3 ms·

The first time I heard of Thinking Machines was on The Daily WTF [0]. It's quite a fascinating story of engineers building an amazing pie in the sky supercomput

The first time I heard of Thinking Machines was on The Daily WTF [0]. It's quite a fascinating story of engineers building an amazing pie in the sky supercomputer but then having the company helmed by someone who was woefully unqualified to be managing anything. The CEO's obsession with design eventually led to the Thinking Machine's most notable role as the big blinkenlights supercomputer prop in Jurassic Park, though as far as I can tell it was not in fact a UNIX system. I had to go to the Wayback machine to find the cited 'The Rise and Fall of Thinking Machines' article [1], which is also a entertaining read. [0] https://thedailywtf.com/articles/Thinking-Machines https://thedailywtf.com/articles/Thinking-Machines [1] https://web.archive.org/web/20080220113344if_/https://www.inc.com/magazine/19950915/2622_Printer_Friendly.html https://web.archive.org/web/20080220113344if_/https://www.in...

dm00 5y ago
That WTF article was wrong on a few counts --- the Connection Machine not only ran Fortran, but Thinking Machines engineers were on the Fortran 9x standards body, and I think one of Thinking Machines' Gordon Bell prizes for parallel computing was won by a Fortran program (a seismic data code from a petroleum company). I believe Thinking Machines faded away for the same reason Cray did --- its supercomputers were fine for doing physics, and a few sales a year to the national labs could sustain a company for a long time. But the end of the Cold War reduced the demand for such machines and they weren't successful in their attempts to shift to commercial applications. The early Connection Machines were array processors, not that different from GPUs today. Later machines were SPARK-based. Both systems relied on UNIX front-ends.
convolvatron 5y ago
the cm5 was really a unix system. there were a few solaris machines stuffed in to act as front ends (compile and launch jobs). since I'm here, the lights on the 5 weren't interesting - there was just a controller that ran some standard patterns. the earlier machines had an led hooked up to each of the 16 core chips and you could indeed tell what was running by the patterns - I don't recall what signal they were wired to, it most likely had something to do with the network (?).

dm00 5y ago

That WTF article was wrong on a few counts --- the Connection Machine not only ran Fortran, but Thinking Machines engineers were on the Fortran 9x standards body, and I think one of Thinking Machines' Gordon Bell prizes for parallel computing was won by a Fortran program (a seismic data code from a petroleum company). I believe Thinking Machines faded away for the same reason Cray did --- its supercomputers were fine for doing physics, and a few sales a year to the national labs could sustain a company for a long time. But the end of the Cold War reduced the demand for such machines and they weren't successful in their attempts to shift to commercial applications. The early Connection Machines were array processors, not that different from GPUs today. Later machines were SPARK-based. Both systems relied on UNIX front-ends.

convolvatron 5y ago

the cm5 was really a unix system. there were a few solaris machines stuffed in to act as front ends (compile and launch jobs). since I'm here, the lights on the 5 weren't interesting - there was just a controller that ran some standard patterns. the earlier machines had an led hooked up to each of the 16 core chips and you could indeed tell what was running by the patterns - I don't recall what signal they were wired to, it most likely had something to do with the network (?).