gate-all-around technology built by IBM and partners Samsung and Globalfoundries.
Source: IBM
Samsung reiterated plans to begin using EUV lithography in mass production in the second half of this year with its 7nm Low Power Plus process. Samsung is expected to be the first chipmaker to put EUV — which the industry has been developing for many years — into commercial production. TSMC and Globalfoundries have announced plans to use EUV in commercial production starting in 2019.

While lithography tool vendor ASML and leading-edge chip makers have demonstrated the ability to overcome the source-power issue that plagued EUV development for years, the supporting technologies needed to deploy EUV in commercial production are still being developed and fine-tuned.

Yongjoo Jeon, a principal engineer with Samsung Foundry, said Tuesday that Samsung will use an internally developed EUV mask inspection tool. This is a significant advantage for Samsung, since no commercial tool has been developed, Jeon added.

Jeon said that Samsung will initially deploy EUV without the benefit of a pellicle to protect EUV photomasks from particle contamination, another technology that is still in development. Jeon said Samsung is making progress on developing an EUV pellicle and that he is confident that the company will eventually have one to deploy in its EUV process.

Samsung is also still developing EUV photoresist and is on track to be able to achieve the target defect density for mass production later this year, Jeon said.

Samsung's process technology roadmap also includes 5nm FinFET production in 2019 and 4nm FinFET production in 2020.

— Dylan McGrath is the editor-in-chief of EE Times.
ambitiously wrote:
Samsung Plans...(恕刪)


GAA在次世代的電晶體是一種選擇

要了解先進設計
就一定要了解FinFET (鰭式電晶體)

FinFET的祖師爺是UC Berkeley的電機系 胡正明教授與Tsu-Jae King-Liu、Jeffrey Bokor研發出來的
what-is-finfet/



胡在2001年至2004年間擔任台積電首任技術執行長
而投奔三星的梁孟松
是胡的博士班學生
https://www.cw.com.tw/article/article.action?id=5063951

梁不得志
部份是由於個性
但台積電人才太多也是因素之一
我一個高中同學隨梁去三星
多少聽他在唬爛一些...

而GAA的由來
也是由FinFET演化而來的

這是FinFET的祖師爺UC Berkeley的ppt解釋FinFET
早在2002年就在解釋10nm線寬的finFET...
https://www3.nd.edu/~gsnider/EE666/666_05/QZhang_FinFET.ppt

而GAA就是FinFET double gates的一種變化...


在15:40解釋 GAA如何由FinFET變化而來...


有EE Times的人在看,不能太不專業...


健人就是腳勤
健人就是腳勤

NQQegg wrote:
GAA在次世代的電...(恕刪)

補充一下
其實以前2d就是只有一個面gate 包圍控制si channel
Finfet 是三個面
Gaa 就是四個面全包住了
Leakage 就更小

bluesky641 wrote:
補充一下其實以前2d...(恕刪)


沒錯
FFET是三個面
但總面積還是增加有限
所以才會有multi-Fins的設計
GAA也是如此~

健人就是腳勤

bluesky641 wrote:
補充一下其實以前2d...(恕刪)


Finfet 是三個面
Gaa 就是四個面全包住了
Leakage 就更小

目前tsmc有gaa的相關研發嗎?

ambitiously wrote:
Finfet 是三...(恕刪)


沒有就不用混了⋯
健人就是腳勤

NQQegg wrote:
沒有就不用混了⋯...(恕刪)


魏哲家宣示 台積5奈米明年試產
https://udn.com/news/story/7240/3182433

2018-06-06 00:53經濟日報 記者張瑞益/台北報導


台積電總裁魏哲家(左) 記者杜建重/攝影
台積電總裁魏哲家昨(5)日表示,台積電持續投資研發及產能,讓公司能在行動裝置、高效能運算、物聯網與車用半導體領域掌握商機。他並透露,台積電5奈米將在2019第1季試產、2020年開始量產。
台積電目前全球市占率超過五成,公司預估5奈米將在2019第1季試產,再度領先同業,也是台積電未來營運持續成長的動能。
魏哲家說,台積電去年持續在先進製程技術有顯著進展,10奈米以有史以來最快的速度進入量產,其量產第1年營收即占年度所有晶圓銷售的10%;領先業界的7奈米已從研發邁入製造階段,已於今年開始量產;7奈米強化版也將隨後試產。
台積電晶圓18廠於今年1月動土,將大規模使用極紫外光(EUV)微影技術生產的5奈米製程,預計於2020年開始量產。
他說,台積電去年成功推出7奈米,客戶7奈米產品設計定案超過十件,預計今年底前客戶產品設計定案將超過50件。台積電7奈米強化版也預計在今年推出。相較於7奈米技術,7奈米強化版256M SRAM晶片已經展現了相同的良率水準。
此外,台積電5奈米技術開發,目前符合2019第1季試產的目標,晶片效能與SRAM開發載具良率的提升皆符合進度,客戶測試晶片已進入生產。


經濟日報提供

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

TSMC的5nm還是用FinFET, GAA還沒用上
TSMC一向是保守,因為要開的出產能,良率要高~
太先進的技術很多技術要突破要有高良率不容易
所以還是穩健中前進...
不像三星要導入高技術節點,結果產能與良率不行,還是被台積電吃掉訂單...

甚麼都是假的
有產能有良率才是重點

NQQegg wrote:
魏哲家宣示 台積5...(恕刪)



SRAM和DRAM哪個技術等級較高?

台積蓋5奈米廠 砸7,500億
http://www.chinatimes.com/newspapers/20180622000249-260202

台積電資本支出及研發費用


2018年06月22日 04:10 工商時報
涂志豪/台北報導
台積電目前在南科興建的5奈米12吋晶圓廠Fab 18,預估總投資金額達250億美元,約折合新台幣7,500億元,將創下台灣科技業投資新高紀錄。台積電副董兼總裁魏哲家強調,半導體製程走到5奈米製程後,全世界可能只會剩下1~2家半導體廠有能力投資,但其中1家一定會是台積電。
■強調從來不跟客戶競爭
台積電昨(21)日舉行2018年台灣技術論壇,由魏哲家主持,他強調,台積電跟客戶合作,從來不跟客戶競爭,台積電擁有1,500位設計服務人員,如果下定決心要做晶片應該有機會成功,但這樣就不會是客戶的忠誠合作夥伴。魏哲家開玩笑說,「沒有(既是)夥伴又(是)客戶競爭(者的),(這樣會)背後拿刀砍你」、「你們知道我在講誰」,引起台下一陣笑聲。與會客戶則說,這就是在暗指三星,自己做手機晶片又替別人代工手機晶片。   魏哲家指出,台積電在先進製程上維持領先,包括28、16、10、7奈米等製程世代,都是最先進入市場的領先者,現在7奈米已經進入量產,5奈米正在興建新廠,在進度上持續領先競爭同業。
■研發經費與人員陣容驚人
魏哲家表示,台積電去年研發人員達6,145人,較2008年的2,069人增加近2倍,至於近3年研發經費每年都超過20億美元,去年達到26.52億美元,約新台幣780億元,創下新高紀錄。由於未來技術研發難度愈高,研發費用也會持續提高。
業者分析,台積電之所以能夠在7奈米製程幾乎囊括所有訂單,就是領先同業率先推出先進製程並進入量產。由此來看,台積電5奈米若能如期在明年底或後年初進入量產,代表2020年5奈米晶圓代工訂單也可望由台積電通吃。
魏哲家說,半導體是一個非常基本的行業,就像創辦人張忠謀所言,半導體就像是日常生活不可或缺的麵包或米飯。台積電看好智慧型手機、高效能運算、汽車電子、物聯網等4大領域,將是未來帶動半導體及台積電成長的4大動能,而且隨著5G與人工智慧等新技術來臨,世界會發生重大變動。
■明年底或後年初量產
魏哲家表示,台積電唯一目的就是幫助客戶的產品成功,即使技術不斷創新、產品不斷創新,都要跟客戶一起合作,合作是永遠不變的道理。所以,台積電從來不跟客戶競爭。魏哲家提及台積電是客戶產能與技術的供應夥伴,7奈米製程正大量生產,5奈米製程將在明年底或後年初大量生產,去年總產能達1,100萬片12吋約當晶圓,今年還會持續增加。
(工商時報)


健人就是腳勤
文章分享
評分
評分
複製連結
請輸入您要前往的頁數(1 ~ 18)

今日熱門文章 網友點擊推薦!