Tony King-Smith of Imagination Technologies discusses the need for RF IC designers, IP platforms, IoT as the new embedded and fabric-based SoCs.
By John Blyler, Editorial Director, JB Systems
At this year’s Imagination Summit 2015 in Silicon Valley, John Blyler, editorial director for “IoT Embedded Systems” sat down with Tony King-Smith, the Executive VP of Marketing at Imagination, to talk about RF design, IP platforms, embedded IoT and fabric-based SoC trends. What follows are excerpts from that conversation. – JB
Blyler: What is the prospect for RF IC engineers in the IoT market?
King-Smith: The prospects are very good for RF engineers because the fundamental of IoT is the connectivity of everything. This connectivity needs to be very cost engineered which means that the RF-to-baseband system has to be on-chip for many of the solutions. And that means that the demand for RF engineering should rise. As evidence, we are looking at RF ecosystem partners to complement what we do. Further, we’ve built our own RF team in-house (part of our Ensigma team that’s develops communication physical IP) so we know the importance of the cost effective design of a communication engine. We offer a full antenna and RF front-end to baseband system.
Blyler: The RF side of the IoT already seems pretty well defined. Do you see the need for RF innovation? Or, since IoT connectivity will require many more radios, will the RF challenge be one of integrating a lot of different RF devices?
King-Smith: You must remember that when you’re integrating on chip it isn’t just about the definition of the radio. Rather, it’s adapting and tuning everything for the individual SOC. Every SOC will have different characteristics, for example, the noise characteristics in the substrate or packaging requirements and the like. Also, the behavior of the RF block will change depending upon its location on the floor plan. And you must consider the complex feedback loops between the RF and the baseband. All of these things require RF expertise and often a lot of customizing of services as part of getting the RF design right.
Blyler: How will IoT platforms be positioned today to meet the changing hardware-software nature of the semiconductor supply chain? What needs to be different from the past?
King-Smith: IP platforms take care of the hard stuff so that designers can focus on product differentiation. Today, there is less of a need to understand CPUs and communication engines but more of a need to deal with a particular sensor subsystem or the like. Customers need to focus on their expertise, which is different from embedded systems of the past. With IoT design you have to start with right hardware complemented with the right software. And with every step of the way you need an ecosystem of partners. Our IP platforms enable that eco-system to flourish. We even see a number of IoT players that don’t understand anything about chips but what they do understand is the system and that’s why they want enabled.
This was Hussein’s message at this year’s (2015) Imagination Tech Summit in San Jose, CA, i.e, of the re-engineering of the supply chain for some vertical markets but not all of them. IoT is one of those markets, where the economics line up by using an IP model instead of just a custom chip model. This allows IoT developers to incorporate their design expertise, be it hardware or software – optimized for the hardware or further up the application stack. It’s all about achieving differentiation.
Blyler: You said that the new term for embedded is IoT. What did you mean?
King-Smith: Originally, I put that phrase into one of my team’s presentations because a lot of people are into IoT wearables. We actually separate them out of the IoT category as wearables are a very distinct area. But I don’t think they should be separated because everything is going to be connected. When we are talking about putting meaning into IoT, it is really what we used to call embedded. It’s just connected embedded. And so that’s why I said yes, IoT is the new embedded.
Blyler: What do you mean by fabric-based SoCs? Old-timers in the industry might remember fabric-based chip companies like Morphic, Mathstar, Ambric and others.
King-Smith: There has been lots of similar technologies over the years. Both coherent and non-coherent fabric technologies are maturing. You need these fabrics to make it easier to assemble heterogeneous SoCs.[Editor’s Note: Today, most server microprocessors integrate coherent interconnects for creating shared memory systems. Non-coherent I/O is used to connect the sever processor to external devices such as GPUs, networking, and storage. There is a move to develop similar coherent systems for non-server applications, like mobile.]
Consider a network on chip, which might start as a switching fabric and be built up from there depending upon the requirements. It could be everything from a glorified bus on up. But our attention is increasingly focused a true network-on-chip (NoC) in which various layers are both coherent and non-coherent. Coherency is really coming down from the server level and going mainstream into mobile over the next few years. In preparation, we’ve been doing a lot of work around coherent fabrics. That was one of the reasons why we’re a founder of the Heterogeneous System Architecture (HSA)Foundation, which is actually built on a coherent fabric providing effective APIs.[Editor’s Note: HSA allows developers to apply these hardware resources in today’s complex SoCs. The goal is to enable applications to run faster and at lower power across the range of computing platforms spanning mobile devices, desktops, high-performance computing (HPC) systems and servers.
Blyler: Thank you.
Originally post on Chipestimate.com “IP Insider”