Productivity And The IoT

Experts at the table, part one: Chips have to get to work quickly and cheaply for the IoT to work. How do we get there?


The market for devices that connect almost everything to the Internet of Things is projected to explode, creating opportunities for companies that haven’t been traditional chip developers to decide to start developing devices. Semiconductor Engineering sat down to discuss this topic with Jack Guedj, corporate VP of Tensilica products at Cadence; Bill Neifert, CTO at Carbon; Marc Evans, director of customer solutions at CEVA; Drew Wingard, CTO at Sonics; and John Koeter, VP of marketing and AEs for IP and Systems at Synopsys. What follows are excerpts from that discussion.

SE: How do we realize the opportunity of IoT? How do we get there from a productivity point of view? Where does productivity come into play with the IoT?

Neifert: I think it has to be because you are looking at something like this that’s on a massive scale. It’s things. It’s inherently consumer in a lot of aspects and the second you talk consumer, you’re talking hundreds of millions or billions of these things. You do it wrong and you’ve made a rather large mistake that comes back and bites you later on. There are a bunch of things that come into this. It’s the normal problems that you have with all this stuff on a scale that needs to be turned around quickly because it’s consumer and needs to be turned around cheaply. All of the problems that you’re trying to solve for other leading edge SoCs, they may not be as complex but it doesn’t mean you can take a back seat on any of this stuff because of the massive scale and how quickly you’ve got to have it turned around. That to me is where the productivity comes in on this.

Wingard: To me, productivity is tightly tied to agility. I’ve been using this phrase, ‘premature integration’ but I think the world of IoT right now is the world of premature integration. The problem is that so many of the IoT applications require a form factor and battery life characteristic that can only be achieved, essentially, through full integration — and that’s unusual. In the history of our business, integration is usually done to try to reduce cost. Sometimes to improve performance but usually to reduce cost. In the case of IoT, integration is what enables the application and that would be great if only we knew what to build. The problem is we don’t know what to build so we don’t know what the compelling feature sets are, we don’t know what users are actually going to buy, so people are forced to make bets. There’s been a couple of presentations — Hossein Yassaie of Imagination has done the most stuff on it to say, ‘Don’t bring your smartphone application processor to IoT,’ the problem is those devices are so heavily overdesigned for this space that they don’t actually meet the power requirements. They can’t fit into the form factors, they can’t deliver the battery life so we’re stuck with this very difficult challenge of how to do it. So productivity, agility comes in by the fact of we need to be able to react to what the consumer will decide in an incredibly fast fashion in order to make IoT work.

Evans: I agree there’s definitely this huge power constraint especially as we look towards the wearable markets and very constrained battery profiles that aren’t improving capacity. There’s also some baseline ingredients. One is connectivity. The device has to connect to something, whether it’s using a local hub on the person or whether it’s more machine to machine type of connection. I think in some domains — in the wearables — we’ll see a lot of natural user interface given the screen profiles are becoming too small to do anything useful with. There’s going to be a lot of natural user interface, natural voice, natural language, gesturing sort of things — they are going to form some baseline ingredients in this space.

SE: How are we going to figure out what to build and are we going to be able to do it fast enough?

Wingard: My perspective is that this will end up being standard products in the semiconductor business but it can’t be at this point. It’s the domain of the systems companies. It has to be the domain of people who are going to be selling the end devices because only they can get the data quickly enough about what the users will accept. Only they will have the economic benefit of actually taking the risk of doing it. Most of the standard product semiconductor companies are saying, ‘This is what I have, maker community, go do something useful with this,’ because they really don’t know what to go build. So it will be the systems companies that will have to go out and take a stab. We’re going to see — at least for a short period of time in this space — the renaissance of the ASIC-style model.

Guedj: That’s what I was going to ask. ASIC or … so they have standard cells or full custom but they also have structured arrays that enable a much quicker time to market and also lower cost if you want to do thousands or tens of thousands of units. On the tool side, also to be able to enable the agility — it’s probably some changes from making ten million gate SoCs that you’re going to tape out once a year or once every two years versus something you tape out more frequently. Same on the board design; you might also have some things that enable you to be nimble and create those boards quickly. Then for IP cores that go into the SoC, you also want to be nimble so you can quickly reconfigure to be able to get to the sweet spot of what you are looking for.

Neifert: The interesting thing on IoT is, what’s it really going to be? It’s a great area where the marketing buzz has really preceded the real application. Personally, I’m wearing a Fit Bit here because I’m doing this for the ARM Challenge but prior to this I haven’t worn anything on my wrist since the 8th grade. They’re talking about watches as one of the killer apps and [I keep thinking] I don’t want to wear a watch. After this week, I’ll probably never wear it again. I’ve had it for two days, and I’m happy with any geeky object for three days — we’ll see how it lasts after a week. It’s the same for any of these IoT applications…What really is IoT? What will be successful because a lot of these answers really depend on where the IoT label will finally land with something that’s successful. I think there’s a lot of stuff out there that kind of is floating around in there, and it’s great, but the application people always bring up is your refrigerator will be able to order your milk for you. I don’t really want that to happen. I travel enough, my milk’s going to go bad no matter what I do with it.

Guedj: The question is, is productivity to create IoT subsystems or systems or is productivity for consumers like you’ve talked about. Which question are you asking? Because at the end of the day, there’s productivity, there’s use, that’s the big question: What is it?

SE: How do we prepare for something that we don’t even know what it is yet?

Wingard: Obviously it’s difficult to do. I think there’s a couple of sub-areas within the gigantic umbrella that seem like they make some sense. I think the healthcare-related ones are easy to point at. The moment where the wearable healthcare appliances is the moment when the insurance companies figure out they can save money by buying you one to check your glucose level or check your heart rate or to check your current cholesterol or blood pressure level. When they realize they will save money, then suddenly there will be a class of those things that will just happen naturally. I think there is another class that’s probably more in the machine to machine part of this, which clearly shows up in the next generation of factory automation and inventory control, and all of those things. We go from the RFID style of things, which are completely passive to something that’s much more active — I think those are almost a no-brainer once the silicon technology exists to make them cheap enough.

Koeter: You’re going to see an explosion of new applications driven by ubiquitous connectivity. I was on a breakfast panel and a guy was defining IoT in a way that I personally hate, which is ‘from sensor to server’ because there’s nothing unique — in my opinion anyway — in the datacenter that’s relative to IoT. When I think about it from an IP provider perspective, I think about IoT on the client side. Specifically, we’re looking at wearables and machine-to-machine applications right now and I think you’re going to see this amazing burst of creativity driven by very low power devices with connectivity built in. I’ll give you two examples I’ve heard of recently. One is a smart blanket, where you wrap a newborn baby in a blanket that will monitor the temperature and breathing patterns. What new parent wouldn’t want that? The other example, at the other end of the spectrum, is connected slippers…with built in accelerometers on it that are Bluetooth enabled and it is for the elderly. When the slip goes literal, that makes sense…It’s an indication of the tremendous degree of creativity that the basics of the process and the IP and the technology is going to enable.