Computer History — Specific Machines
Phase 1
Phase 2
Phase 3
CDC 1604
1958
Seymour Cray
first all-transistor computer
IBM 7030 Stretch
1961
300 KFLOPs
Phase 4
CDC 6600
1964
Seymour Cray
100 ns cycle time, 1 MFLOPs
60-bit words, parallel processing, RISC-like design
CDC 7600
1969
Seymour Cray
27.5 ns cycle time, 10 MFLOPs (36 MFLOPs peak)
Cray-1
1976
Seymour Cray
12.5 ns cycle time, 136 MFLOPs (250 MFLOPs peak)
SGI R10000
1996?
SPECint95: 11.4 SPECfp95: 19.1
Cray X-MP
1983
9.5 ns cycle time, 2 CPUs, 400 MFLOPs
Phase 5
Cray-2
1985
1-2 GFLOPS
ETA Systems ETA-10G
early 1987
7 ns cycle time, 10 GFLOPs
Cray-3
1989
4-5 GFLOPS
1 GHz GaAs processors
Cray-4
1995
smaller than the human brain
One of the "Seymour Stories" related in the wake of Seymour Cray's death was told by Larry Smarr, Director of the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign:
When Seymour left Cray Research to form Cray Computer in the early 1990's, I was the fifth visitor he invited to Colorado Springs to review his plans for the Cray 3. He had a private lunch with me, one I will remember all my life. I asked him at one point what the next qualitative step for supercomputing would be. He paused and thought for a moment, then said, "I think it will be biological computing--using DNA and proteins as the computing elements just as Nature does." This was before the first tentative steps in that direction were announced by frontier researchers in this yet-to-be-born field. I asked him if he thought that he would be involved in building such machines and he said matter-of-factly, "no, I don't think I will live to see that day..."
Phase 6
ASCI project: ASCI stands for Accelerated Strategic Computing Initiative, and is a US DOE-DP (Department of Energy, Defense Programs office) program created in 1995 and (as currently planned) involving $1 billion over ten years (from 1995 to 2004). Under this program a 1.3 TFLOP machine was created by the end of 1996.
The ASCI program was created mainly to facilitate U.S. participation in the Comprehensive Nuclear Test Ban treaty, by allowing us to simulate a nuclear weapon test entirely on the computer. The program is roughly evenly divided between the three DOE-DP laboratories (Los Alamos, Lawrence Livermore, and Sandia) which have historically had a major role in nuclear weapons research as well as other big scientific computing projects. Each of the three laboratories is involved in a different supercomputer system project, and each has a different commercial contractor:
| Laboratory | Contractor | ASCI Project name | Architecture |
| Sandia | Intel | Option Red | based on Intel Paragon, with Pentium-Pro nodes |
| Los Alamos | SGI (Cray division) | Option Blue Mountain | linked Origin 2000 systems (MIPS R10000 nodes) |
| Lawrence Livermore | IBM | Option Blue Pacific | based on POWER Parallel, POWER2 nodes |
The remaining two current projects (the option Blue systems) should be complete by the end of 1998 or early 1999 and run at 3 TFLOPS each.
Another ASCI project (called PathForward) is working on developing the technology for faster machines. It focuses on developing those parts of the system which cannot be built from mass-market parts (mainly interconnect speed, debugging techniques and archival mass storage capacity). The schedule calls for 10 TFLOPS by the end of 1999, 30 TFLOPS by 2001 and 100 TFLOPS by 2004. Note that these might be accomplished by linking two or more machines together, if the interconnect bandwidth is high enough (the initial Blue Mountain system works that way, for example). Throughout the program the systems developed will have approximately the following performance characteristics (expressed as ratios): 1 TFLOPS performance / 16 TB/sec cache bandwidth / 3 TB/sec memory bandwidth / 1 TB memory / 100 GB/sec I/O bandwidth / 10 GB/sec disk bandwidth / 50 TB hard disk / 1 GB/sec archival storage bandwidth / 10 PB archival storage. Description of PathForward, with first reference to the word 'PetaFLOPS'
ASCI Option Blue Pacific (initial)
199604 (delivered)
IBM
256 processors (66 MHz POWER2)
SGI Origin 2000
199610 (introduced)
Cray Research division of Silicon Graphics
(successor to the Challenge and POWER CHALLENGE architecture)
up to 128 processors (R10000)
2 processors per node
nodes linked by hypercube network of parallel bus links
bisection bandwidth: 20 GB/sec for a 128-processor system
ASCI option Blue Mountain (initial)
199612 (delivered)
Cray Research division of Silicon Graphics
description: 7 SGI Origin 2000 systems with a total of 256 processors (R10000) and 32 GB RAM
linked by 800-MB/sec HIPPI links through four 16-way crossbars
ASCI Option Red (partial)
199612 (partial system)
Intel
(successor to the Paragon architecture)
7264 processors (200 MHz Pentium Pro)
2 processors per node
ASCI Option Red (at Sandia)
199706 (operational)
Intel
(successor to the Paragon architecture)
US $55 million, "peak" speed 1.8 TFLOPS
9072 processors (200 MHz Pentium Pro)
288 GB memory
2 processors per node
nodes linked by 38 x 32 x 2 mesh, 800 MB/sec over each link
bisection bandwidth is 60.8 GB/sec (cutting horizontally), or 51.2 GB/sec (cutting vertically)
ASCI option Blue Mountain (at Los Alamos)
199812 (projected delivery)
Cray Research division of Silicon Graphics
3072 processors (R10000)
ASCI Option Blue Pacific (at Lawrence Livermore)
199812 (projected delivery)
IBM
US $96 million, "peak" speed 3.2 TFLOPS
4096 processors (66 MHz POWER2)
Sources
Cable News Network. [[email protected]]. "Seymour Cray, father of supercomputer, dies." [http://www.cnn.com/TECH/9610/05/cray.obit/]. 5 October 1997.
Cave, Robert. [[email protected]]. "Cray Memoriam." [http://www.cray.com/PUBLIC/SEYMOUR/STORIES/0028.html]. 17 October 1996
Cray Research Corporation. [[email protected]]. "Seymour Cray: Memorial" [http://www.cray.com/PUBLIC/SEYMOUR/].
Cray Research Corporation. [[email protected]]. "Seymour Cray: Biography." [http://www.cray.com/PUBLIC/SEYMOUR/seymour_bio.html].
Fujitsu Limited. "VPP700 Series Specifications." [http://www.fujitsu.co.jp/hypertext/Products/Infoprocess/hpc/vx-e/vpp7sp-1-e.html].
Hitachi Limited. "HITACHI SR2201 Massively Parallel Processor." [http://www.hitachi.co.jp/Prod/comp/hpc/eng/sr1.html]. 5 February 1997.
Howe, Denis. [[email protected]]. "The Free On-line Dictionary of Computing." [http://wfn-shop.princeton.edu/foldoc/]. 1997.
Intel Corporation. "Overview of the DOE Accelerated Strategic Computing Initiative TFLOPS system." [http://www.ssd.intel.com/sc95booth/tflop2.html]. 1995.
Lawrence Livermore National Laboratory, IBM Corporation. [savoye1llnl.gov, rgormanvnet.ibm.com]. "Livermore and IBM in $93 Million Deal
to Build World's Fastest Supercomputer." [http://www.llnl.gov/PAO/NewsReleases/July96/NR-96-07-04.html]. 7 July 1996.
Makino, Jun. "The GRAPE-4 System." [http://butterfly.c.u-tokyo.ac.jp:8080/pub/people/makino/grape4.html]. 22 November 1996.
Mercury News Wire Services. [[email protected]]. "Supercomputer inventor dies." [http://www.sjmercury.com/news/nation/cray1005.htm]. 6 October 1996.
NEC Corporation. [[email protected]]. "SX-4 series system specifications." [http://www.nec.co.jp/english/product/computer/sx/].
Smarr, Larry. [[email protected]]. "No subject." [http://www.cray.com/PUBLIC/SEYMOUR/STORIES/0009.html]. 7 October 1996.
Summer, Frank. [[email protected]]. "A Brief Summary of Workshop on GRAPE:TNG Design, Applications, and Funding."
[http://www.astro.umd.edu/~nemo/peta/summers2.html]. 11 December 1995.
Summer, Frank. [[email protected]]. "Request for Comments: Petaflops Astrophysical Particle Simulations on GRAPE: The Next Generation." [http://www.astro.umd.edu/~nemo/peta/summers1.html]. 19 October 1995.
U.S. Department of Energy. [Chris Kielich, Matthew Donoghue, 202/586-5806]. "DOE's 'Ultra' Computer Reaches 1 Trillion Operations Per Second
Milestone." [http://www.doe.gov/html/doe/whatsnew/pressrel/pr96178.html]. 16 December 1996.
This page was written in the "embarrassingly readable" markup language RHTF, and was last updated on 2012 Nov 28.
s.27