It’s good to see start-ups doing something dramatic. So many do incremental stuff – 30 per cent more performance, 30 per cent less power etc. Grand aspirations were what the semiconductor industry used to be about.
“One of the things that can make it very disruptive is because it makes it unnecessary to have any standards,” says XMOS co-founder and CTO, Professor David May, Professor of Computer Science at BristolUniversity and the architect of the Inmos Transputer.
The flexibility of XMOS’ chips, called Software Defined Silicon (SDS) means they can be re-programmed in C before or after installation to comply with anyone’s system. And the SDS chips can contain significant functionality e.g. 200Mbps Ethernet or 7MSPS 16-tap FIR. A 4-core chip delivers 1600 MIPS.
Another fine aspiration is CEO’ James’ Foster’s maxim: “What if you could start a semiconductor company with $100,000 again? Our NREs are less than $100,000 and our prototype lead-time is 30 seconds.”
“It now costs $100m to do a state of the art ASIC, and no investor will put $100m into an ASIC chip company”, points out May, “even big companies find it difficult to make a business case to do it.”
Another good aspiration is to get the electronics hobbyist back into business. “We want to allow people to be electronics hobbyists again”, says Mark Lippett, vice president of engineering, “it has been dead for many years, because electronics design has become so complicated. We’re giving the opportunity for a great deal of creativity once again. It’s a uniquely affordable technology.”
And another aspiration is to give the customer something he always wants: differentiation. “Customers tell us”, says Noel Hurley, XMOS co-founder and vice president of marketing, “‘I receive my reference design, I’ve got the software drivers already ported to it. The big issue is that, so do all my competitors. So how do I differentiate my product?'”
The answer: Put all those software engineers you employ who understand C to programme the chips to do the differentiation.
“Consumer electronics customers have built up teams of software engineers who understand C, very few understand RTL”, says Hurley.
A time-line XMOS draws is:
The Age of TTL (’60s) – 100,000 designers using pen and paper with little flexibility.
The Digital Age (’80s) – 1 million designers, utilising schematics and HDLs, designing FPGAs and SOCs around gate-level flexibility.
The Software Age (Now) – 10m+ coders/designers in the world use C to deliver system-level flexibility.
The thing is there’s an awful lot of guys who can use these things.
“I see growth in people rolling out generic technologies”, says May, “there’s no need for them to have their own IP. They configure the platform, they don’t have to do hardware design or manufacture. It sounds like the ideal business to me. We all ought to be in it.”
TOMORROW MORNING: THE TEN MOST STUPID REMARKS OF THE YEAR