guru@ut-emx.uucp (chen liehgong) (03/07/91)
I am not sure whether this topic was ever discussed in this newsgroup. Well, anyway, here it is: Why were the CDC 6600 and the TI ASC (Advanced Scientific Computer) failures? I would like to have your opinions on this. -r.guru prasadh (guru@emx.utexas.edu)
lhughes@b11.ingr.com (Lawrence Hughes) (03/07/91)
In article <45252@ut-emx.uucp>, guru@ut-emx.uucp (chen liehgong) writes: > > Why were the CDC 6600 and the TI ASC (Advanced Scientific Computer) > failures? I would like to have your opinions on this. > > -r.guru prasadh > (guru@emx.utexas.edu) I'm not sure how you define a "failure" if you include the CDC 6600 one! It was THE machine of choice in large scientific sites, such as NCAR, LLL, FermiLabs, etc., as well as in many larger universities (those who didn't have IBM cr*p foisted on them by a legislature/BOR). It was also VERY popular at "spook sites" ("unpublicised" government / military sites), as it was then ONLY machine which could provide air-tight hardware level security between jobs or for the OS (the OS was actually resident in a group of 10 or so "perhiperal processors"). In terms of performance, it was the fastest, most powerful machine available when it was released. In terms of impact on the industry, it was the foundation for much of the "hot iron" that has followed, including the 7600, Cyber 73/74/76, Cray I/II, and distinct echoes of it can still be heard in the 860, 88000, etc. It pioneered many concepts of parallelism, register architecture and instruction (pre) fetching that are considered basic today (such as the "stunt box" scheme). All in all, I would consider this to be one of the MOST important machines in the history of the field. And I'm pretty sure CDC didn't exactly lose money on it! The TI ASC is another matter - fell flat on its face right out of the starting gate - a few nice ideas, but poor implementation, no visible support and no sales to speak of. Basically, TI was way out of their league, in virtually every respect. ------------------------------------------------------------------------------ Lawrence E. Hughes | ASK DR. MATH - He knows MORE than you do! Remember, Intergraph Corp. CR1100 | He's not a REAL Doctor: "I have a Bachelor's Degree, Huntsville, AL 35984 | in MATH!" * uunet!ingr!b11!lhughes * (205) 730-4167
wilson@schaefer.math.wisc.edu (Bob Wilson) (03/08/91)
In the early 70's I read that "90% of all computations done since the origin of mankind" had been done on CDC 6600's. I don't know if it was literally true (hard to verify!) but the mere fact that it was believable makes it hard to call the 6600 a "failure"! One might also see the special section on Seymour Cray at the Computer Museum in Boston. The "original" CDC machine, the 1604, and the 6600, both attributable in large part to Cray and not to some design committee, have always seemed to me the cleanest as well as the most innovative machines from CDC. Bob Wilson wilson@math.wisc.edu <usual disclaimer>
kludge@grissom.larc.nasa.gov ( Scott Dorsey) (03/08/91)
In article <45252@ut-emx.uucp>, guru@ut-emx.uucp (chen liehgong) writes: > > Why were the CDC 6600 and the TI ASC (Advanced Scientific Computer) > failures? I would like to have your opinions on this. > > -r.guru prasadh > (guru@emx.utexas.edu) The CDC 6600 was not a failure at all! It was an extremely popular machine that a lot of sites purchased. The CDC 6000 architecture certainly isn't a failure in general; there are still a lot of sites using Cyber 170 series machines or 180 machines in 170 mode which are essentially based on the 6000 architecture. We just surplussed a 170, and have three more left. Want one? The TI ASC was a failure because TI had no experience in building high speed machines. Not enough money was one problem. Generally bad system design was another. I recall NASA having looked into purchasing the machine but discovering that they couldn't get a single program to pass the Fortran compiler.... --scott
henry@zoo.toronto.edu (Henry Spencer) (03/08/91)
In article <45252@ut-emx.uucp> guru@ut-emx.uucp (chen liehgong) writes: >Why were the CDC 6600 and the TI ASC (Advanced Scientific Computer) >failures? I would like to have your opinions on this. Assuming that you have confused the 6600 (successful) with the Star 100 (failure), the most basic reason why the Star and the ASC failed was that they focussed on long-vector performance and ignored scalar performance, on the theory that everything would vectorize extremely well. The Cray 1 succeeded where they failed because it was a blazing-fast *scalar* machine first and an even-faster vector machine second. -- "But this *is* the simplified version | Henry Spencer @ U of Toronto Zoology for the general public." -S. Harris | henry@zoo.toronto.edu utzoo!henry
brooks@physics.llnl.gov (Eugene D. Brooks III) (03/11/91)
In article <1991Mar7.215545.430@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes: >Assuming that you have confused the 6600 (successful) with the Star 100 >(failure), the most basic reason why the Star and the ASC failed was that >they focussed on long-vector performance and ignored scalar performance, >on the theory that everything would vectorize extremely well. The Cray 1 >succeeded where they failed because it was a blazing-fast *scalar* machine >first and an even-faster vector machine second. This is the correct viewpoint. One should also take note of the fact that the fastest scalar processors these days are Killer Micros, and they do it an incredibly small fraction of Cray's price for a CPU. No one will survive the Attack of the Killer Micros!
guru@ut-emx.uucp (chen liehgong) (03/12/91)
There was an obvious error when I asked why the CDC 6600 was a failure. It is indeed a popular m/c as many of you pointed out. The TI ASC, of course, was not a success!! My apologies to all those (designers, fans etc.,) connected with CDC 6600. -r.guru prasadh PS: Thank you for your opinions regarding the TI ASC, though.