Oral history of British science
Furber, Steve (Part 4 of 5)
The British Library Board acknowledges the intellectual property rights of those named as contributors to this recording and the rights of those not identified.
Legal and ethical usage »
Interviewee's office, University of Manchester
Furber, Steve, 1953- (speaker, male)
Lean, Thomas (speaker, male)
Part 4 [1:45:54][9 November 2012] Comments on: anticipated users for BBC micro; early market for personal computers in UK; anticipated uses for early personal computers, learning about computers, limited anticipation of networking; success of Acorn Atom and associated software support; [05:00] feedback from users, small company nature of Acorn; addressing technical problems from users, difficulty soldering Atom; astonishing success of BBC Micro, anecdote about retailers selling BBC Micro at no margin; position of Acorn and BBC Micro in early 1980s computer market, comparisons with Sinclair, value of BBC to Acorn brand; overstretch of Acorn in trying to expand into new markets in USA and Germany, problems of BBC Micro in USA market meeting electromagnetic emissions tests; [11:57] Acorn learning from experience of problems selling to US market; reasons for Acorn failure in US market, competition with Apple, scale of USA; losses from write off of American venture and unsold Acorn Electrons, after problems producing Acorn Electron in 1983, leading to takeover by Olivetti. [15:00] Comments on Acorn Electron: market driven initiative; cost reductions from BBC micro, cheaper memory chips, better ULAs; technical details of machine; anecdote about small cardboard mock-up case; financial success thwarted by technical problems; technical group considering the electron a retrograde step driven by competition with Sinclair; benefits of experience for SF in developing microchip development skills; [20:45] school holiday pupil developing field-programmable gate array [FPGA] computer development tools; [21:15] Detailed description of designing ULAs; [26:40] Remarks on: school vacation pupil John Cox who wrote development tools; description of FPGA, field-programmable gate array, and Ferranti uncommitted logic array ULA used in BBC Micro, multi-purpose chips whose functionality was defined at the final stage. [29:05] Comments on BBC Micro: development of specification by BBC and Acorn; expansion routes of machine; Acorn scheme for adding a second processor through The Tube port, to allow Z80 processor and CP/M; BBC tele-software connection using PRESTEL network; floppy disc interface; comparisons of excitement around BBC Micro and current Raspbery Pi; use by professionals. [34:10] Remarks on SF limited direct contact with BBC: anecdote about SF and Sophie Wilson at first recording of 'The Computer Programme'; SF involvement with Tube and ARM designs. [35:35] Remarks on BBC Doomsday project, c1986: plan to store data on Britian collected by school children and stored on laser discs; anecdote about Doomsday Reloaded allowing access to project today, which was too expensive at the time. [37:50] Comments on rivalry with Sinclair in early 1980s: Chris Curry earlier role in Sinclair; success of Acorn overwhelming most competition; SF never meeting Clive Sinclair and limited contact with Sinclair workers as he lived outside Cambridge; Chris Curry and Hermann Hauser positions on rivalry. [41:55] Remarks on SF limited involvement with BBC micro updates, central role in Electron. [42:35] Story about start of ARM: SF and Wilson thinking beyond 8-bit processors, c1983; SF unimpressed with existing 16-bit processors, importance of memory bandwidth to computer performance, comparisons with 6502 interrupt performance on BBC Micro; Hermann Hauser supplying papers on Berkeley RISC [reduced instruction set computer] machine; RISC's optimisation for use in a microcomputer rather than a minicomputer, by keeping instruction set simple and using pipe-lining; description of pipe-lining of instructions; influence of Berkeley example of creating a high performance simple processor; Stanford developed Microprocessor without Interlocking Pipeline Stages [MIPS]; [50:43] Wilson sketching instruction sets; desire to make commercial microprocessor; commercial microprocessor development normally needing huge teams, examples of National Semiconductor design centre in Israel; anecdote about SF and Wilson trip to Phoenix, USA, to visit processor development centre, and realising that development could be done successfully on a small scale; [54:15] Remarks on SF early opinions on RISC: expecting RISC to widely taken up quickly; early developments to inform Acorn's choice of which RISC processor to buy; industry poor reactions to RISC ideas; concerns over code density; Wilson approach incorporating academic RISC ideas and commercial experience. [57:05] Story outlining development of first ARM chip: SF developing micro architecture for Wilson's instruction set, then block specifications; Acorn chip design team of Robert Heaton, Jamie Urquet, Harry Oldham and Dave Howard; use of double metal layers in chips; outline of ARM dates between 1983 and 1985; anecdote about visit to VLSI offices in Munich for 'taping out' of the chip; testing of first ARM chip in 1985. [1:02:05] Remarks on: transistor count and performance of ARM compared to Motorola 68020; ARM's use of trick in memory array layout to improve bandwidth; support from Hauser for developing ARM; split of technical and commercial activities in Acorn between Hauser and Curry; Hauser recruiting Jim Mitchell and setting up the Acorn Palo Alto research laboratory to develop operating system for ARM; rapid growth of Acorn; [1:07:45] expecting ARM development to be a learning exercise rather than successful chip development at first; software testing team to support hardware development, including Hugo Tyson, John Thackray, David Sewell; anecdote about moving testing from Wilson's software emulator to SF's hardware reference model. [11:11:10] Description of instruction set: abstraction layer between software and hardware; guiding philosophy of ARM's instruction set, compared to National Semiconductor 32016; anecdote about greater complexity of Intel processor compared to ARM; [1:16:26] anecdote about Hermann Hauser giving ARM team the advantages of no money and no people, necessitating simple design for ARM; examples of ARM's simplicity and compromise in design from pure RISC. [1:19:00] Remarks on: early expectation of ARM as to power Acorn's products; Palo Alto operating system research centre, closed when Acorn encountered problems; limitations of early Archimedes Arthur operating system, which became more powerful RISC OS; anecdote about poor performance of modern computers compared to older RISC OS computers; description of RISC OS use of cooperative rather than pre-emptive multitasking, comparisons with Apple iPhone multitasking. [1:23:19] Comments on interaction between hardware and operating system designers: SF working closely with Wilson at Acorn; influence of operating system design on chip design, with description of example paged memory system development for ARM1, which left chip with feature that was initially redundant; [1:29:00] description of context switches. [1:30:20] Remarks on decision making process as a team: meetings including SF, Wilson, Tudor Brown, Mike Muller, at Acorn; ARM's use with Video, IO, and memory controller chips; typical consensus decisions; Andrew McKernan leading group for a time; SF promotion to lead hardware team under research director Jim Mitchell; flat structure; anecdote about objecting to management attempts to impose different bonus levels on development team; value of dissenting voices from Wilson and Mike Muller in team discussions; involvement of Hauser. [1:35:05] Remarks on daily work activities: anecdote about SF recently rediscovering original ARM reference model, spending time in 1980s working on it; travel to Phoenix and Israel; Wilson, SF and Chris Turner touring Britain with a BBC Micro seminar in early 1980s; anecdote about giving a seminar on BBC to an overfilled Savoy Place; [1:39:35] SF feelings about designing a processor, great fun but challenging; tight Acorn development team, comparisons with later university experience; anecdote about daily visits to pub for lunch; trauma of Acorn financial problems, Olivetti having little awareness of ARM before takeover; anecdote about journalist's disbelief at ARM in 1985; climax of chip development process in 1985 when working chips came back from manufacturer; anecdote about tradition of champaign for successful chips; naming of ARM as Acorn RISC Machine, then Advanced RISC Machine.
Interview with computer scientist, Steve Furber