VERIFIED FACTS — Post 49 “The Attack of the Killer Micros”
VERIFIED FACTS — Post 49 “The Attack of the Killer Micros”
Verification pass 2026-07-20. Treat this table as CANONICAL for the write phase. Verdicts: CONFIRMED | DISPUTED | CORRECTED | COULD NOT VERIFY (writer must OMIT). All Wayback URLs below are live snapshots confirmed via the Wayback availability API this session (Seymour_Cray freshly captured 2026-07-20; the rest have recent 2026 snapshots).
FRAME — killer micros & vector-industry collapse
| Fact | Verdict | Primary URL | Wayback URL | Confidence | Notes |
|---|---|---|---|---|---|
| “Attack of the Killer Micros” coined by Eugene Brooks, LLNL | CONFIRMED | https://en.wikipedia.org/wiki/Killer_micro | http://web.archive.org/web/20260107030924/https://en.wikipedia.org/wiki/Killer_micro | High | Name/affiliation solid. |
| Talk delivered at Supercomputing 1990 | CONFIRMED (with caveat) | https://en.wikipedia.org/wiki/Killer_micro | (as above) | Medium-High | Wikipedia states SC1990. Some lore cites SC’89 (Reno) and the phrase appears on Usenet comp.arch Oct 1989. Safe wording: “at Supercomputing 1990” or “around 1989–1990.” Do not assert SC’89. |
| Full name “Eugene D. Brooks III” | COULD NOT VERIFY | — | — | Low | Wikipedia gives only “Eugene Brooks.” The “III” is widely repeated in secondary sources but NOT primary-confirmed. Writer: use “Eugene Brooks” (drop the suffix) or hedge. |
| Verbatim quote “Nobody will survive the attack of the killer micros” | DISPUTED (secondary only) | — | — | Low | Attributed via Jargon File/FOLDOC, not a primary Brooks source. Do NOT present as a verbatim primary quote; paraphrase the thesis or attribute to “the catchphrase.” |
| Brooks birth date | COULD NOT VERIFY | — | — | — | OMIT. |
| Seymour Cray born 28 September 1925, Chippewa Falls WI | CONFIRMED | https://en.wikipedia.org/wiki/Seymour_Cray | http://web.archive.org/web/20260720114545/https://en.wikipedia.org/wiki/Seymour_Cray | High | |
| Cray’s fatal car accident 23 September 1996 | CONFIRMED (with caveat) | https://en.wikipedia.org/wiki/Seymour_Cray | (as above) | Medium-High | Wikipedia = 23 Sep; some wire copy = 22 Sep. Safe: “late September 1996” or cite 23 Sep per Wikipedia. |
| Accident: Jeep Cherokee merging onto I-25 near USAF Academy (Colorado Springs); another driver overtaking, cited for careless driving; rollover | CONFIRMED | https://en.wikipedia.org/wiki/Seymour_Cray | (as above) | High | |
| Seymour Cray died 5 October 1996 (age 71) | CONFIRMED | https://en.wikipedia.org/wiki/Seymour_Cray | (as above) | High | Age 71 correct (some wire copy said 70 — CORRECTED to 71). |
| Cray Computer Corp filed Chapter 11 24 March 1995 | CONFIRMED | https://en.wikipedia.org/wiki/Seymour_Cray | (as above) | High | cray-history.net says “27 March”; the filing date per Wikipedia/Seattle Times is 24 March 1995. Use 24 March 1995. |
| Cray-3 (GaAs) delivered 1993, essentially a prototype; Cray-4 (1 GHz) cancelled, never produced | CONFIRMED | https://en.wikipedia.org/wiki/Seymour_Cray | (as above) | High | Research note “Cray-3 delivered to NCAR May 1993, on loan, no paying customer” is consistent; the precise NCAR loan-not-sale detail is from secondary obits — safe to state generally. |
| Thinking Machines Corp filed Chapter 11 August 1994; hardware assets → Sun Microsystems | CONFIRMED | https://en.wikipedia.org/wiki/Thinking_Machines_Corporation | http://web.archive.org/web/20260718071732/https://en.wikipedia.org/wiki/Thinking_Machines_Corporation | High | No exact day in Wikipedia; WashPost reorganization story 16 Aug 1994. Use “August 1994.” |
| Cray Research acquired by SGI, February 1996, $740 million | CONFIRMED | https://en.wikipedia.org/wiki/Cray | http://web.archive.org/web/20260716013119/https://en.wikipedia.org/wiki/Cray | High | |
| TOP500 first published June 1993 | CONFIRMED | https://www.top500.org/lists/top500/1993/06/ | http://web.archive.org/web/20260624003053/https://www.top500.org/lists/top500/1993/06/ | High | |
| June 1993 #1 = CM-5/1024, Los Alamos, 59.70 GFLOP/s (Rmax) | CONFIRMED | https://www.top500.org/lists/top500/1993/06/ | (as above) | High | #2 achieved 30.40 GFLOP/s. |
ENGINE — MPI, MPICH, Gustafson
| Fact | Verdict | Primary URL | Wayback URL | Confidence | Notes |
|---|---|---|---|---|---|
| MPI standardisation workshop, Williamsburg VA, 29–30 April 1992 | CONFIRMED | https://en.wikipedia.org/wiki/Message_Passing_Interface | http://web.archive.org/web/20260711192054/https://en.wikipedia.org/wiki/Message_Passing_Interface | High | |
| MPI-1.0 released June 1994 | CONFIRMED (with caveat) | https://en.wikipedia.org/wiki/Message_Passing_Interface | (as above) | High | Wikipedia = June 1994. The standard DOCUMENT is dated May 1994 (brief said “May 1994”). Both defensible; safest wording “in 1994” or “the MPI-1.0 standard document, May 1994, released that June.” |
| MPI-2 released 1997 | DISPUTED (month/year phrasing) | https://en.wikipedia.org/wiki/Message_Passing_Interface | (as above) | Medium | Wikipedia phrases MPI-2 as “completed in 1996”; the widely-cited MPI-2.0 release is 1997. Safe: “MPI-2 followed in 1997” (common citation) OR hedge “in the mid-1990s.” Do not over-precise the month. |
| Pre-MPI dialects: PVM, Intel NX, TMC CMMD, Express, p4, PARMACS, Zipcode (and IBM/nCUBE vendor libs, Chameleon) | CONFIRMED (partial) | https://en.wikipedia.org/wiki/Message_Passing_Interface | (as above) | High | Wikipedia explicitly lists IBM, Intel, nCUBE, PVM, Express, p4, PARMACS. NX/CMMD/Zipcode/Chameleon corroborated across research sources. All safe to name. |
| MPI Forum ≈ 80 people / 40 organisations; conveners Dongarra & Walker | CONFIRMED | https://en.wikipedia.org/wiki/Message_Passing_Interface | (as above) | Medium-High | Draft submitted by Dongarra, Hey, Walker; Wikipedia does not name a single “convener” — “conveners Dongarra & Walker” is from the Williamsburg workshop organisation, defensible. |
| MPICH = MPI + CHameleon; Gropp & Lusk (Argonne + Mississippi State), reference implementation, dev from 1992 | CONFIRMED | https://en.wikipedia.org/wiki/MPICH | (snapshot exists; not separately listed) | High | |
| Gustafson, “Reevaluating Amdahl’s Law,” CACM vol. 31, no. 5, May 1988, pp. 532–533 | CONFIRMED | https://en.wikipedia.org/wiki/Gustafson%27s_law | http://web.archive.org/web/20260713010539/https://en.wikipedia.org/wiki/Gustafson%27s_law | High | ACM DOI 10.1145/42411.42415 (403 to fetch); johngustafson.net primary = SSL failure. Citation confirmed via Wikipedia. Written Oct 1987. |
| Gustafson at Sandia National Laboratories; experiment on 1024-processor nCUBE hypercube; ~1020× speedups on 3 apps | CONFIRMED (well-attested) | http://www.johngustafson.net/pubs/pub13/amdahl.htm (SSL fail) | — | Medium-High | Primary URL not fetchable this pass (SSL); corroborated across all three research files + SAND 88-0317 (Benner, Gustafson & Montry). Safe to state. Co-author Edwin Barsis (Sandia) confirmed. |
| Gustafson birth date | COULD NOT VERIFY | — | — | — | OMIT. |
SCAFFOLD — MPP machines & NWP ports
| Fact | Verdict | Primary URL | Wayback URL | Confidence | Notes |
|---|---|---|---|---|---|
| Cray T3D introduced 27 September 1993; DEC Alpha 21064 (EV4) 150 MHz; 32–2048 PEs; 3-D torus; needs a vector front-end | CONFIRMED | https://en.wikipedia.org/wiki/Cray_T3D | http://web.archive.org/web/20260706032107/https://en.wikipedia.org/wiki/Cray_T3D | High | |
| Cray T3E launched late November 1995; DEC Alpha 21164 (EV5); fully self-hosting MPP | CONFIRMED | https://en.wikipedia.org/wiki/Cray_T3E | http://web.archive.org/web/20260716015947/https://en.wikipedia.org/wiki/Cray_T3E | High | |
| T3E = first supercomputer >1 TFLOP/s on a real application (1480-proc T3E-1200, 1998) | CONFIRMED | https://en.wikipedia.org/wiki/Cray_T3E | (as above) | High | Distinct from IFS’s own 1 Tflop on IBM p690+ in 2004 (Salmond) — do not conflate. |
| CM-2 (1987) SIMD; CM-5 (1991) MIMD fat-tree of SPARC | CONFIRMED | https://en.wikipedia.org/wiki/Connection_Machine | (snapshot exists) | High | Detailed specs from research, secondary but reliable. |
| Intel Paragon (~1992–93), i860 XP, 2-D mesh; IBM SP1 (Feb 1993)/SP2 (1994), HPS switch | CONFIRMED (partial) | https://en.wikipedia.org/wiki/IBM_RS/6000_SP | — | Medium-High | From research (netlib/Wikipedia); safe at the level of dates+architecture, avoid over-precise per-node FLOPS unless needed. |
| ECMWF path: Cray C90-16 (1992, shared-memory vector) → Fujitsu VPP700 operational Sept 1996 (distributed-memory parallel-vector) → Fujitsu VPP5000 (2000) → IBM p690/Power4 (2002–2003) → IBM p690+ (2004) | CONFIRMED | https://www.ecmwf.int/sites/default/files/elibrary/81678-the-critical-role-of-highperformance-computing-in-medium-range-weather-forecasting-half-a-century-of-technology-innovation.pdf | — (ECMWF elibrary; not Wayback-saved this pass) | High | VPP700 = the Post 47 4D-Var machine (1997). VPP700 compute-PE count: ECMWF HPC-history doc says 39 compute PEs (~46 total); Post 47 memory says “46 PEs” — describe as “~46 PEs (39 for computation)” to reconcile. |
| Met Office Unified Model ported to Cray T3E (grid-point → halo exchange, MPI/SHMEM); UM v4.2 first T3E release; Cray-proprietary → IEEE number format | CONFIRMED (partial) | https://en.wikipedia.org/wiki/Unified_Model | — | Medium-High | From research (UM User Guide + Wikipedia); safe. |
| Spectral transform needs all-to-all “transpose” comms (MPI_Alltoallv); IFS uses transposition strategy over 4 data states (grid-point → Fourier → Legendre → spectral); reduced Gaussian grid for load balance | CONFIRMED | https://www.ecmwf.int/sites/default/files/elibrary/2004/12066-computer-architectures-and-aspects-nwp-models.pdf | — | High | Salmond ECMWF paper (fetched by NWP agent). Transpose-strategy canonical ref: Barros & Kauranne, Parallel Computing 20 (1994). |
| IFS parallel code: Barros, Dent, Isaksen, Robinson, Mozdzynski, Wollenweber, “The IFS model: a parallel production weather code,” Parallel Computing 21 (1995/96) | CONFIRMED (partial) | https://www.sciencedirect.com/science/article/abs/pii/0167819196800020 | — | Medium-High | Author list from citation; George Mozdzynski confirmed as IFS parallelisation figure. Verify exact author order if listing all names. |
PEOPLE — verification status (birth dates all COULD NOT VERIFY → OMIT)
| Person | Verdict | Notes |
|---|---|---|
| Eugene Brooks (LLNL) | CONFIRMED name/role | “III” suffix COULD NOT VERIFY — drop it. No birth date. |
| John L. Gustafson (Sandia) | CONFIRMED | Gustafson’s law 1988. No birth date. |
| Jack Dongarra | CONFIRMED | b. 18 July 1950, Chicago; Turing Award 2021; MPI/PVM/TOP500. |
| Bill Gropp, Rusty Lusk | CONFIRMED role | MPICH co-authors, Argonne. Lusk: B.A. Notre Dame 1965, Ph.D. Maryland 1970, Argonne from 1982. Gropp birth date OMIT. |
| Marc Snir | CONFIRMED role | Ph.D. Hebrew U 1979; IBM SP; IEEE Seymour Cray Award 2013. Birth date OMIT. |
| David Walker, Rolf Hempel, Tony Hey, Steve Otto | CONFIRMED role | MPI 1993 ORNL draft authors. Hey CBE NOT verified — omit honour. Birth dates OMIT. |
| George Mozdzynski (ECMWF) | CONFIRMED | IFS parallelisation / spectral-transform scalability. Birth date OMIT. |
| Deborah Salmond, Tuomo Kauranne, Saulo Barros | CONFIRMED | Cited authors. |
| Geerd-R. Hoffmann (ECMWF) | PARTIAL / hedge | ECMWF affiliation + editor of the ECMWF “Use of Parallel Processors in Meteorology” workshop proceedings (1992, 1994) CONFIRMED. Exact title “head of computing / Head of Computer Division” and tenure NOT primary-confirmed. Writer: hedge (“who led ECMWF’s computing side” / “organised ECMWF’s parallel-processing workshops”) rather than assert a precise job title. Birth date OMIT. |
CROSS-LINK SLUGS (verified against _posts/ this session)
| Post | Slug/URL | Status |
|---|---|---|
| Post 42 “The First in Bracknell” (Cyber 205) | /weather/hpc/history/2026/05/15/The-first-in-Bracknell.html | CONFIRMED |
| Post 39 “Thirty-eight Years on the Sphere” (spectral) | /weather/hpc/history/2026/05/14/Thirty-eight-years-on-the-sphere.html | CONFIRMED |
| Post 47 “Eleven Years from the Adjoint” (4D-Var/VPP700) | /weather/hpc/history/2026/05/30/Eleven-years-from-the-adjoint.html | CONFIRMED |
| Post 48 “The Hemisphere That Caught Up” (radiance/software motif) | /weather/hpc/history/2026/07/20/The-hemisphere-that-caught-up.html | CONFIRMED |
| Post 26 Cray-1 “The Machine That Looked Like Furniture” | /weather/hpc/history/2026/04/27/The-machine-that-looked-like-furniture.html | CONFIRMED — CORRECTION: research files wrongly listed “Serial Number One” as the Cray-1 post. |
| CDC 6600 “Thirty-Four People, Including the Janitor” | /weather/hpc/history/2026/05/05/Thirty-four-people-including-the-janitor.html | CONFIRMED (CDC 6600) |
| CDC 7600 “Freon and Wire” | /weather/hpc/history/2026/05/07/Freon-and-wire.html | CONFIRMED |
| “Serial Number One” | /weather/hpc/history/2026/05/04/Serial-number-one.html | = CDC 1604, NOT Cray-1. Do not use as a Cray-1 link. |
| “First Cray in Europe” (Cray-1A at ECMWF) | /weather/hpc/history/2026/05/09/First-Cray-in-Europe.html | CONFIRMED — safe extra Cray link. |
MUST-OMIT (COULD NOT VERIFY)
- Birth dates of Brooks, Gustafson, Gropp, Snir, Walker, Hempel, Otto, Mozdzynski, Hoffmann.
- The “III” in “Eugene D. Brooks III”.
- Any verbatim primary Brooks quote (“Nobody will survive…” is secondary/Jargon-File only).
- Tony Hey’s CBE/knighthood (unverified).
- Geerd-R. Hoffmann’s exact ECMWF job title/tenure — hedge, do not assert “head of computing” as a formal title.
- Supercomputing ‘89 as the killer-micros venue — use SC1990 (or “around 1989–1990”).