Minutes of the meeting held on 23/
7/1968 at State House, London
-
Present
- Lord Halsbury (Chair)
- Dr R A Buckingham
- Professor T Kilburn
- Dr T G Pickavance
- Mr L F Rutterford
- Dr J Howlett
- Mr C Jolliffe
- Dr J H Price
- Mr R St J Walker
- C M Wilson ICL
- J C Deas ICL
-
Apologies
- Dr A H Halsey
- Professor E S Page
- Professor R E Peierls
- Professor M V Wilkes
1 Congratulations
Before starting the formal business of the meeting,
the Committee recorded its very great pleasure in the award
of a Knighthood to Sir Rudolf Peirls.
2 Purchase of New Computer
- The meeting was called to discuss a single item, the purchase
of a new computer for the Atlas Laboratory. at the
previous meeting held on 26th April 1968 the Committee had discussed
the proposal for an ICT P.51 system described in
ACC/68/4, but had not been able to reach a decision because of
lack of information from ICT (Minute 6 of the meeting refers).
Since that time the merger of ICT and English Electric
Computers had been effected and the new company ICL formed
by the merger had settled its production policy for the 1900
and System 4 ranges and given the Committee (in confidence for
the time being) a technical specification for the P.51 system.
the offer described in ACC/68/4 was now confirmed in every
respect - performance, price and delivery - and could therefore
be taken as a firm basis for discussion.
The fast processor would in future be called the 1908A
- The position after the 26th April meeting had been reported
to the UST Board at its meeting on 25th June 1968. The
Board (Minute 207) agreed to support the proposal if the Atlas
Committee, after being given the information it needed, recommended
it to do so. The Board appointed a sub-committee
(Sir Ewart Jones, Lord Halsbury, Professor Ford, Dr Gavin) to
act on its behalf on receiving the Atlas Committee's decision.
- The meeting was conducted on the basis of ACC/68/4, supplemented
by the technical document from ICL, a copy of
which had been sent to each member. At the suggestion of the
Chairman, ICL was invited to send representatives to answer
questions on the proposed system. Dr C M Wilson and Mr J C Deas attended in this capacity.
- Questions to the ICL representatives, and discussions
arising from the replies, took up the greater part of the
meeting. The main points made were as follows:
- ICL were firmly convinced that the whole project was
sound and that the delivery dates they had offered were
realistic. They were confident of getting adequate
supplies of the advanced components which they needed - in particular
the 330ns store, the 50ns and 100ns
buffer stores, the mass core store and the 1ns integrated
circuits. They were expecting to sell some
30-40 1906A installations, but knew that several potential
customers would not buy the machine unless they knew that
the 1908A was to be produced.
- The twin-processor system could be either 1906A + 1908A
(as proposed) or 1908A + 1908A. Provisional twin-system
software would be supplied initially (i.e. in mid-1972
for the Chilton installation) which would allow, for
example, sharing of the file store by the two processors,
but a fully integrated operating system would not be available
until mid-1973.
- The first prototype 1906A had power on now and should be
running as a complete machine, with peripherals, by
January 1969. the second prototype, which would have
the proper production platters from their automated plant,
should be at the same stage of development as the first
by march 1969. Software would be developed on these
machines and on a specially-built 1905E with paged store,
due to be delivered to Putney in February 1969. Present
production plans were:
- First production 1906A in June 1970, thereafter
approximately one a month with plans for ten in 1971.
- First production 1908A in April 1972, with total of
four in 1972 and six in 1973.
- ICL had made a detailed comparative study of the architectures
of the 1900 and System 4 series and had concluded
that the former, which was much simpler, was the better
basis for the powerful machine. To provide the same computing
power with the System 4 design (which is the same as
the IBM System 360) would have to involve more complex
design, cost more and take longer.
- Asked about ICL's plans for the future, Dr Wilson said
that they were expecting to introduce a new product set
about 1975 or 1976, but much study was needed before its
form could be decided. they were very conscious of the
need to make changeover easy both from 1900 and System 4
machines, so that the very big investment in programs,
both of the Company and by users, should not be wasted.
He felt that this need to preserve the value of the
investment would ensure a long life for all the large
families of computers, including the ICL 1900 series
and especially, because of the numbers involved, the IBM 360.
His view was that there was little risk of
rapid obsolescence of these machines.
- ICL recognised that there were deficiencies in some
of the basic software of the 1900, in particular in the
operating systems and the Fortran compilers; they were
working on these and were confident of making considerable
improvements. They did not agree with the view that the
short character-length (6 bits) and small number (3) of
index registers of the 1900 were a serious drawback to
this series.
- Professor Page had re-emphasised in his letter the suggestion
he had made at the previous meeting: namely, that
the Atlas Laboratory should install the 1906A and peripherals
as proposed in ACC/68/4, and the mass store, as soon as possible,
and rent from ICL a second 1906A until the 1908A was available.
The Laboratory could then start to develop the twin-system
operating system, preferably in collaboration with
ICL, with the expectation of having this working by the time
the new processor was delivered. This would also provide more
processing power in the years 1970-72, when there was virtually
certain to be a serious shortage throughout the universities.
The ICL representatives said this seemed feasible but needed
study; the Company was not proposing, in general, to rent
1906A systems. Mr Walker said there were financial problems,
especially as it was now certain that the Council's grant would
be cut. Dr Howlett undertook to discuss all aspects of the
suggestion with ICL and Finance Branch.
- After the ICL representatives had left the meeting, the
Committee briefly reviewed the arguments - discussed in
the preceding three meetings - in favour of a new machine at
Chilton, and agreed that these were still valid; and agreed
that ICL had now removed the doubts and uncertainties about
the new project. They recorded their decision to recommend
to the UST Board that the proposal, as already put to the
Board provisionally, should be supported. Details of the make-up
and prices of the proposed system are given as Appendix I to these
minutes.
3 Appendix I: Prices of New System
|
£000 |
£000 |
1906A System |
1906A with 128K store, 750ns includes extended precision Floating Point
Unit, paging, 1 high-speed channel, 5 extra slow channels |
|
460.0 |
Peripherals |
|
|
1 × 4-deck cluster 80Kcs/9 track magnetic tape |
42.0 |
|
1 × 2-deck cluster 60Kcs/7 track magnetic tape |
26.5 |
|
1 × 420M ch disc file with control |
132.0 |
|
1 × 2M ch magnetic drum with control |
49.0 |
|
2 × 1,600 cpm card readers (with binary image) |
19.0 |
|
2 × 1,000 ch/sec paper tape readers |
12.8 |
|
1 × 110 ch/sec paper tape punch |
4.5 |
|
1 × 100 card/min card punch |
7.0 |
|
2 × 1,350 line/min printers |
44.0 |
336.8 |
Communications |
PF504 control computer with
63 telegraph and
16 telephone lines |
|
61.6 |
Total: 1906A System |
|
858.4 |
1908A System |
Processor with 128K store, 330ns |
|
1,120.0 |
Twinning hardware (see Note 1) |
|
50.0 |
Peripherals |
1 × 2M ch magnetic drum with control |
49.0 |
|
1 × 4-deck cluster 160Kcs/9 track magnetic tape |
55.0 |
104.0 |
Mass store, 512K, 1µs |
|
250.0 |
Total: 1908A system |
|
1,524.0 |
Note 1: This refers to the switch and connections which
link the two processors and allow
the peripherals to be assigned
to the processor in any way.