See also: Cutting power consumption in complex FPGAs
The programmable logic sector has been lacklustre for a few years: no new ideas, not much growth, no real success in solving its basic drawbacks of cost and power. But the X-factor – the Xilinx and XMOS factor – could be about to take effect.
“We are perceived to be stuck in a $3.6bn niche,” says Moshe Gavrielov, CEO of Xilinx. Xilinx has decided to try to break the $3.6bn stranglehold by making its parts more accessible to more people.
“In the past we have focused on the underlying technology in the FPGA,” says Brent Przybus, director of product marketing at Xilinx. “We have said: ‘How great it is to have a six-input LUT’. Now, when we talk to customers we are not just talking to FPGA customers any more.”
What Xilinx is doing is producing tools with which non-FPGA users can very quickly – in a week or so – learn how to programme an FPGA.
“Engineers who have not used FPGAs before can use them using these evaluation kits, and they can learn to do it in a week,” says Przybus. “Everything is there to enable engineers to go to the next step.”
The kits are rapidly expanding the Xilinx customer base. “In six months we have added 650 early access customers – people who haven’t used FPGAs before,” says Przybus. Xilinx now has 1,000 early access customers, compared with 350 at the end of last year.
That is not hugely additive to Xilinx’s 37,000 customers, but it is a start. Coming along are more specific evaluation kits aimed first at engineers in particular technical areas, and later on at engineers in particular markets.
The first of these, called “domain-specific”, will be for connectivity, embedded and DSP applications; then come “market-specific” kits aimed at communications (both wired and wireless), video processing, broadcast, aerospace and defence, automotive, surveillance, and consumer.
“That is how we get out of the $3.6bn number and to a bigger number,” says Przybus.
See also: Actel to offer first ARM Cortex-M3 processor-based FPGA
Programmability
The other X-factor is the XMOS proposition that because many engineers can programme in C, if you give them an inexpensive, low-power, easy-to-use C-programmable chip which can perform significant functions, they can get to market quickly and inexpensively.
The Bristol-based start-up’s latest chip costs $5, has eight threads, performs 400MIPS, has 64kbyte of SRAM, and uses less than 500µA in sleep mode and 15mA in standby. It has dual-core and four-core chips on the market.
XMOS already has one company which has put its own product design into an XMOS chip and is selling it under its own brand.
“They got a product to market in nine months at a cost of $100,000 and they are now shipping 25,000 to 30,000 ICs a month,” says James Foster, CEO of XMOS.
Value proposition
With the problems which venture capitalists are having in finding the millions of dollars needed to fund a conventional chip start-up, XMOS feels it has a value proposition in allowing a semiconductor start-up to get a product to market for $100,000.
Apart from the XMOS customer which is already shipping its product design in volume using an XMOS chip, XMOS is working with two other semiconductor companies to get their products into the market using XMOS ICs.
“It can take only two to three months to get a product from concept to shipping chips,” says Foster. “If we can help people bring new innovative IP to market, we will help them and work with them – we are all entrepreneurs here. It is the perfect time to go to market because the market is so predisposed to change. People are looking at innovative ways to bring products to market.”
So, after a dreary period for the programmable industry, the green shoots of innovation could be emerging. It all depends on the X-factor.
See also: Altera offers tamper-proof Cyclone III FPGAs