Chiplet（芯粒）技术是近年来兴起的新一代集成电路技术，因其具有提升良率、突破光罩极限、芯片架构灵活、芯片组件技术供应货架化等特点，受到产业界的广泛重视。为进一步推动chiplet技术在中国的发展，梳理了chiplet技术的应用场景，分析了chiplet中的各种核心组件技术，阐述了在chiplet技术开发中可能出现的各种技术挑战，回顾了中国chiplet标准的发展情况，最后针对中国发展chiplet技术提出了建议。
项少林
,
郭茂
,
蒲菠
,
方刘禄
,
刘淑娟
,
王少勇
,
孔宪伟
,
郑拓
,
刘军
,
赵明
,
郝沁汾
,
孙凝晖
. Chiplet技术发展现状[J]. 科技导报, 2023
, 41(19)
: 113
-131
.
DOI: 10.3981/j.issn.1000-7857.2023.19.013
Chiplet technology is a new generation of integrated circuit technology that has emerged in recent years. It has been widely recognized by the industry due to its excellent characteristics such as improving yield, breaking the limit of reticle size, flexibility in chip architecture, and standardization of IP supply chain. To further promote the development of chiplet technology in China, this article summarizes the application scenarios of chiplet technology, analyzes various core component technologies, elaborates on various technical challenges that may arise in the development of chiplet technology, reviews the development of China’s chiplet standards, and finally, puts forward some suggestions for the development of chiplet technology in China.
[1] Intel PR. Intel architecture day 2021[EB/OL]. (2021-0819) [2023-06-20]. https://www. intel. com/content/www/us/ en/newsroom/resources/press-kit-architecture-day-2021. html.
[2] Peng V. 4th Gen AMD EPYC™ processor architecture[EB/ OL]. (2022-11-11) [2023-06-20]. https://www. amd. com/ en/campaigns/epyc-9004-architecture.
[3] Xia J, Cheng C, Zhou X, et al. Kunpeng 920: The first 7nm chiplet-based 64-core arm soc for cloud services[J]. IEEE Micro, 2021, 41(5): 67-75.
[4] 张建锋 . 倚天云服务器[EB/OL]. (2021-10-19) [2023-06-20]. https://www.aliyun.com/product/ecs/yitian.
[5] 张戈 . 龙芯 3D5000[EB/OL]. (2023-04-08) [2023-06-20]. https://www.loongson.cn/product/show?id=21.
[6] Liao H. DaVinci: A scalable architecture for neural network computing[C]//Hot Chips Symposium 2019. Stanford University, CA, USA: IEEE, 2019: 1-44.
[7] 陈天石. 思元370芯片[EB/OL]. (2021-11-03) [2023-0620]. https://www.cambricon.com/index.php?m=content&c= index&a=lists&catid=360.
[8] Peng V. AMD Instinct ™ MI series accelerators[EB/OL]. (2021-11-09)[2023-06-20]. https://www.amd.com/en/graphics/instinct-server-accelerators.
[9] Huang J. NVIDIA H100 tensor core GPU[EB/OL]. (202203-22) [2023-06-20]. https://www.nvidia.com/en-us/datacenter/h100/.
[10] Hong M, Xu L J. Biren BR100 GPGPU: Accelerating datacenter scale AI computing[C]//Hot Chips Symposium 2022. Online, USA: IEEE, 2022: 1-22.
[11] Liu R, Feng C. AI compute chip from enflame[C]//Hot Chips Symposium 2021. Online, USA: IEEE, 2021.
[12] Intel PR. Introducing intel agilex M-series FPGAs[EB/ OL]. (2022-03-07) [2023-06-20]. https://www.intel.com/ content/www/us/en/newsroom/article/introducing-intelagilex-m-series-fpgas.html.
[13] Agrawal A, Kim C. Intel Tofino2-A 12.9Tbps P4-programmable ethernet switch[C]//Hot Chips Symposium 2020. Palo Alto, CA, USA: IEEE, 2020: 18-22.
[14] Cook T. Apple unveils M1 Ultra, the world's most powerful chip for a personal computer[EB/OL]. (2022-03-08) [2023-06-20]. https://www.apple.com/tn/newsroom/2022/ 03/apple-unveils-m1-ultra-the-worlds-most-powerfulchip-for-a-personal-computer/.
[15] Peng V. AMD Ryzen™ PRO desktop processors[EB/OL]. (2023-06-13) [2023-06-20]. https://www. amd. com/en/ ryzen-pro.
[16] Feldman A. CS-2: A revolution in AI infrastructure[EB/ OL]. (2021-04-06) [2023-06-20]. https://www. cerebras. net/product-system/.
[17] 胡振东 . 多物理场耦合技术的研究进展与发展趋势 [EB/OL]. [2023-06-20]. https://www. renrendoc. com/paper/205171074.html.
[18] 黄永安, 尹周平, 熊有伦. 热-电-位移耦合多体系统的多物理场分析[C]//第九届全国振动理论及应用学术会议暨中国振动工程学会成立 20 周年庆祝大会 . 杭州: 浙江大学, 2007.
[19] 武传松, 陆皓, 魏艳红. 焊接多物理场耦合数值模拟的研究进展与发展动向[C]//第十六次全国焊接学术会议. 镇江: 中国机械工程学会, 2011: 1001-1382.
[20] Miki S, Taneda H, Kobayashi N. Development of 2.3D high density organic package using low temperature bonding process with Sn-Bi solder[C]//IEEE Electronic Components & Technology Conference-2019. Baltimore, Maryland, USA: IEEE/ECTC Proceedings, 2019: 15991604.
[21] Yu D. TSMC packaging technologies for chiplets and 3D [C]//Hot Chips Symposium 2021. Online, USA: IEEE, 2021.
[22] Mahajan R, Sankman R, Patel N. Embedded multi-die interconnect bridge(EMIB): A high-density, high-bandwidth packaging interconnect[C]//IEEE Electronic Components & Technology Conference-2016. Las Vegas, Nevada, USA: IEEE/ECTC Proceedings, 2016: 557-565.
[23] Key Chung C. SPIL technology, fanout-WLP, FO-EB [EB/OL]. (2020-09-18) [2023-06-20]. https://www. spil. com.tw/TechnologyData/FanOutWLP.
[24] Peng V. AMD 3D V-Cache technology: Innovative 3D stacking technology for server and desktop applications [EB/OL]. (2021-12-14)[2023-06-20]. https://www.amd. com/zh-hans/technologies/3d-v-cache.
[25] Wang T Q, Feng F, Xiang S L, et al. Application defined on-chip networks for heterogeneous chipletchiplets: An implementation perspective[C]//2022 IEEE International Symposium on High-Performance Computer Architectur(HPCA 2022). Seoul, South Korea: IEEE, 2022: 1198-1210.
[26] Parasar M, Jerger N E, Gratz P V, et al. SWAP: Synchronized weaving of adjacent packets for network deadlock resolution[C]//The 52nd Annual IEEE/ACM International Symposium. Columbus, USA: ACM, 2019.
[27] Yin J M, Lin Z F, Kayiran O, et al. Modular routing design for chipletchiplet-based systems[C]//2018 ACM/ IEEE 45th Annual International Symposium on Computer Architecture(ISCA), CA, USA: IEEE, 2018.
[28] 虞振洋. 基于多物理场耦合特性的电气设备分析与设计[D]. 南京: 南京航空航天大学, 2015.
[29] Pu B. Design of 2.5D interposer in high bandwidth memory and through silicon via for high speed signal[J]. IEEE, 2020, doi:10.36227/techrxiv.12950261.
[30] Kim J, Chekuri V C K, Rahman N M, et al. Chiplet/interposer co-design for power delivery network optimization in heterogeneous 2.5-D ICs[J]. IEEE Transactions on Components, Packaging and Manufacturing Technology, 2021, 11(12): 2148-2157.
[31] 张鹏, 孙晓冬, 朱家和, 等 . 集成微系统多物理场耦合效应仿真关键技术综述[J]. 电子与封装, 2021, 21(10): 46-58.
