pj@litp.UUCP (Pierre JOUVELOT) (03/19/88)
Hello everyone, I'm looking for any information concerning the up-to-come Cray-3 such as clock period, technology, architecture, cooling system, and/or any interesting feature about this machine. Does anyone have any clue about it ? Thanks in advance, Pierre -- Pierre Jouvelot Centre d'Automatique et Informatique Ecole des Mines 60, bvd St-Michel 75272 PARIS France ARPA: JOUVELOT@XX.LCS.MIT.EDU BITNET: JOUVELOT@FREMP11 USENET: ...mcvax!litp!pj Tph: 64.22.48.21 (Work) 45.43.85.67 (Home)
upton@ole.UUCP (Mike Upton) (03/23/88)
The folowing data was presented at the 1987 IEEE GaAs Symposium. The paper was entitled: CRAY-3: A GaAs Implemented Supercomputer System. Dave Kiefer-CRAY John Heightley-Gigabit Logic For anyone who doesn't know, Gigabit is a Gallium Arsenide foundry with a line of SSI and MSI GaAs parts, they are doing the fab work for CRAY for the CRAY 3. This is Table 1 as persented in the paper: ATTRIBUTE CRAY3 CRAY2 GFLOPS >10 1.2 Clock 500MHz 244MHz Power 150kW 150kW Size 32" octagon 53" diameter 34" high 45" high Technology GaAs DFET Si ECL 200 gates/chip 16 gates/chip prop delay 80ps internal 350ps internal 200ps external 650ps external packageing Direct Die to Packaged devices board other numbers cited in the paper include memory: 1 Gbyte "local memory only requires 6ns access time ..." 40K chips Logic: 50K chips, ~200 different designs. the paper goes into some depth about the yield of the GaAs process. It didnt say in the paper, but I believe the 3 will contain 16 processors. CRAY is a registered trademark of Cray Research, Inc. -- Michael Upton@Seattle Silicon (uucp: ...uw-beaver!tikal!ole!upton) /* Semi-conducting our business since 1983 */
biagioni@mckinley.cs.unc.edu (Edoardo Biagioni) (03/24/88)
In article <417@ole.UUCP> upton@ole.UUCP (Mike Upton) writes: >Clock 500MHz means a 2 ns clock. >Size 32" octagon > 34" high Since light takes 1 ns to travel 12", they must have lots of fun distributing the clock.... On the other hand, maybe only the CPU is clocked at that speed, and it might be much smaller than 32". Ed Biagioni biagioni@cs.unc.edu Department of Computer Science seismo!mcnc!unc!biagioni Chapel Hill, N.C. 27514, USA