Experts At The Table: Mobile Design Challenges

First of three parts: Power displaces performance as top priority; iPhone vs. Android; cloud-based video; the advantages of stacked die; defining the next killer app.


By Ed Sperling
Low-Power Engineering sat down to discuss the increasing challenges of designing for mobile devices with Qi Wang, technical marketing group director at Cadence; Cary Chin, director of technical marketing for low-power solutions at Synopsys; Bernard Murphy, CTO of Atrenta; and Dave Reed, senior director of marketing at SpringSoft. What follows are excerpts of that conversation.

LPE: As we move toward the next process nodes and stacked die, we have lots of issues ranging from physical effects to software and hardware convergence. Are we making progress solving these issues?
Wang: The reality is that engineers are smart enough to find solutions and workarounds. The mobile market has changed everything—hardware, software, design methodologies, low power and mixed signal.
Murphy: There are a lot of challenges, but a couple of the obvious vectors for any mobile device are whether you can fit into a power, performance and cost profile, and whether you can finish verifying it before it becomes irrelevant. That will decide a lot of what you’re going to use to build it and what methodologies you’re going to use. With things like Android, some of the software choices are more obvious than they used to be.
Reed: We don’t run out of challenges to solve because we don’t run out of things we like. Who thought they needed to be able to watch a TV show on their cell phone? I just got back from Asia, and over there if you don’t have something to stare at on the train then it’s a long ride. We can and have solved any problems that have come up, and we will continue to do that.
Chin: That’s exactly what engineers do. We solve problems. One of the things we learned in the last five years or so, though, is that some of these problems that make chips fly off the shelves aren’t solved by purely technology. The iPod is a good example. It wasn’t the first MP3 player available in the market. The iPhone wasn’t the first smart phone, and the iPad wasn’t the first tablet. But those devices were flying off the shelves even though they didn’t contain the latest technology. As engineers we have to expand our vision beyond just technology as we move into this space of mobile devices and consumerism and what people are looking for, as opposed to who has the fastest processor or more storage. That’s a fundamental change in the market.
Reed: I agree that performance is not the driver anymore. It’s super integration. Consumers want a smaller footprint and they want better power. One thing I’m unhappy with on my mobile device is how fast the battery runs out.
Chin: We haven’t crossed the threshold of acceptability with power yet. I have chargers everywhere because it’s too painful to think about having my phone die. We’re not high enough on the power scale to say we have enough yet. Playing music back today is largely solved with the amount of storage and processing. It no longer taxes performance or storage. Video is where we’re running into problems today.
Reed: Data communications are a problem.
Murphy: If you look at the whole Apple lineup, it’s not leading-edge implementations of anything. It’s a new way of looking at the user experience. But I see that as a disruptive jump, and I don’t see every other player in the market doing disruptive jumps. All of the Android platforms are essentially clones of the iPhone or iPad.

LPE: This is like business drafting?
Murphy: Yes, exactly. You get some business jumps out of the iterative company like Apple, but the majority of the market is not doing that.
Wang: I agree that other phones are clones, particularly on the hardware. But I do see differentiation in the software. With hardware platforms it’s going to go the way of foundries. There are not too many of them in the world. Eventually they will all have the same DNA. But the software will differentiate the features, the look and feel and the power consumption. Where Apple did a good job is packing new technologies inside a user interface. If you look at the Macbook Air, there’s a video of someone in Japan using it to chop fruit because it’s so sharp. But I’m also amazed by the graphics and the battery life. I still believe performance is driving the business. Unfortunately, those fancy technologies are crucial to the semiconductor industry but we don’t get recognition. People don’t care about a 3D IC. They care about being able to watch a video. It should be more application-driven as opposed to 20nm. That’s not the driving force anymore. It’s the application, the software and the platform—and all of it put together to create a unique value.

LPE: We say performance isn’t important, but no one wants to wait for downloads. It’s no longer just a device. It’s how it interfaces with everything else, right?
Chin: But even that is in the context of the human interface. The speed at which I can push the button is on the order of seconds before you get upset about it. That’s different than the billionths or trillionths of a second we’re working on down at the engineering level. As you get new applications coupled with different things phones can do, you have to think about how those relate to power usage and efficiency. That has to be above a certain threshold, and today pushing 3D IC as a manufacturing technology to allow lower power is a good goal. There are other ways of doing that, too.
Murphy: If you look at 3D movies on a cell phone, then stacking die is the solution. If you have to do that much processing, that puts a lot of emphasis on memory. You can solve that with a big external DRAM, but battery life will suffer. If you stack a DRAM on top of an interposer, you can get good performance and battery life.
Reed: When you reach a threshold then continuous improvements lose value. With interactive tools, you need enough performance where people perceive it’s instant. Once you reach that point then you don’t have to do better than that. With bandwidth we certainly haven’t reached the threshold. Processing power, maybe. On my home computer I used to look for an upgrade every couple of years. I don’t look for that anymore.
Murphy: But on your handheld device that’s different.
Reed: I prioritize battery life over processor speed.
Chin: And that’s the interesting part. In the generation we’ve grown up with in technology, there hasn’t been a competitor to performance. Faster was always considered better.

LPE: How about cost?
Chin: In the consumer space, $1,000 for a PC was always good enough, and $2,500 for a high-end one was acceptable. That’s why a one-inch IC was the maximum you could build for 30 years. Those things didn’t change for a long time, despite these technological advances. But power is the one thing that is changing. You’d rather have it last an extra hour than run faster because you rarely run a spreadsheet on your phone.
Murphy: You don’t want it to run slower, though, for everything you’ve already got.
Chin: Above that line it can run slower because what I’ve gained is more time for the stuff that’s critical. That will change when we hit the next killer app. If everyone needs on-the-fly video encoding and recognition, that will change things.
Murphy: That’s cloud-based video.
Chin: Yes, and it’s coming. But we’re at the point where we need to define the next killer app. We did a really good job over the years, between Intel and Microsoft, of creating new versions of processors and apps that required each other. Today the stuff I really care about is not plugged into the wall, so power is a big issue. That doesn’t mean we’re done looking at technology. The piece that goes hand in hand with mobile is the cloud infrastructure on the other side of the communications pipe.
Reed: Low power does not equate to low technology, though. If you see the extremes people have to go to in order to manage power down at the silicon level, it’s extreme.

LPE: So how do we get the power down? Will it be 2.5D and 3D stacking? Will it be better software?
Murphy: All of the above and more.
Wang: There is no silver bullet. At the system level you have to go beyond the IC. In a mobile device the screen sucks up the most power. That needs to be tackled first. So if you can’t solve that, you add solar to the screen so it powers itself. In software, people are starting to recognize that’s an issue and they’re worrying about it. They’re developing tools and methodologies to solve that. Then if you go down a level the challenge is the packaging problem, which is why we’re talking about 3D ICs and wide I/O. You have to work all the angles. You have to squeeze out every bit of it.