First of three parts: Future changes will be significant, both in technology and how we use it. Expect more regulation, more unusual design, and more integration with software.
Semiconductor Engineering sat down with Jim Hogan, long-time industry venture capitalist; Simon Bloch, senior director at Samsung Electronics; Sumit DasGupta, formerly Si2 senior vice president of engineering; and Mike Gianfagna, vice president of marketing at eSilicon (VP of corporate marketing at Atrenta when this roundtable was held). What follows are excerpts of that discussion.
SE: What’s going to change in electronics in the future?
Bloch: My world is surrounded by smart phones and tablets and TVs that are connected in a contextual way. Device-to-device communication is a big thing at Samsung because we are accelerating devices. At the same time, we’re dealing with basic technological issues that deal with power consumption and performance and how to make all things make better. So when you consider what’s coming, what are the horizontal and what are the vertical technologies. People are talking about automotive or smart phones. Those are verticals. Distributed compute systems are horizontal. In my opinion, there is a lot going on in both, and a combination of both. Computer today not only computes, but it also has a networking function and distributed systems that are located all over. Everything is changing in that horizontal area around what is the compute system. In the verticals, there is a lot of work in extracting valuable content from data. Some people call that analytics.
DasGupta: I’ve been following RTL with 3D architectures for several years. I attended a conference where I told them it isn’t ready yet, and they nearly drove me out. When is 3D going to be real? From an EDA perspective we are just about ready, but the products are not there. I remember three years ago saying this is not going to be ready for at least five years. I also make an incendiary statement that 2.5D is going to be here for a very long time. It’s time we delivered both of them. We’ve done 2.5D before. When I was at IBM, we had an interposer. It’s not unknown technology, but we need to make it ubiquitous. In my opinion those will be very real.
That’s one big change. A second change will be in the area of photonics. A large number of companies are looking at it for a number of applications. IBM, Oracle and Intel are looking at it for high-speed inter-processor communications. We got invited to an EU-funded project in Europe where we are the only non-European company. We’re going to add photonics components to OpenAccess to enable electronic photonic design. That’s very exciting, and it’s going to be a huge improvement to communication between processes, not just processors. Outside the chip we need huge bandwidth, and photonics has enormous potential. There’s a third thing that’s coming—automotive. The percentage of electronic components is going up dramatically. There is real potential there for autonomous vehicles. There is a huge potential for large numbers of electronics components in each vehicle. And these are all new applications.
Gianfagna: The Internet of Things is interesting. The whole interconnected world of devices is certainly an opportunity for more bandwidth and more silicon and more ubiquitous use of that technology. We’re going to see a lot more of that. Everything is going to have an IP address. Even today if I have a gadget with a wall switch, there’s probably a device that I can plug in so I can control it with my iPhone. The world is going that way, but there’s a more fundamental shift underway here. You don’t need to be constrained by form factors. At one point we were constrained by the terminal connected to the VAX connected to the mainframe. Now we’re constrained by iPhones and Androids. When transducer and networking technology get sophisticated enough, you can do things like wearable electronics. You won’t need a lot of stuff because it’s already here. You won’t need to carry a generator with you. The technology needs to evolve to have the ultra-cheap, ultra-reliable sensors and very small processing units and new networking sensors. It will take a while, but it fundamentally changes the whole dynamic. When the semiconductor industry starts building the equivalent of drywall and lightbulbs, that’s a really different industry.
Bloch: If you go that far into the future, it becomes the Internet of people. Wearable electronics will be on us, but it’s not clear where.
Hogan: I went to Japan last year for a private equity deal and they’re still not turning on the air conditioning. The escalators in those three-story deep train stations still aren’t powered up. They’re regulating brownouts because they don’t want to build another power plant, and they’re paying a lot of money to Iran for energy. That’s not sustainable. They also have to compete against China for energy. In the end, that’s going to be a battle they’re going to lose. We talk about carbon footprint. The real issue is energy—economic energy. The world will get to a point where you will have budget that’s assigned to you and you will comply with that. A TV calls 100 watts. If it pulls more than that, you don’t care. But it will be regulated. If you shut off your set-top box today, it draws 40 watts. Why? Because the disk is still moving. The challenge for us in the SoC world is to be able to shut off the disk and pull 1 watt. Maybe you have an LED that shows it’s in deep sleep, not standby mode. That becomes the central issue going forward, and it’s what changes the world as we know it—it’s the consumption and regulation of energy.
SE: How much of this is really new technology?
DasGupta: Let’s go back a few years. The microwave oven was invented for the space program. NASA funded it. So if you talk to an astronaut in the 1960s, that was current technology. We didn’t buy our first microwave oven until 1984. We needed something to heat the food quickly. The gestation period for it to become ubiquitous was long. Many of the developments happening today will not come to ubiquitous use for 20 or 30 years. That will likely be the case with autonomous vehicles. How much redundancy is required? What’s the business model? What is the legal model? If a child is killed, who pays for that? Is it the driver, the car manufacturer, who? The same goes for 3D. What is the legal model for different layers? Who pays for a flawed chip? Known good die may not be good once they’re put together.
Gianfagna: You have assembly issues and then the finger pointing after that.
DasGupta: Exactly. And that is the bigger hurdle.
Bloch: Everything we do has hurdles. Flexible displays exist already. The obstacle to putting those into production are making everything else behind the display flexible, too. That means the boards, the chips, the actives and passives. With phones, there was the brick phone, the simple phone and the smart phone. The smart phone is about seven years old. TVs today are still in the generation of the brick phone. There is no evolution other than better resolution. Even smart TVs are just connected to WiFi. There is an expectation that some revolution will come with a wall computer. The biggest difficulty there is the user interface—how you connect the analog wall with pointing technology. Another area that will change is software-defined everything. A lot of these after-market companies develop hardware for an automobile that two or three years down the road will be obsolete. Some companies are moving to software-defined hardware. We need to expand that to software-defined devices.
Gianfagna: Today the arrow points from hardware to software. You build the hardware and turn it over to the software guys. What if the arrow pointed the other way—here’s my application, build me a piece of hardware that optimally uses it. Here’s what we want to create for the user experience. Today people write embedded software code on some operating system and that’s the user experience. What that runs on has a profound impact on the efficiency and the robustness of that experience. What if the software guy was king and this is the executable spec for the hardware. This would be EDA prime or double prime. The cool thing about that world is that you’re hitting the bull’s-eye on what’s squarely needed, and all the software guys will be buying EDA tools.