What would that price be equivalent to today?
(quick search later) According to http://www.westegg.com/inflation/infl.cgi it would be $38 million to $64 million in 2006 dollars.
You smell that? That's the tangy aroma of aging silicon, cigarettes and loneliness. This week we're traveling back to the 1960s to look at one of the crown jewels of the high performance computing world. You'll notice that — Hey! Don't touch that. Has Jean Claude Van Damme's Time Cop taught you nothing about tooling with the …
What would that price be equivalent to today?
(quick search later) According to http://www.westegg.com/inflation/infl.cgi it would be $38 million to $64 million in 2006 dollars.
Interesting article, although Columbia University in New Yawrk claims to have had the fastest computer in 1969 with the IBM 360/91:
However, I'll let the experts duke it out, I'm just glad to see old iron that for once is older than me, if just by a few years.
Hey, I want to congratulate you on writing this article.
1) There is one (Well, OK. A later, but binary-compatible one) still in use at cray-cyber.org. They give you user accounts. They are also excellent folks. Feel free to dump money at them so that they can do cool things for us.
2) IIRC, there was actually only one peripheral processor, but it was a barrel processor, that is, it executed the responsibilites of each logical CPU. This changed, at some point. And they weren't speedy at all.
3) It might be worth noting that the displays are actually vector displays. That is, every letter is drawn as a vector, and the CRT ray is controlled as a vector, not as a drawn raster. It's also pretty damn fast, and can monitor RAM as code is executing in it. Which is a lot of fun to watch.
4) The CDC 6600 was FAST. Holy shit was it ever fast. It was an order of magnitude (in some instructions, two OOMs) faster than anything IBM had at the time. Very very fast. Whoosh.
5) You mention "aging silicon" - actually, I think the CDC6600 was based on germanium, not silicon, transistors. Don't quote me on that, though.
I love this column. Old computers with El Reg's sense of humour.
Much better than that Second Life weirdo with the mesh outfit.
My dad was a CDC salesman back in the 1960's. He sold a bunch of 6600's, almost all bringing the computer into a new industry or science area that had never been automated before. As a child, I used to go with him on business trips to customer sites, and often had a chance to "play" on the 6600 with one of his analysts. (Oh, the usual: drawing pictures on the CRT, making the line printer play Christmas songs, punching out my name in the holes of punch cards...) This experience is probably what got me to follow in my father's footsteps and pursue a career in IT.
Oh, and don't forget about the other CDC boxes that made impacts in our day-to-day lives: the 3600 and the 1700...
Yeah, it does sound better shouted, doesn't it?
A real shame computer makers these days put about 3 seconds into design - Pay attention up the back there, Mr. Jobs, this is for your benefit too!
It's great to see someone out there looks after all this marvelous old gear.
Wikipedia has a good write up (http://en.wikipedia.org/wiki/CDC_6600), but don't miss out on this classic written by Thornton himself:
I know a couple CDC staff who remember being around for the launch of the 6600. Anymore floating on El Reg?
I was not around for the launch of the 6600 but I did work on the first one to be delivered to a customer. Something the article did not mention is that, in addition to plunking down millions to buy the computer, the customer had to pay heaps more for a team of on-site customer engineers to keep it running. It was a great machine to program; I wrote lots of assembly language for it. Those were the days!
If the 6600 is like the 6400, the machine has hardware opcodes for floating point arithmetic. The 60 bit words stored floats. It was a scientific number cruncher.
"3) It might be worth noting that the displays are actually vector displays. That is, every letter is drawn as a vector, and the CRT ray is controlled as a vector, not as a drawn raster. It's also pretty damn fast, and can monitor RAM as code is executing in it. Which is a lot of fun to watch."
I wonder if there is a video or any simulation to show what the operator at the console would have seen?
And surely there were video games running on it? Those two scopes just shout out for a good space duel game :-) Anyone remember the Vectrex, an 1980's console game that also had a vector display.
Did my internship/degree work experience with Digital Equipment Corp. and one of my projects was to interface a PDP 11/70 as a front-end processor used to load jobs and retrieve output. Went on and did the same for a 7600 as well.
Never forget the wiring, point to point shielded twisted pair and coax!!
I used to use one of those back in the 70s via Teletype/tape/card from a remote site. Never knew what the beast looked like though - it could have been vast ranks of slide-rule wielding pixies for all I knew. So 30+ years later, I know. Thank you, Reg
Were the on-site engineers the same group that had to program the machines with all the punchcards? I hear -lots- of stories of a certain person dragging down a cart with a whole lot of punchcard boxes at a time for the programmers to run.
<- Not old enough to have been there, but likes the stories of washing-machine sized memory
...is that even the wikifiddlers have more correct information than presented here. Things like 'where's the any key?' designed to lighten the load on the reader only works for people who have 1) a shallow sense of humor and 2) never seen Type 5 keyboards. If that cliche'd crap is meant to be funny, you're failing, because the issue with shallow coverage in computing has been covered more times than the maximum core storage capacity of the machine in question.
El Reg staff, please don't make this another Reg Dev. And categorize articles correctly -- AFAIK there are no chips in this machine and it's certainly not a PC. Just the /idea/ of wikifiddlers being more thorough than El Reg is repulsive, but something tells me we shouldn't shun making that comparison. Here you are.
Being an old computer hand I love this column but in a way it makes me sad too.
Where has the fun gone? These days all computers are the same. A PC is a PC is PC maybe with slight difference in keyboard layout. The "Intel" instruction set is so dominant there is a generation that has never programmed on anything else - that is assuming they have even got close to the instruction set and not just programmed in C or one of it's close relatives (C++, Java, C# all the same to a first approximation). There are really only 2 operating systems going anywhere (Windows and Unix/Linux derivatives). Yes I know big blue still has OS/z limping along and Unisys probably has an antique OS or 2 around too but where are the exciting new developments with really DIFFERENT feeling OSs from mainstream vendors?
to one of your points - into Servers it goes. i'll leave austin to handle the rest, although i suggest some measure of relaxation is in order
``into Servers it goes.''
``although i suggest some measure of relaxation is in order''
Nah, I'm having way too much fun 8)
Always grand to hear. Although, unlike other publications, we expect our readers to do the real dirty work, elevating our paltry contributions well above the wikifidler stage. If the truth is out there, you guys have it.
He was even less of a commodity than the CDC 6600 design was a commodity design.
the 6600 was a johnny-come-lately; I think Adelaide uni had one. I worked at the Commonewealth Bureau of Census and Statistics, what had a brace of 3200s, some 3300s, a 3500 and a 3600.
The 3200s ran SCOPE and essentially program at a time; there was also a background, interrupt-driven program to mange off-line printing.
The 3300s came (at our establishment) in two flavours, the A running MSOS, a bit more capable than SCOPE and having disk, and the B, bigger and running a full multitasking OS, MASTER. It did clever things with special registers to extend memory addressing.
The 3600 ran SCOPE which wans't mich like the 3200-scope or theat which appeared later in the 60-bit boxes.
The low 3000 series (up to 3500) had 32-bit words, the high 3000s (except I think there as a 48-bit 3400) had, um, 48-bit words.
the little 3200 in Perth was the first one I used, I later programmed the 3500 (in compass) and the 3600 in fortran.
Those who remember IBM's S/360 and S/370 might like to google for hercules. it emulates IBM's mainframe, from then to now, is FOSS, runs on Linux and, er, Winders.
Having got that, one can then choose from OS (I think it's around), MVS, VM and (I think) DOS variants. MVS runs somewhat faster than we ran it in the 70s.
And Linux, of course, but that's too modern for these reminisces.
I wasn't even born when this machine was designed, or even when it was scrapped as obsolete, but I'm still impressed with the terminology, and the look of the thing; why can't we have an operating system today called KRONOS, in capital letters? I imagine a full set of instruction and service manuals would fetch a pretty penny on eBay. I picture this machine calmly working out a plan to order human society, like Colossus from "The Forbin Project". Control Data Corporation has a awesome, and slightly sinister, ring of power and command. It saddens me that modern computing terminology is so cute and infantile.
"cute and infantile"
E.g. perhaps Microsoft could rebrand themselves as The Microcomputer Software Corporation. Their new operating system will not be called Vista any more, it will be called KALIBAR VI. The dialogue boxes will not have "yes, no, cancel", they will have "proceed, decline, ABORT"; if the user picks ABORT he will be compelled to fill out a lengthy form explaining why he chose this action.
When the machine is shut down it will not say "it is safe to turn off your computer", it will say "Rotate system key A to SAFE and unlock the system switch. Do not exceed stated voltage. Failure to comply will result in major system failure, possible elevation of DEFCON level, and is a notifiable offence".
The retail box will include an orange jumpsuit and a badge that gives the user Code Bikini Black Special security clearance. And a pistol.
This, or a 512mb C2D running vista...
Another thing to remember about the 6600 is that it was the machine that ran PLATO, the early 70's system where just about everything you love about the internet was invented -- nationwide email, newsgroups, chat and IM, multiplayer graphic games and much more.
You can actually log on to a PLATO system (running on a CDC machine that is emulated on a workstation, and runs faster than the original!) and check it out for yourself, using an emulation of the original PLATO 512x512 dot plasma display (yeah, plasma, back in the early 70's!), at http://www.cyber1.org/
The college had received it just the previous summer, and with the purchase of the machine, got one year's loan of a programming instructor. If you had the choice of a lab with a couple of DEC PDP-11s that you had to sign up for time in the middle of the night, or learning on a machine with hardware floating point and remote job entry, which would you chose? Big iron, all the way! KRONOS with a "Tempo" front end minicomputer to preprocess the data streams from up to 100 Teletypes connecting over phone lines.
Thanks for the memories...
Err, email started on an IBM 7094, and much of the Internet technology was developed on a mish-mash of IBM, DEC and Honeywell kit. CDC kit was out of the price range of most of the universities and companies that worked on the DARPA projects. Minicomputers like DEC's PDP 11 were much more appropriate hosts for a network of machines than a CDC 6600 anyway.
I followed the link to Thornton's book on the CDC Design from 1970, it seems things really don't change; for instance to quote paragraph 2:
"Early in digital computer history circuit technology advanced so rapidly that giant strides were made in equipment performance with little variation in design structure."
Mmm... X86/PC architecture anybody?
Ah, back when computers were computers and men were men;) I loved the old machines with blinking lights and displays!
I was a university student from 1970 through 1978, and followed up as a research assistant. I used a 6400 to run Fortran and Iftran using a card punch. I rewrote an analysis program originally written for an IBM 360 and the first thing I did was dump all the double precision references since the 6400 with its 60 bit word was equivalent to that to begin with. Once it got the program running correctly it was OPT=2; can you do that anymore? Then on to another location with the 170 where I had full resources open to me since the researcher I worked for had a project that justified the purchase of the machine. The card database filled a good sized room, but I worked from a tape copy. Well those card punch tab cards wore out fast on the clerks entering the data, and sometimes they got punched wrong. I had to shift columns of data to get the decimal point where it belonged; probably in excess of 250,000 data points were corrected long after they thought they had a clean database. It sure was fun.
The CDC 3000 machines were actually either 24-bit (not 32-bit) or 48-bit machines, and the instruction set was similar for both sizes, but not identical. As for the rest, I see that the Wikipedia article on that machine gives it silicon transistors, and notes that the peripheral processors were what we would term one peripheral processor with ten threads today, as noted above.
OK, I've double-checked. It used planar silicon transistors, and the logic family was DCTL, direct-coupled transistor logic.
There is a book _Design of a Computer, the CDC 6600_ by J.E. Thornton, 1970. Good luck finding it, as Amazon calls it "unavailable"; I found one once at a flea market in Pennsylvania but gave it away.
Never did touch a CDC 6600, but the first big machine I got to play with was a 6400. Those ASR33 Teletypes were such a Great Thing! Oh Boy! Interactive Computing with a rhythm sectin thrown in! Way too cool.
Later the University of Calgary acquired a Cyber 172 which (IIRC) was more or less an integrated-circuit update of the 6000 series hardware.
Regrettably I have no pictures of the 6400, but you can find a (rather poor) picture of the Cyber 172 console at:
I've seen the same display on a 6400 screen. The console display was driven by a dedicate PPU (Peripheral Processor Unit), basically a mini-computer that front-ended the CPU. At the time the console was the most advanced graphics device available on the system. I believe our configuration had 12 PPU's that mostly managed I/O peripherals.
One amusing oddity of the CDC machines was that one did not "boot" them: one performed a "deadstart":
which loaded a program into PPU-0 (again, IIRC) from a panel of toggle switches.
They were fascinating machines, but I have to say I was glad to see the back end of the last one they wheeled out. The next box ran Multics....
<Good luck finding it, as Amazon calls it "unavailable">
Fortunately, Tommy Thorn provided a working link to a full PDF copy of it, in the 7th comment above.
Where's the "Doh" icon? The airhead will have to do.
>Were the on-site engineers the same group that had to program the machines
>with all the punchcards? I hear -lots- of stories of a certain person dragging down
>a cart with a whole lot of punchcard boxes at a time for the programmers to run.
No, the CEs were not programmers. I actually have little notion of what they did -- but they did plenty of it. There were times when I had programs on huge decks of punch cards but I did most of my 6600 programming on Intercom, the interactive subsystem. I worked with SCOPE, KRONOS and NOS. Before my time there was another os called Chip (for Chippewa). I also remember PLATO.
I was in grad school in the early 80's, and we had a CDC6600 that was still in use. CDC had sold the school a STAR64 that never worked. The 66 soldiered on, past retirement, until it was replaced by a CYBER170.
The particular school, not to be named, had Seymour as an unpublicized alum. He was unpublicized because he had been denied entry to the grad school because his math scores were not good enough.
I was quite familiar with the 170, having used one as an undergrad at a different school. Barry D. had used the CDC APL environment as a vehicle for an undergrad linear algebra course. I do not think I learned much linear algebra (orthonormal basis of a WHAT ?), but I did learn quite a bit about APL and programming. Dick R. had developed a liking for APL when he was using the ILLIAC 4.
In the mid-70's I spent part of a summer in Fort Collins at Colorado State University at an NSF-sponsored science program. One of the classes was COMPASS (assembly language on the 6600); we punched up our homework and ran on the university's dual-6600.
I fondly remember the weird architecture - to read from memory, you wrote the 18-bit (IIRC) address to an "A" register (A0-A5 only), and the data showed up in the corresponding "X" register (X0-X5). If you wrote an address to A6 or A7, the data in X6 or X7 was written to memory. The COMPASS textbook we used was devoted to the A6 and A7 registers, "... without whom this book could not have been written".
There were 15, 30, and 60-bit opcodes, so part of the task of optimizing assembly code was to order the instructions so that your code packed optimally into the 60-bit words.
Macro Rodent asked:
> I wonder if there is a video or any simulation to show what the operator
> at the console would have seen?
> And surely there were video games running on it? Those two scopes just
> shout out for a good space duel game :-) Anyone remember the Vectrex,
> an 1980's console game that also had a vector display.
Yes. Run your own 6600. To get started, go to
Tom Hunter's simulator is faster than the real iron was.
I worked on CDC machines deom about 1969 at the Hebrew University in Jerusalem as a System Programmer (an exalted position, enabling one to get System Time during the night - thats when I learned to drive while sleeping every other second).
Later I became a CDC Israel employee as a Systems Analyst (the same, but the nights were distributed over several sites).
If anybody is interested I've got any number of stories about these machines, such as the procedure for manually polishing the disk surfaces after a head crash, or how the machine was deadstarted using a 12x12 array of toggles.
It was a good solid foundation, still serving me well today.
Yeah, I played with PlayDoh on the plasma terminals for a while, till I got bored with it.
The most impressive thing I remember about it was the eerie green tint the world had after my play session was over and I had to rejoin the real world. That orange plasma screen did real strange things to your vision!
I joined CDC UK in 1970 and was first employed at the University of London Computer Centre which had a 6600 and later a 6400 and 7600. SIA in Victoria also had a 6600. I remember being blown away at the way the Cray team had designed the hardware. Brilliant!!! I still remember tuning clocks and 7 inches being equivalent to a Nanosecond. Ho Hum
In 1966 I was a senior Math major at UT in Austin. To take classes in which you could learn/use programming, you had to take them as math or science or engineering classes because the UT CS Dept. didn't open until fall of '66, which was after I graduated. To help pay for my math degree I worked as a student computer operator at the UT Computation Center (UTCC) in 1965 and 1966.
Until summer of '66 the "mainframe" we had at UTCC was a CDC 1604. As the predecessor to the UTCC CDC 6600, it featured 48-bit words (eight 6-bit characters/word) and had 32K words of memory composed of tiny magnetic doughnuts commonly called cores. 1604 I/O was via eight Ampex 7-track, pinch-roller (read noisy!) tape drives (200, 556, and 800 BPI). It had no disks and the tape drives had to be PM'ed every AM by one of the two resident CDC engineers. After a successful PM "The Eyes"* was played through the 1604's operator console speaker. Upon hearing "The Eyes", the 1604 operator would know it was time to go to work for the day.
Summer of '66 saw the UTCC installation of the CDC 6600, which replaced the 1604 (and most of its auxiliary equipment, including the Ampex drives). As a UTCC employee I was privileged to help with the 6600's installation (mostly by staying out of the way!). Several more CDC engineers were on hand to accomplish the install. I remember internal wire length as being a critical issue to the optimal functioning of the new machine. As the pics in this article show, there were many internal wires. My recollection is that the 6600 used no integrated circuits; circuitry was implemented in discrete components, hence all the wires and internal cabinet cooling for the circuitry.
Our 6600 had 128K 60-bit words (ten 6-bit characters/word) of memory. I don't remember if that included memory for the ten peripheral processors. At the time, it just seemed incomprehensibly large!
Directly-attached, fast CDC tape drives, drum line printers, card readers and the sexy-in-its-day looking dual CRT operator console gave an enormous boost to UTCC throughput. I don't know how long the 6600 stayed in production at UTCC, but I suspect it was many years because of its superlative utility.
Our 6600 came with a new-to-us type of hardware - disk storage. The disk got everybody's attention, in more ways than one. On the plus side it was way fast; on the minus side its MTBF was way small.
Years later I learned from a CE whose job it had been to keep those disks running, that they were known among the CEs as giant disk brakes! The failure rate proved to be unacceptably short and the failures could be catastrophic, eventuating in a complete mechanical overhaul. Not only would data be lost by these failures, but physically rebuilding the drive could take an unacceptably long time. The read/write heads were hydraulically actuated, which made the head-positioning mechanism complicated and cumbersome to maintain.
After I graduated at the end of the summer of '66 I lost track of the 6600 and its disk problems, but my guess was the hydraulic disk was replaced by another, more reliable disk solution. To its credit the disk was not a CDC product; apparently CDC had arranged with the disk manufacturer to adapt their drive to the 6600.
The promise of supercomputing attendant with the UTCC's 6600 installation was immediately recognized by UTCC staff, UT faculty and students. A mathematical convergence Fortran test program I wrote for a math class ran orders of magnitude faster on the 6600 than on any other computer on campus at that time. In my first job out of college we used a CDC 3150 computer, which was certainly a step up from the old UTCC 1604, but having used the 6600 as a student/employee made me long for my UTCC days on the 6600.
*"The Eyes of Texas", of course.
Fun article (particularly since I live and work in Minneapolis, erstwhile home of many supercomputing machines, including the Cray headquarters for awhile).
However, I particularly liked your nod to Joel and the Bots there in the last line. (Also fellow Minneapolitans, come to think of it...)
There is a high-performance emulator called "Desktop CYBER" for the CDC 6000 and CYBER series available at:
The emulator supports a wide range of peripherals and runs all 60 bit operating systems including COS, SCOPE, KRONOS, NOS 1, NOS 2 and NOS/BE.
There used to be an open source version of the emulator, but it has disappeared from the website above.
Outstanding reminder of my past days at Purdue University. We had a state-wide timeshare network running on IBM-7094 hardware that ran as a batch processing front end to the CDC machine.
On one experiment in our Physics Department, I consulted with the graduate student and his professor because they were burning all of their NSF grant money up on the IBM SYS/360 they used. We recoded the electron density probability study (large matrices and double precision floating point) in FORTRAN on the CDC and it ran like greased lighting -- 4 hour runs reduced to 15 minutes.
You could actually run from the 8 word instruction stack on the processor. If you hand-optimized your code, it was possible to make an inner-product matrix multiplication loop run inside the stack. Of course, this also balanced performance on the two ALU's in the processor.
If I am not mistaken, the initial description of the CPU and PPUs was incorrect in the fact that the PPUs were actually 15-bit computers instead of the referenced 12-bit.
We had Serial #001 CDC 6500 at Michigan State University in the early 70s. Learned Fortran, LISP, and played wicked Star Trek on that beast. I loved the SCOPE OS. It was nice. It was clean. It was almost as good as Unix! Anyway, thanks for this blast from the past, as an operator, we could also have great fun getting the tape drives to open and close in sequence. Grand old machines. You are right, it's a shame modern 'puter designers can't take a lesson from the greats of the past!
It was a lovely machine to write code for. I used one at the University of London in the later 1960s - I've still got my SCOPE, FTN and Ascent F manuals. Is there anybody else out there who used the SYSTEMC routine ? This was a wickedly useful procedure which allowed a running program to get directly at the OS and do things like over-riding fatal errors, mess around with your job control limits and other naughty things. CDC tucked the documentation for this practically out of sight in an appendix to the SCOPE reference manual. Once again, it was a wonderful machine to work with.