It's a good book, but with a heavy emphasis on design at the logic level.
(James Thornton, the author, is the engineer that helped Cray's designs become real, while at CDC. He focused more on detailed execution of the design, etc.)
Thank you for sharing this. I'm especially impressed by the gorgeous diagrams. My favorites are on pages 67, 87, 142, and 170. At first I thought they were all hand drafted until I reached the diagram on pg 170, which has dithered grey filled table cells. Does anyone know what method or software might have been used for generating diagrams like these? The publishing date is 1970 and AutoCAD didn't get released until 1982.
In the days of hand-setting you could buy books of dithering (half-tone) printed on plastic film which you would cut and press onto a page. Some brands were Letratone and Zip-A-Tone, but it was generically called shading film.
The first two machines I touched were a CDC 6400 and DEC 20 at Lehigh University in the 70s.
Both - while built by large teams - "felt" like they were designed by a single person, complete consistency across hw, OS, tools, and apps (programs). It's so foreign from the experience we have now it's difficult for me to describe.
You noticed the conceptual integrity of the systems, as described by Fred Brooks in The Mythical Man-Month.
From Chapter 4:
> "I will contend that conceptual integrity is the most important consideration in system design. It is better to have a system omit certain anomalous features and improvements, but to reflect one set of design ideas, than to have one that contains many good but independent and uncoordinated ideas."
> "Because ease of use is the purpose, this ratio of function to conceptual complexity is the ultimate test of system design. Neither function alone nor simplicity alone defines a good design. This point is widely misunderstood."
> "All my own experience convinces me, and I have tried to show, that the conceptual integrity of a system determines its ease of use. Good features and ideas that do not integrate with a system's basic concepts are best left out. If there appear many such important but incompatible ideas, one scraps the whole system and starts again on an integrated system with different basic concepts."
Its successor, the CDC 7600 seems to be in the ballpark of a 386 PC, similar CPU speed and similar memory. I wonder if the CDC 7600 (or even the 6600) could run Wolfenstein 3D or even Doom...
Those machines are vectorized, batch computers. They process data in an entirely different way and also lack the required IOs. I doubt Doom on an ASR-33 would be much fun.
The CDC 6600 could run multiple functional units (adder, multiplier, logic, etc.) at the same time, and the CDC 7600 added pipelining to the mix.
These machines were organized in such a way that the main CPU did not run (most of) the OS and was not responsible for I/O. Those operations were all handled by an array of secondary processors optimized for the purpose. A PPU (Peripheral Processing Unit) would interact with an I/O device, load data into main memory, and then notify the main CPU.
Vector came later, with the Cray 1 (and the CDC Star, less effectively).
For general-purpose integer computations, a 386 PC would have been similar or even better in speed.
For floating-point operations, a CDC 7600 would have been at least ten to twenty time faster (but less precise) than the 80387DX coprocessor. A 386 PC without coprocessor could have been one thousand times slower that CDC 7600 at FP operations.
Only with Pentium (the first Intel consumer CPU with pipelined FPU, they previously had the non-PC CPU Intel 80860 that was faster than CDC 7600) the PCs have exceeded the FP speed of the ancient supercomputers.
> During this period, CDC grew from a startup to a large company and Cray became increasingly frustrated with what he saw as ridiculous management requirements.
I learned to program FORTRAN on a CD 3100 and a 3400 in the 60's, at the Université de Montréal. Also programmed on a 6600 at Dalhousie University in Halifax.
> The only running CDC 6000 series machine has been restored by Living Computers: Museum + Labs.
Wonder what's become of it, now that the Paul Allen foundation deemed the LCM a needless expense and who knows what happened to all the donations.
Seriously, who knows what happened to all the donations? A lot of folks in tech reached in their proverbial closets for pieces of history, worked on these restorations, etc, presumably with the expectation that work would be preserved for more than a few years via this institution endowed by a vast fortune.
I always wanted to visit the LCM and it's like feeling the knife twist again every time I see a line about "this extremely cool thing from the history of computers was restored and you can see it yourself at - oh, that's the one in Seattle isn't it."
There's a running CDC6000 somewhere in the USA, or at least there was. Is it still in Seattle? Is it still in the museum building? Does it still run? Is anyone responsible for it? Has anyone turned it on and felt the hum since 2020?
The computers seem to still be in the building, and many of them are online. 'ssh cdc6500@tty.livingcomputers.org' seems to have trouble connecting to this particular machine right now.
By connecting via ssh to 'menu@tty.livingcomputers.org' you can try the other machines -- I was able to log into the TOPS-20 system running on the TOAD2, for instance.
Wow, that is really cool. Thank you - I frankly feel a lot of relief knowing the collection is still being maintained, even if I can't visit it. Maybe they need volunteers.
19 comments
[ 3.8 ms ] story [ 85.1 ms ] threadhttp://www.bitsavers.org/pdf/cdc/cyber/books/DesignOfAComput...
Glad I have a 6 hour flight upcoming.
(James Thornton, the author, is the engineer that helped Cray's designs become real, while at CDC. He focused more on detailed execution of the design, etc.)
Here's a 1974 Zip-A-Tone catalog: https://www.peculiarmanicule.com/zipatone
Here's an example of modern use of zipatone for manga: https://www.youtube.com/watch?v=Q2U4EfKCfjI
Edit: After reading the wikipedia page, I think it was the 7600. I think the summer before I worked on it, they went to the 7600. Been a long time.
Both - while built by large teams - "felt" like they were designed by a single person, complete consistency across hw, OS, tools, and apps (programs). It's so foreign from the experience we have now it's difficult for me to describe.
From Chapter 4:
> "I will contend that conceptual integrity is the most important consideration in system design. It is better to have a system omit certain anomalous features and improvements, but to reflect one set of design ideas, than to have one that contains many good but independent and uncoordinated ideas."
> "Because ease of use is the purpose, this ratio of function to conceptual complexity is the ultimate test of system design. Neither function alone nor simplicity alone defines a good design. This point is widely misunderstood."
> "All my own experience convinces me, and I have tried to show, that the conceptual integrity of a system determines its ease of use. Good features and ideas that do not integrate with a system's basic concepts are best left out. If there appear many such important but incompatible ideas, one scraps the whole system and starts again on an integrated system with different basic concepts."
These machines were organized in such a way that the main CPU did not run (most of) the OS and was not responsible for I/O. Those operations were all handled by an array of secondary processors optimized for the purpose. A PPU (Peripheral Processing Unit) would interact with an I/O device, load data into main memory, and then notify the main CPU.
Vector came later, with the Cray 1 (and the CDC Star, less effectively).
But on those two round CRT's that show up in CDC 6600 pictures it might look pretty cool, even if they're vector drawing
For floating-point operations, a CDC 7600 would have been at least ten to twenty time faster (but less precise) than the 80387DX coprocessor. A 386 PC without coprocessor could have been one thousand times slower that CDC 7600 at FP operations.
Only with Pentium (the first Intel consumer CPU with pipelined FPU, they previously had the non-PC CPU Intel 80860 that was faster than CDC 7600) the PCs have exceeded the FP speed of the ancient supercomputers.
Plus ça change
Wonder what's become of it, now that the Paul Allen foundation deemed the LCM a needless expense and who knows what happened to all the donations.
Seriously, who knows what happened to all the donations? A lot of folks in tech reached in their proverbial closets for pieces of history, worked on these restorations, etc, presumably with the expectation that work would be preserved for more than a few years via this institution endowed by a vast fortune.
I always wanted to visit the LCM and it's like feeling the knife twist again every time I see a line about "this extremely cool thing from the history of computers was restored and you can see it yourself at - oh, that's the one in Seattle isn't it."
There's a running CDC6000 somewhere in the USA, or at least there was. Is it still in Seattle? Is it still in the museum building? Does it still run? Is anyone responsible for it? Has anyone turned it on and felt the hum since 2020?
By connecting via ssh to 'menu@tty.livingcomputers.org' you can try the other machines -- I was able to log into the TOPS-20 system running on the TOAD2, for instance.