Monday, June 9, 2014
High Performance Mobile SoCs Enabled by 10nm SoC Technology
Organizers/Chairs: Geoffrey Yeap, Qualcomm and Ken Uchida, Keio University
Organizers/Chairs: Geoffrey Yeap, Qualcomm and Ken Uchida, Keio University
The 10nm generation will push optical lithography and the FinFET architecture to the very limits, challenging many scaling assumptions. Multiple tradeoffs will need to be made to ensure proper SoC power, performance, and cost scaling, while allowing for fast yield maturity and time to market. This shortcourse will cover scaling trends in 10nm-class SoC technologies, show the process challenges at these tight pitches, and discuss the innovations needed to ensure a healthy and scaled FEOL integration.
Bio: Andy Wei received his B.S. in E.E. from the University of Arizona, Tucson, AZ, and his M.S. and Ph.D. in E.E. from the Massachusetts Institue of Technology, Cambridge, MA, USA, in 2000. He worked in Process Integration and Technology at AMD through 2009, taking leading-edge devices from conception to manufacturing, spanning the traditional scaling, stressor-based scaling, and high-k eras. Since 2009 he has been with GLOBALFOUNDRIES Technology Development, with focus on process scaling, architecture, and pathfinding for leading edge technologies in the FinFET era and beyond.
As the dimensions shrink from technology node to node, the interconnect resistance and capacitance need to scale as well to minimize product performance impact. To enable this, there has been innovation in the areas of materials, processes and integration such as low-k dielectrics, conductor materials and multiple patterning. With this, there has been a need to balance the performance, reliability and manufacturability requirements. This short course is intended to provide insight to the innovation behind 10nm backend of line (BEOL) interconnect scaling.
Bio: Terry Spooner received a Bachelor of Science in Chemistry from Rochester Institute of Technology in 1985, and was awarded a Ph.D. in Materials Engineering from Rensselaer Polytechnic Institute in 1990 for his dissertation on the measurement of excess noise in metal films. From 1990 to 1998, he was employed by Digital Equipment Corporation serving as a reliability engineer, process development engineer and technical manager at the Hudson, MA site. Terry joined International Business Machines in 1998 and has worked in the areas of unit process development, BEOL integration and technical management in East Fishkill, Yorktown Heights and most recently, Albany, NY. He has contributed to the early definition and development of several technology nodes incorporating copper and low k dielectrics for advanced bulk and high performance SOI products.
FinFET design requires new design skills to deal with the transistor 3D structure, the quantization of device width and double patterning. This short course will start out with the advantages of FinFET transistor:
BIO: Charles Young joined TSMC in 2012, as Director in charge of Design and Technology Platform Leadership. He has been working on TSMC’s leading nanometer process technology, 20SoC, 16FF and beyond since. His responsibilities include process technology and design co-optimization; ensuring TSMC’s technology area-power-performance leadership. Prior to TSMC, he was with NVIDIA for more than 10 years, as Director of Hardware Engineering. He was in charges of NVIDIA Standard Cell, SRAM Compiler and custom digit circuit designs. He also established and managed NVIDIA Shanghai Research and Development Center from 2005 to 2010, serving as General Manager. He holds a BS degree in EECS from University of California, at Berkeley.
Embedded memory has evolved during decades and plays a important role in a mobile System-On-Chip. Scaling, speed and power-consumption have been improved with the help of technology and design-architecture. eSRAM is still the main memory for CPU and GPU with most integration techniquie occupying about 30% over SoC. Then eDRAM is making the challenge to replace eSRAM for higher bandwidth and capacity. eNVM provides the key functionality of chip ID and redundancy resource in SoC. The STT-MRAM is getting attention as the new memory. This short-course covers all kinds of eMemory for applications and features in a commercial point of view. Especially, eSRAM with assist circuits and eDRAM for low-power design are in introduced.
Bio: Gyuhong Kim received the B.S degree in electrical engineering from Yunsei University, Seoul, Korea in 1990. He joined the Semiconductor Research and Development Center, Samsung Electronics Company Ltd., Yongin, Korea, in 1990. From 1990 to 1997, he was enageged in the development Fastpage/EDO DRAM and SDR/DDR SDRAM. Since 1997, he has been working on the development of embedded memory such as embedded DRAM, embedded NVM and embedded SRAM at the System LSI Division of Samsung. Now he is a principle engineer and director of Library IP(eMemory, Standard Cell & GPIO) development team.
This presentation will focus on the source and effects of variability in the upcoming technology nodes and on the DFM techniques aimed at combating the inherent variability to result in robust high yielding products. We will cover all key physical mechanisms for random and systematic variations in logic, embedded memory and analog/RF components of the mobile platform SOC’s. We will also demonstrate techniques for characterizing these variation sources and show results from the state-of-the-art fabrication processes. Then we will present effective DFM techniques to achieve high manufacturing yield and reliability targets without sacrificing power, performance or SOC chip area. These techniques have been implemented in the design flows of the most advanced circuits and will be illustrated by the real design examples.
Bio: Andrzej J. Strojwas is Joseph F. and Nancy Keithley Professor of Electrical and Computer Engineering at Carnegie Mellon University. Since 1997 he has served as Chief Technologist at PDF Solutions, Inc. He has held positions at Harris Semiconductor Co., AT&T Bell Laboratories, Texas Instruments, NEC, HITACHI, SEMATECH, KLA-Tencor and PDF Solutions, Inc. He received multiple awards for the best papers published in the IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, IEEE Transactions on Semiconductor Manufacturing and IEEE-ACM Design Automation Conference. He is also a recipient of the SRC Inventor Recognition Award. He was the Editor of the IEEE Transactions on CAD of ICAS from 1987 to 1989. He served as Technical Program Chairman of the 1988 ICCAD and Conference Chairman of the 1989 ICCAD. In 1990 he was elected IEEE Fellow.
A common feature of applications such as mobile internet, high density and 3D displays, cloud computing, gaming and high performance computing is the ever increasing need for more data and more data transfer. 3D integration technologies allow for a significant power reduction together with a strong increase in memory-logic interconnect bandwidth. This is achieved by reducing both the interconnect connection pitch and wire-length. This can be done at different levels of the electronic system hierarchy and by repartitioning the electronic system. Because of thermal limitations, some applications require the use of a so-called interposer substrate to realize high bandwidth interconnects between subsystems. This approach will also allow to take the next step in high bandwidth communication. In this presentation will review the state-of-the-art and progress in the development of 3D technology components such as TSV, micro-bumps and advanced packaging and review the benefits and challenges for 3D system design. We will report on aspects such as thermal modeling, mitigation of mechanical stress of TSV and packaging, design metrics (Keep-out-Zone), compatibility with advanced CMOS, testability and signal/power integrity to enable system designers to develop optimized 3D systems.
Bio: Geert Van der Plas obtained the M.Sc. and Ph.D. degrees from the Katholieke Universiteit Leuven, Belgium, in 1992 and 2001, respectively. He joined imec, Belgium, in 2003, where he is a principal scientist. He has been working on energy efficient data converter, power and signal integrity and design enablement of scaled CMOS and 3D integration technologies. He has authored and co-authored over 150 papers in journals and conference proceedings and serves on the technical program committee of the symposium on VLSI circuits.
2014 Silicon Nanoelectronics Workshop will be held on June 8-9, 2014 as a satellite workshop at the same location where Symposium takes place. Silicon Nanoelectronics Workshop Website
2014 Spintronics Workshop on LSI will be held on June 13, 2014 and will focus on VLSI-implementable Spintronics Technology Spintronics Workshop Website