TY - JOUR
T1 - NS3K
T2 - A 3-nm Nanosheet FET Standard Cell Library Development and its Impact
AU - Kim, Taehak
AU - Jeong, Jaehoon
AU - Woo, Seungmin
AU - Yang, Jeonggyu
AU - Kim, Hyunwoo
AU - Nam, Ahyeon
AU - Lee, Changdong
AU - Seo, Jinmin
AU - Kim, Minji
AU - Ryu, Siwon
AU - Oh, Yoonju
AU - Song, Taigon
N1 - Publisher Copyright:
© 1993-2012 IEEE.
PY - 2023/2/1
Y1 - 2023/2/1
N2 - Nanosheet FETs (NSFETs) are attracting attention as promising devices that can replace FinFETs beyond the 5-nm node. Despite the importance of the devices, few studies analyze the impact of NSFETs at the block-level. In this article, we introduce NS3K, the first 3-nm NSFET standard cell library, and examine the results on a block-level scale. In addition to the overall process of designing a full library, we extended the scope of the buried power rail (BPR) to better layout designs. We showed that BPR, originally proposed to overcome power delivery problems, is also an effective solution for standard cell hegith reductions. Using BPR, we highlight that 4-track height standard cell designs have a negligible impact on power delivery and signal routing. Overall chip results show that the 3-nm NSFET outperforms the 5-nm FinFET by -27.4% in power, -25.8% in total wirelength, -8.5% in the number of cells, -47.6% in area, and 34.7% performance, respectively, owing to better device performance and interconnect scaling. However, careful device/layout designs and new interconnect structures must be applied to continue the scaling trend and maximize the advantages of 3-nm technology.
AB - Nanosheet FETs (NSFETs) are attracting attention as promising devices that can replace FinFETs beyond the 5-nm node. Despite the importance of the devices, few studies analyze the impact of NSFETs at the block-level. In this article, we introduce NS3K, the first 3-nm NSFET standard cell library, and examine the results on a block-level scale. In addition to the overall process of designing a full library, we extended the scope of the buried power rail (BPR) to better layout designs. We showed that BPR, originally proposed to overcome power delivery problems, is also an effective solution for standard cell hegith reductions. Using BPR, we highlight that 4-track height standard cell designs have a negligible impact on power delivery and signal routing. Overall chip results show that the 3-nm NSFET outperforms the 5-nm FinFET by -27.4% in power, -25.8% in total wirelength, -8.5% in the number of cells, -47.6% in area, and 34.7% performance, respectively, owing to better device performance and interconnect scaling. However, careful device/layout designs and new interconnect structures must be applied to continue the scaling trend and maximize the advantages of 3-nm technology.
KW - Buried power rail (BPR)
KW - FinFET
KW - library
KW - nanosheet
KW - nanosheet FET (NSFET)
KW - standard cell
UR - https://www.scopus.com/pages/publications/85146254508
U2 - 10.1109/TVLSI.2022.3229442
DO - 10.1109/TVLSI.2022.3229442
M3 - Article
AN - SCOPUS:85146254508
SN - 1063-8210
VL - 31
SP - 163
EP - 176
JO - IEEE Transactions on Very Large Scale Integration (VLSI) Systems
JF - IEEE Transactions on Very Large Scale Integration (VLSI) Systems
IS - 2
ER -