Difference between revisions of "Ferranti Mark 1"

From Computer History Wiki
Jump to: navigation, search
m (Jnc moved page Ferranti Mark I to Ferranti Mark 1: Seems to have been its formal name then)
m (Further reading: +Lavington, ''Early Computing in Britain")
 
Line 1: Line 1:
The '''Ferranti Mark 1''' (whether this was its formal name at the time, or the one now often seen, the '''Mark I''', is unclear - a scan of a Ferranti marketing document from July, 1961 used 'Mark 1'; just to maximize the confusion, early on at Manchester it was called the '''Mark II''') was a very early [[computer]], the first one ever commercially available anywhere. There were two variants, the original Mark 1, and the '''Ferranti Mark 1*''', which had minor improvements.
+
The '''Ferranti Mark 1''' (whether this was its formal name at the time, or the one now often seen/used, the '''Mark I''', is unclear - a scan of a [[Ferranti]] marketing document from July, 1961 used 'Mark 1'; just to maximize the confusion, early on at Manchester it was called the '''Mark II''') was a very early [[computer]], the first one ever commercially available anywhere. There were two variants, the original Mark 1, and the '''Ferranti Mark 1*''', which had minor improvements.
  
 
The Mark 1 was a re-engineered version of the [[Manchester Mark I]]; work on it started after November, 1948 (although the final design was started in mid-1949, and basically finished by November, 1949). The primary improvements over the Manchester version were:
 
The Mark 1 was a re-engineered version of the [[Manchester Mark I]]; work on it started after November, 1948 (although the final design was started in mid-1949, and basically finished by November, 1949). The primary improvements over the Manchester version were:
Line 8: Line 8:
 
* 8 [[Williams tube]]s of [[main memory]], but with only one 32-[[word]] 'page' per tube (for increased reliability);
 
* 8 [[Williams tube]]s of [[main memory]], but with only one 32-[[word]] 'page' per tube (for increased reliability);
 
* a larger [[drum]], with 512 'pages' (8 times that of the last Manchester drum);
 
* a larger [[drum]], with 512 'pages' (8 times that of the last Manchester drum);
* the basic instruction time was 1.2 msec, instead of 1.8 msec (50% faster)
+
* the basic [[instruction]] time was 1.2 msec, instead of 1.8 msec (50% faster)
 
* a [[hardware]] random number generator.
 
* a [[hardware]] random number generator.
  
 
A few details on the improved multiplication unit: in the Manchester machine, the multiplication time was variable, and depended on the number of l's in the multiplier. An analysis of some of its typical problems showed that use of that implementation would result in about half the machine's time being spent in multiplication. A new parallel multiplier was designed, which operated in a fairly short, fixed time - on average, five times faster than the original. That multiplier was fairy complex, and its use increased the number of tubes in the machine by about ten per cent, but it was felt that the increased cost was worth the overall performance benefit.
 
A few details on the improved multiplication unit: in the Manchester machine, the multiplication time was variable, and depended on the number of l's in the multiplier. An analysis of some of its typical problems showed that use of that implementation would result in about half the machine's time being spent in multiplication. A new parallel multiplier was designed, which operated in a fairly short, fixed time - on average, five times faster than the original. That multiplier was fairy complex, and its use increased the number of tubes in the machine by about ten per cent, but it was felt that the increased cost was worth the overall performance benefit.
  
The drum stored data at a density of 165 bits per inch, around a track. An instruction was provided which could check the correctness of a transfer.
+
The drum stored data at a recording density of 165 bits per inch, around a [[track]]. An instruction was provided which could check the correctness of a transfer.
  
The Ferranti Mark 1, like the Manchester Mark I, also had 20-bit [[instruction]]s, stored two per word. It had 48 [[operation code]]s; instructions contained a 6-bit op-code, a 3-bit B-line selection, and a 9-bit memory [[address]]. (Two bits were apparently unused.)
+
The Ferranti Mark 1, like the Manchester Mark I, also had 20-bit instructions, stored two per word. It had 48 [[operation code]]s; instructions contained a 6-bit op-code, a 3-bit B-line selection, and a 9-bit memory [[address]]. (Two bits were apparently unused.)
  
 
''Note: There are inconsistencies between sources as to some of the details; take everything here with a grain of salt!''
 
''Note: There are inconsistencies between sources as to some of the details; take everything here with a grain of salt!''
Line 27: Line 27:
 
==Further reading==
 
==Further reading==
  
 +
* Simon H. Lavington, ''Early Computing in Britain: Ferranti Ltd. and Government Funding, 1948-1958'', Springer Nature, Cham, 2019 - covers the Mark 1 in some detail, including many of the sites
 
* Simon H. Lavington, ''A History of Manchester Computers'', National Computer Centre, Manchester, 1976
 
* Simon H. Lavington, ''A History of Manchester Computers'', National Computer Centre, Manchester, 1976
 
* Simon H. Lavington, ''[https://ed-thelen.org/comp-hist/EarlyBritish.html Early British Computers]'', Manchester University, Manchester, 1980
 
* Simon H. Lavington, ''[https://ed-thelen.org/comp-hist/EarlyBritish.html Early British Computers]'', Manchester University, Manchester, 1980
Line 32: Line 33:
 
* B. V. Bowden (editor), ''Faster Than Thought: A Symposium on Digital Computing Machines'', Pitman Publishing, New York, 1964 - the Mark 1 is covered in Chapter 6, pp. 117-124
 
* B. V. Bowden (editor), ''Faster Than Thought: A Symposium on Digital Computing Machines'', Pitman Publishing, New York, 1964 - the Mark 1 is covered in Chapter 6, pp. 117-124
 
* Tom Kilburn, [https://archive.org/details/in.ernet.dli.2015.67697/page/n95/mode/1up ''The New Universal Digital Computing Machine at The University of Manchester''], ''Nature'', Vol. 168, pp. 95—96 - scan of the original
 
* Tom Kilburn, [https://archive.org/details/in.ernet.dli.2015.67697/page/n95/mode/1up ''The New Universal Digital Computing Machine at The University of Manchester''], ''Nature'', Vol. 168, pp. 95—96 - scan of the original
 +
* John F. Wilson, ''Ferranti: A History - Volume I: Building a Family Business, 1882–1975'', Carnegie Publishing, Lancaster, 2001 - not much technical content, but does cover the organizational context
  
 
==External links==
 
==External links==

Latest revision as of 02:16, 27 March 2024

The Ferranti Mark 1 (whether this was its formal name at the time, or the one now often seen/used, the Mark I, is unclear - a scan of a Ferranti marketing document from July, 1961 used 'Mark 1'; just to maximize the confusion, early on at Manchester it was called the Mark II) was a very early computer, the first one ever commercially available anywhere. There were two variants, the original Mark 1, and the Ferranti Mark 1*, which had minor improvements.

The Mark 1 was a re-engineered version of the Manchester Mark I; work on it started after November, 1948 (although the final design was started in mid-1949, and basically finished by November, 1949). The primary improvements over the Manchester version were:

  • 8 'B-line' index registers, instead of 2;
  • B-lines could be used for loop control;
  • an improved multiplication unit;
  • 8 Williams tubes of main memory, but with only one 32-word 'page' per tube (for increased reliability);
  • a larger drum, with 512 'pages' (8 times that of the last Manchester drum);
  • the basic instruction time was 1.2 msec, instead of 1.8 msec (50% faster)
  • a hardware random number generator.

A few details on the improved multiplication unit: in the Manchester machine, the multiplication time was variable, and depended on the number of l's in the multiplier. An analysis of some of its typical problems showed that use of that implementation would result in about half the machine's time being spent in multiplication. A new parallel multiplier was designed, which operated in a fairly short, fixed time - on average, five times faster than the original. That multiplier was fairy complex, and its use increased the number of tubes in the machine by about ten per cent, but it was felt that the increased cost was worth the overall performance benefit.

The drum stored data at a recording density of 165 bits per inch, around a track. An instruction was provided which could check the correctness of a transfer.

The Ferranti Mark 1, like the Manchester Mark I, also had 20-bit instructions, stored two per word. It had 48 operation codes; instructions contained a 6-bit op-code, a 3-bit B-line selection, and a 9-bit memory address. (Two bits were apparently unused.)

Note: There are inconsistencies between sources as to some of the details; take everything here with a grain of salt!

The Ferranti Mark 1* had only minor changes from the original, most of them intended to make life simpler for the programmers; the instruction set thus contained only 30 instructions, instead of the original 48. It also supported only signed numbers, instead of both signed and un-signed.

Two of the original Mark 1 machines were produced (the first one, for Manchester, was delivered in February, 1951); and seven of the improved Mark 1* type (deliveries from September, 1953 through October, 1957).

Further reading

  • Simon H. Lavington, Early Computing in Britain: Ferranti Ltd. and Government Funding, 1948-1958, Springer Nature, Cham, 2019 - covers the Mark 1 in some detail, including many of the sites
  • Simon H. Lavington, A History of Manchester Computers, National Computer Centre, Manchester, 1976
  • Simon H. Lavington, Early British Computers, Manchester University, Manchester, 1980
  • Raúl Rojas, Ulf Hashagen (editors), The First Computers — History and Architectures, MIT Press, Cambridge, 2000 - the Mark 1 is covered in Part IV-2, pp. 373-377
  • B. V. Bowden (editor), Faster Than Thought: A Symposium on Digital Computing Machines, Pitman Publishing, New York, 1964 - the Mark 1 is covered in Chapter 6, pp. 117-124
  • Tom Kilburn, The New Universal Digital Computing Machine at The University of Manchester, Nature, Vol. 168, pp. 95—96 - scan of the original
  • John F. Wilson, Ferranti: A History - Volume I: Building a Family Business, 1882–1975, Carnegie Publishing, Lancaster, 2001 - not much technical content, but does cover the organizational context

External links