Microchip is 30 years old

By Anand Parthasarathy

KOCHI, NOV. 17.A development contract from a Japanese company, for a set of chips needed to power an electronic calculator, triggered off what became the world's first computer-on-a-chip: the microprocessor. Thirty years ago this week, Intel, then a small Santa Clara(US)-based manufacturer of memory and switching devices, unveiled a thumbnail-sized slab of silicon encased in a ceramic casing with 16 pins: it was called the 4004 - and it launched the era of general purpose desktop computing.

The contract with Busicom, required Intel to deliver the electronics for a calculating machine, as a set of 12 custom- built chips. The electrical engineer assigned to handle the task - Dr. Marcian ``Ted'' Hoff, a Ph.D from the Stanford University - had an inspired thought. Why make a set of chips for just one application? Why not generalise the design so that other computing tools could be built? In the process, Dr. Hoff also suggested that the composite elements of a computer, as it was then understood - a unit to do the arithmetic, a small memory, input and output interfaces - could be combined in a single slab of silicon.

When Busicom had cash flow problems, Intel bought back the design it had created and slipped the product into its own catalogue. The 4004 chip had 2300 transistors on board, flipped 108,000 times a second (108KHz) and cost $299 apiece.

The potential of this (for those days) pricey product was first realised by government agencies working on mission- critical applications. When the specification sheets of the new chip were distributed in India in late 1971, by Intel's Indian agent, the Hyderabad-based ``Electronics International'' (later Microelectronics International), the first orders were received from Indian defence research and aerospace laboratories.

The selling price was around Rs 7,000 a chip. But even as Intel upgraded that first microprocessor with the 8008 a year later and then with the 8080 series which saw the 286,386 and 486 chips, the Personal Computer (PC) had come - and the microchip found its mass market.

In end August this year, Intel launched the latest successor to the 4004: the latest version of the Pentium 4 which is approximately 20,000 more powerful - with 42 million transistors on board and a clocking speed of 2 GHz (2 billion times a second). It costs about twice what the 1971 chip did - around Rs 15,000.

Meanwhile, Intel had become the world's number one chip manufacturer, with its processors powering eight out of 10 PCs sold worldwide - five million of them in India alone.

Over this two-decade span, all these Intel chips, as well as some landmark processors from other manufacturers - the ``PowerPC'' from Motorola, the ``Alpha'' from Digital, the ``Athlon'' from AMD as well as the power-efficient ``Crusoe'' from Transmeta - have remained true to Moore's Law.

This was a rule-of-thumb coined by the Intel co- founder, Dr. Gordon Moore, predicting that microchips would double in complexity and computing power, every 18-24 months. But for how long?

Future technologies

Today's thinking is that chips will continue to be manufactured by the current silicon process for another decade at least. At the recently-concluded Development Forum in Bangalore, the Director, Microcomputer Software Labs, Intel, Dr. Richard Wirt, told this correspondent that the next chip on the company's roadmap, ``McKinley'', which will house a whopping 220 million transistors, was due to be unveiled in 2002. By 2003, Hewlett Packard is due to launch its dual processor-on-a-chip ``Mako'', while AMD is set to introduce the ``Hanover'' - a chip based on the premise that if the signals are to flow even faster, the plumbing must be improved.

After that, it may be a case of `thus far and no further' as far as silicon is concerned; scientists apprehend that a physical lakshman rekha will be reached, beyond which no more transistors can be squeezed on a given slab of the material, without the interconnecting copper or aluminium leads shorting.

When this limit is reached, chip makers may have to switch to radical new techniques.

The New York Times reported last week that two different solutions are in the works: the IBM-led effort is looking at rolled-up cylinders of carbon, known as ``nanotubes'', as a successor to silicon. However, a team at the Harvard University, led by Dr. Charles M. Lieber, hopes to stick with silicon - but instead of carving millions of transistors on the silicon, they hope to ``grow'' these transistors as tiny crystal rods or ``nanowires'' of the material and then squirting them on a silicon oxide wafer. The work is described in a paper by the Lieber team in the latest issue of the American journal Science.

But whatever the technology that finally emerges from the shakeout, one thing is clear: in the fourth decade of the microchip, a significant proportion of applications that run on these chips and the software that helps customer migrate to faster newer processors, will originate in India.

Send this article to Friends by E-Mail

Section  : Business
Previous : Dr. Reddy's loses exclusivity
Next     : Ion Exchange shifts focus to value addition