SystemX

SystemX Seminar: Materials and device innovations in the scaling and post-scaling eras

Topic: 
Materials and device innovations in the scaling and post-scaling eras
Abstract / Description: 

With creative innovations and significant technical effort, semiconductor technology scaling is now continuing deeper into nm dimensions. The ultimate lateral dimensions, or ultimate number of layers in 3D stacking may be under debate, but not the fact that there are fundamental or practical (technical and economic) limits to exponential improvements. The industry is already transitioning towards an era in which innovations are enabling advantages for just one or two generations. This talk presents an overview of scaling showing examples of how innovations in materials, devices and design-technology co-optimization enabled scaling and continue to do so towards the 5nm and 3 nm nodes. We also discuss some of the fundamental limits of pitch scaling as well as perspectives on beyond pitch scaling approaches, 3D stacking and heterogeneous and system level integration that will allow to continue to enhance system capabilities, and how emerging applications such as neuromorphic computing impact and drive hardware requirements and development, and open new growth opportunities.

Date and Time: 
Thursday, December 7, 2017 - 4:30pm
Venue: 
Huang 018

SystemX Special Seminar: THz Wireless Communications: New Opportunities and Challenges

Topic: 
THz Wireless Communications: New Opportunities and Challenges
Abstract / Description: 

Ultra-broadband Millimeter (mm) and Terahertz (THz) wireless communication systems are expected to help satisfy the ever-growing need for smaller devices that can offer higher speed wireless communication anywhere and anytime. The large bandwidth paired with higher speed wireless links opens up the door to a large number of novel applications such as 1) ultra-high-speed cellular links, 2) wireless short range communications for ultra-high-speed data transfer, 3) secure wireless communication for military and defense applications, 4) on-body sensors for health monitoring systems. To enable future mm- and THz-range wireless communications for these different applications, it is imperative to understand propagation mechanisms and develop good channel models.

This talk compares propagation characteristics of three frequency bands: 30 GHz (26-40 GHz), 140 GHz (110-170 GHz) and 300 GHz (300-316 GHz), discusses propagation mechanisms that are prevalent at these frequencies and proposes techniques for modeling THz wireless channels.

Date and Time: 
Tuesday, November 28, 2017 - 1:00pm
Venue: 
Packard 202

SystemX Seminar: Efficient battery usage for wireless IoT nodes

Topic: 
Efficient battery usage for wireless IoT nodes: Battery lifetime prediction
Abstract / Description: 

The prediction of the life-time of battery-powered IoT devices and wireless sensor networks is almost exclusively based on the assumption that the total charge in a battery (i.e. the mA-h) can be linearly consumed in time. This is not the case in reality. Batteries are complex electro-chemical systems and their discharge behavior depends heavily on the timing and intensity of the applied load. There is very little empirical data or reliable models available for these kinds of batteries and loads that are typically used in IoT sensor nodes for very long operational time, 5 -10 years.

We characterize the inexpensive CR2032 Li-coin cells using carefully controlled synthetic loads and a wide range of IoT-typical load parameters. We observe that actual lifetimes can differ from predicted linear ones by almost a factor of three. Furthermore, loads with similar average currents can vary significantly in the amount of capacity of the battery they can utilize. We conclude that short duration loads generally are faring better than sustained loads which was not anticipated. We suggest a better prediction model, that captures the non-linear short duration behavior, which can be implemented in constrained IoT devices.

Date and Time: 
Thursday, November 16, 2017 - 4:30pm
Venue: 
Huang 018

SystemX Seminar: Reconfigurable platforms, the thirst for bandwidth, and the future of computing

Topic: 
Reconfigurable platforms, the thirst for bandwidth, and the future of computing
Abstract / Description: 

Today, designers of specialized systems are increasingly tantalized by the enormous energy efficiency of custom silicon solutions...but are just as turned off by the spiraling costs of building and verifying those very chips. Enter the FPGA, a reconfigurable substrate that would be the absolutely perfect solution were it not for two common conventional wisdoms: they're impossible to build, and impossible to use. In this talk we will discuss the motivation and context of FPGAs, notably from the perspective of IO circuits, and what interesting problems and solutions they pose to designers and users.

Date and Time: 
Thursday, November 9, 2017 - 4:30pm
Venue: 
Huang 018

SystemX Seminar: Efficient Machine Learning Hardware Design: From High-Throughput Computer Vision to Ultra-Low-Power Biomedical Applications

Topic: 
Efficient Machine Learning Hardware Design: From High-Throughput Computer Vision to Ultra-Low-Power Biomedical Applications
Abstract / Description: 

In recent years, deep learning algorithms have been widespread across many practical applications. Algorithms trained by offline back propagation using pre-defined datasets show impressive performance, but state-of-the-art algorithms are compute-/memory-intensive, making it difficult to perform low-power real-time classification, especially on area-/power-constrained embedded hardware platforms.

In this talk, we present our recent research on how hardware designs of machine learning algorithms are efficiently customized for two divergent applications. These include deep convolutional neural networks (VGG, ResNet) for high-throughput image/video applications (e.g. autonomous driving), and compressed neural networks for ECG-based ultra-low-power biomedical applications (e.g. wearable devices). Our FPGA and ASIC prototype designs are presented that improve the energy-efficiency by optimizing computation, memory, and communication for representative neural networks.

Date and Time: 
Thursday, November 2, 2017 - 4:30pm
Venue: 
Huang 018

SystemX Alliance Fall Conference - Nov 14-16

Topic: 
SystemX Alliance Fall Conference
Abstract / Description: 

Tuesday, Nov 14 (Day 1) - starts with a SystemX overview given by Prof. Boris Murmann and Prof. Philip Wong, SystemX's faculty directors. This will be followed by overviews of our Focus Areas by their respective faculty leaders. Breakout sessions for the Bio Interfaces and Design Productivity Focus Areas will follow in the afternoon. The day concludes with an FMA student poster session / industry mixer, reception from 4:30 -6:00 pm.

Wednesday, Nov 15 (Day 2) - begins with topical presentations on research into Machine Learning at Stanford. The intention is to be well rounded, and present the challenges as well as the opportunities to be found in ML. The afternoon will hold three breakout sessions for the Computation for Data Analytics, for the Heterogeneous Integration, and for the Photonic & Quantum Technologies Focus Areas, after which member company representatives will assemble for our Business Meeting. The day concludes with the SystemX Dinner also at the Li Ka Shing Center.

Our featured speaker at dinner will be Dr. Wally Rhines, President & CEO of Mentor Graphics, a Siemens business

Thursday, Nov 16 (Day 3) - will hold two breakout sessions in the morning for the Energy/Power Management Systems, as well as, the Internet of Everything Focus Areas. This final day will end at noon.

On Tues and Wed, the conference will be held at the Li Ka Shing Center. The Thursday morning sessions will be held in the Allen buildings. Please consult the preliminary agenda for more details.

 

 

PRELIMINARY SYSTEMX NOVEMBER CONFERENCE AGENDA

PARKING INSTRUCTIONS FOR GALVEZ LOT

PARKING AND CONFERENCE MAP

Date and Time: 
Tuesday, November 14, 2017 (All day) to Thursday, November 16, 2017 (All day)
Venue: 
Li Ka Shing Conference Center, Stanford University

The Vision and AI Startup Landscape: Dispatches from the Entrepreneurial Front Lines [SystemX Seminar]

Topic: 
The Vision and AI Startup Landscape: Dispatches from the Entrepreneurial Front Lines
Abstract / Description: 

No recent technology trend has attracted as much hype as deep neural networks. Thousands of startups have attached themselves to the AI wave, chasing opportunities for new cloud applications for business and social media, new system platforms like self-driving cars, and new silicon platforms. In this talk, we deconstruct the major technologies and end-system functions and examine what startups and applications are most likely to succeed, where entrepreneurship is thriving, and how deep learning innovations in real-time voice and vision, in convolutional, recurrent and generative network structures, and in new high-efficiency neural processor architectures are likely to shape future systems.

Date and Time: 
Thursday, September 28, 2017 - 4:30pm
Venue: 
Huang 018

Maximizing Server Efficiency: from microarchitecture to machine-learning accelerators [SystemX Seminar]

Topic: 
Maximizing Server Efficiency: from microarchitecture to machine-learning accelerators
Abstract / Description: 

Datacenter servers have emerged as the predominant computing platform, impacting nearly every aspect of our daily lives. The need to study, understand, and improve server systems has been the main driver of my research career. I will begin this talk with a brief overview of our work on the CloudSuite benchmarks and Dark Silicon in Servers, which identified a number of challenges and opportunities for server architecture. Then, I will describe how these results have led to our work on Scale-Out Processors and Temporal Memory Streaming, and dive deeper into my students' recent results on instruction sequence memorization and FPGA-based accelerators for machine learning.

Date and Time: 
Tuesday, October 3, 2017 - 4:30pm
Venue: 
Huang 018

Satisfiability Modulo Theories [SystemX Seminar]

Topic: 
Satisfiability Modulo Theories
Abstract / Description: 

I will give an introduction to Satisfiability Modulo Theories (SMT) solvers, including the DPLL(T) architecture that combines a Boolean satisfiability (SAT) solver with a theory solver, techniques for building individual theory solvers, and techniques for theory combination. I will also cover some applications of SMT solvers, including symbolic execution and equivalence checking.

Date and Time: 
Thursday, October 5, 2017 - 4:30pm
Venue: 
Huang 018

Pages

Subscribe to RSS - SystemX