For decades it has been the benchmark by which advancements in computing are measured.

Now Moore's Law - the maxim which states that computers double in speed roughly every two years - has come under threat, from none other than the man who coined it.

Gordon Moore, the retired co-founder of Intel, wrote an influential paper in 1965 called 'Cramming more components onto integrated circuits', in which he theorised that the number of transistors on a computer chip would double at a constant rate.

Silicon Valley has kept up with his widely accepted maxim for more than 40 years, to the point where a new generation of chips, which Intel will begin to produce next year, will have transistors so tiny that four million of them could fit on the head of a pin.

In an interview yesterday, however, Mr Moore said by about 2020, his law would come up against a rather intractable stumbling block: the laws of physics.

"Another decade, a decade and a half, I think we'll hit something fairly fundamental," Mr Moore said at Intel's twice-annual technology conference. Then Moore's Law will be no more.

Mr Moore was speaking as Intel gave its first demonstration of a new family of processors, to be introduced in November, which contain circuitry 45 nanometres - billionths of a metre - wide.

The 'Penryn' processors, 15 of which will be introduced this year, with another 20 to follow in the first quarter of 2008, will be so advanced that a single chip will contain as many as 820 million transistors.

Computer experts said today that a failure to live up to Moore's Law would not limit the ultimate speed at which computers could run. Instead, the technology used to manufacture chips would shift.

The current method of Silicon-based manufacturing is known as "bulk CMOS", which is essentially a 'top-down' approach, where the maker starts with a piece of Silicon and 'etches out' the parts that aren't needed.

"The technology which will replace this is a bottom-up approach, where chips will be assembled using individual atoms or molecules, a type of nanotechnology," Jim Tully, chief of research for semi-conductors at Gartner, the analyst, said.

"It's not standardised yet - people are still experimenting - but you might refer to this new breed of chips as 'molecular devices'."

Anthony Finkelstein, head of computer science at University College London, said, however, that a more pressing problem in the meantime was to write programs which took full advantage of existing technologies. "It's all very well having multicore chips in desktop machines, but if the software does not take advantage of them, you gain no benefit."

"We are hitting the software barrier before we hit the physical barrier," he said.

Mr Moore, who is 78, pioneered the design of the integrated circuit, and went on to co-found Intel in 1968, where he served as chief executive between 1975 and 1987.

Asked what area he would research if he were a graduate today, he told the conference: "I'd probably look at something more in the biology mould. The interface between computers and biology is a very interesting area."

Explore Tech & Web

• Personal Tech

• The Web

• Gadgets & Gaming