Article 1098 of comp.sys.cdc: Path: matra.meer.net!news.spies.com!genmagic!sgigate.sgi.com!uhog.mit.edu!news.kei.com!newsfeed.internetmci.com!in1.uu.net!news.eng.convex.com!not-for-mail From: barnett@news.eng.convex.com (Paul Barnett) Newsgroups: comp.sys.cdc Subject: Re: What About the CDC 160-A? Date: 13 Jan 1996 12:32:46 -0600 Organization: CONVEX News Network, Go Fishing (worm), Richardson, Tx USA Lines: 32 Distribution: inet Message-ID: <4d8tse$mn8@jaguar.convex.com> References: <4d7c0b$sil@dns.plano.net> NNTP-Posting-Host: jaguar.convex.com In <4d7c0b$sil@dns.plano.net> Charles Richmond writes: >I work with an aging crone (ok, he is 10 years older than me) and he >tells me about a minicomputer he worked with in the early 60's. It was a >12-bit minicomputer called the CDC 160-A, a 1963 vintage machine. He >used it when working for NASA to process telemetry data. >Does anyone know any more about this box? By chance, does anyone have an >OSAS-A assembly language manual that accompanied this machine? I've never actually seen a 160A, but in the early 80's I worked for Control Data's OS development at Arden Hills. One of the engineers had a 160A assembly language manual, and I thumbed through it once. The instruction set is very similar to the peripheral processors (PP's) that were used for I/O and stuff on the 6000 series computers and all of its successors -- the Cyber 70, 170, 700, and even the 12-bit mode of the 800's and 900's. The 'PP' had a single 18-bit general purpose register, which made some shifting operations easier. The trick was to use the first 77(base 8) words in memory for as much data as possible, since the instructions to access them occupied only a single 12-bit word, while the rest of memory required a 24-bit instruction (which was slower) for access. There was a total of 4096 words of memory -- which was the maximum addressable anyway. The basic cycle time of the PP's was one microsecond in the first generation, and got as low as 250 nanoseconds in later models. Instructions typically took from 2-5 cycles. I'm not sure how this compares to the speed of the 160A. Article 1100 of comp.sys.cdc: Path: matra.meer.net!news.spies.com!genmagic!sgigate.sgi.com!uhog.mit.edu!news.mathworks.com!newsfeed.internetmci.com!vixen.cso.uiuc.edu!avenir!walker From: walker@avenir.cerl.uiuc.edu (Michael W. Walker) Newsgroups: comp.sys.cdc Subject: Re: What About the CDC 160-A? Date: 14 Jan 1996 05:42:42 GMT Organization: University of Illinois at Urbana-Champaign Lines: 66 Distribution: inet Message-ID: <4da54i$jje@vixen.cso.uiuc.edu> References: <4d7c0b$sil@dns.plano.net> <4d8tse$mn8@jaguar.convex.com> NNTP-Posting-Host: avenir.cso.uiuc.edu In article <4d8tse$mn8@jaguar.convex.com>, Paul Barnett wrote: >In <4d7c0b$sil@dns.plano.net> Charles Richmond writes: > >>I work with an aging crone (ok, he is 10 years older than me) and he >>tells me about a minicomputer he worked with in the early 60's. It was a >>12-bit minicomputer called the CDC 160-A, a 1963 vintage machine. He >>used it when working for NASA to process telemetry data. > >>Does anyone know any more about this box? By chance, does anyone have an >>OSAS-A assembly language manual that accompanied this machine? > >I've never actually seen a 160A, but in the early 80's I worked for >Control Data's OS development at Arden Hills. One of the engineers had >a 160A assembly language manual, and I thumbed through it once. > >The instruction set is very similar to the peripheral processors (PP's) >that were used for I/O and stuff on the 6000 series computers and all of >its successors -- the Cyber 70, 170, 700, and even the 12-bit mode of >the 800's and 900's. > >The 'PP' had a single 18-bit general purpose register, which made some >shifting operations easier. The trick was to use the first 77(base 8) >words in memory for as much data as possible, since the instructions to >access them occupied only a single 12-bit word, while the rest of memory >required a 24-bit instruction (which was slower) for access. There was >a total of 4096 words of memory -- which was the maximum addressable >anyway. > >The basic cycle time of the PP's was one microsecond in the first >generation, and got as low as 250 nanoseconds in later models. >Instructions typically took from 2-5 cycles. I'm not sure how this >compares to the speed of the 160A. > I believe the 160 and 160A were based on the technology of the 1604, CDC's first mainframe. The 1604 had a 48 bit word, so a 12 bit I/O processor meshed well with it. We had both a 1604 and a 160 in the early days of PLATO, the 160 serving as the interface to the removable pack disk drives. I programmed the 1604, but never the 160, so my recollections of it are more vague. Later versions of the 160 were also used as I/O processors for the 3000 series. By the 6000 series, they decided to incorporate multiple I/O processors into the design and give them access to all of CM. That was presumably the reason for going to an 18 bit accumulator, as it had to be able to hold a full CM address. If I remember correctly, the big difference between a 160 and 160A was that the A had a bank switch feature that allowed the accessing of 2 X 4096 words of memory. These were pretty slow machines by today's standard...7+ microseconds per memory reference of 12 bits yields less than 2 Mbits/sec for max memory bandwidth. Both the 1604 and 160 were quite compact for their day. The 1604 CPU fit in a box roughly 8 feet long, 6 feet high and 3 feet deep. This for a 48 bit CPU and 32K of 48 bit CM (real core). A bank of 4 tape units occupied a second box of nearly the same size. The console was a desk with a pre-selectric IBM typewriter, a paper tape reader and punch, and an octal readout for the A, Q, and P registers which displayed only when the machine was stepped/stopped. The major feedback while running was through a speaker wired to the upper bits of the accumulator...you could often tell audibly how your job was progressing. The 160 was built into a desk size console using the same octal readout method, but could be much smaller due to having so many fewer digits to display. -- Michael W. Walker mwwalker@uiuc.edu University of Illinois at Urbana-Champaign