Xilinx is building the adaptable, intelligent world

Xilinx is building the adaptable, intelligent world - Edinburgh Technopole

We want to increase the speed of applications such as machine learning and artificial intelligence, and fit them into conventional data structure platforms. The advantage of our solutions is that they can be easily customised, and even changed remotely in the field.

Colin Carruthers, Director of Xilinx Scotland

Xilinx, Inc., headquartered in California’s Silicon Valley, develops highly flexible and adaptive silicon processing platforms that enable rapid innovation across a variety of technologies. The company – which invented field-programmable gate arrays (FPGAs), hardware-programmable systems on chips (SoCs) and the adaptive compute acceleration platform (ACAP) – has sites all around the globe, including Xilinx Scotland at Edinburgh Technopole, which has the distinction of being its first research and development site outside of San Jose.

Towards the end of the 20th century, the explosion of the dot-com bubble led to a fiercely competitive market and difficulties recruiting high calibre staff . Xilinx realised that there were advantages in establishing R&D facilities outside North America, as this would provide access to an additional talent pool. This led to the 1993 opening of Xilinx Scotland, formed through the acquisition of a small spin-out company from the University of Edinburgh. Since then, the company has gone from strength to strength, outgrowing its original premises and moving to its current Edinburgh Technopole base, set in mature parkland within easy commuting distance of the city.

Usually, a silicon chip is designed to perform one specific function, but there’s an alternative, as Colin Carruthers, Director of Xilinx Scotland, explained: “Xilinx devices are quite different to the conventional silicon chip, as they are more like blank canvases. This means that they continue to offer all the benefits of a custom chip, but have the huge advantage that system designers can configure them to their specific needs. It’s a bit like software programming, but you’re actually configuring the chip to perform different functions at different moments in time. The major benefits for our customers are programmability, flexibility, low latency and accelerated time to market, as they no longer have to design their own chip.”

The focus at Edinburgh Technopole is on product development, using research and innovation to devise radical ways of solving problems and producing novel architectures with a commercial advantage. A large part of this work is related to the communications infrastructure, for example, the emergence of 5G technology. “Historically, we’ve worked a lot on 4G technology for mobile phones and infrastructure, but now there’s a big push on 5G. That’s exciting for us, because it’s a standard that is not quite solidified,” Colin continued. “As it adapts and changes, we have the ability to just tweak the configuration on our devices – there is no need for a complete redesign. The challenge for designers is not only meeting the performance specifications, it’s also about size. If you can make your device smaller and still achieve the same speed, it will consume less power and be a more cost-effective solution for the customer.”

Increasingly, the company is looking at data centre, machine learning and artificial intelligence applications for its devices, the most recent being its Adaptive Compute Acceleration Platform. “We want to increase the speed of applications such as machine learning and artificial intelligence, and fit them into conventional data structure platforms. The advantage of our solutions is that they can be easily customised, and even changed remotely in the field. This dynamic reconfiguration is ideal for data centre applications, and it’s more environmentally friendly too; when something new comes along, customers can simply refactor the existing platform, rather than replacing an entire board,” Colin concluded.