CVA CHIP GALLERY Last updated April 25, 2005

2003: 20Gb/s 0.13um CMOS Serial Link Transmitter This 700um x 1.1mm die uses an integrated LC oscillator to directly modulate the output driver, thereby achieving 20Gb/s serial output data, while burning 150mW.

Patrick Chiang, William J. Dally, Ming-Ju Edward Lee, Ramesh Senthinathan, Yangjin Oh, and Mark Horowitz.


2002: Imagine Stream Processor The prototype Imagine processor is a 21-million transistor chip, implemented in a 0.15 micron CMOS process. At 232 MHz, a peak performance of 9.3 GFLOPS is achieved while dissipating 6.4 Watts with a die size measuring 16 mm on a side.

Brucek Khailany, Scott Rixner, Ujval J. Kapasi, Peter Mattson, Jinyung Namkoong, John D. Owens, Brian Towles, Andrew Chang, and William J. Dally.


2001: Monolithic Chaotic Communications System This 1mm x 2mm 0.25um CMOS pulse position modulation(PPM) transceiver uses chaotic modulation for pulse dithering, increasing security, decreasing power spectral density, and decreased sychronization time.

Patrick Chiang, Ming-Ju Edward Lee, and William J. Dally.


1999: 90mW 4Gb/s Equalized I/O Circuit . This 0.1 sq mm CMOS serial link circuit is fabricated 2.5V 0.25um CMOS NWell process.

Ming-Ju Edward Lee, Patrick Chiang, and William J. Dally.


1998: Multi-ALU Processor (MAP). This 5-million transistor 18.25mm x 18.3mm chip is fabricated in a 3.3V 0.5um Leff, 5-level metal CMOS technology and is the core processing element of each node in the M-Machine multicomputer and includes 7 execution units, 44KB on-chip cache, and hardware support for multithreading.

Stephen Keckler, Andrew Chang, Whay Sing Lee, Nicholas Carter, and William J. Dally.


1997: Fast Links.

William J. Dally and John Poulton (UNC)


1995: Reliable Router. This 100MHz 13.5mm x 15mm network switch element is fabricated in 5.0V 0.8um HP26 CMOS process. It is the core building block for a 2-D Mesh network with a useful link bandwidth of 3.2 Gbit/sec. The Reliable Router chip incorporates adaptive routing, link-level re-transmission, and a unique token protocol to maximize reliability. It implements queueless, low-latency pleisochronous channel interfaces and simultaneous bidirectional signaling to maximize performance.

Larry Dennison, Thucydides Xanthopoulos, Kinhong Kan, Dan Hartman, Jeff Bowers, and William J. Dally.


1991 Message Driven Processor (MDP). This 1.1-million transistor 10.2mm x 15mm chip is fabricated in a 5.0V 1um CMOS process and is the core processing element in each node of the 1024-node J-Machine. The MDP is a tagged architecture and includes a 32bit ALU; integrated network interface and 3D-Mesh Router; and 4KWords of software managed on-chip memory.

William J. Dally, Stuart Fiske, John Keen, Richard Lethin, Peter Nuth, Waldemar Horwat, Michael Noakes, Scott Furman


1983: Quarternary Serial Multiplier.

William J. Dally.


1980: Stamp Processor.

William J. Dally.



achang@cva.stanford.edu