Experts At The Table: Platform-Based Design

Last of three parts: System design automation, complexity vs. business and consolidation.


By Ed Sperling

System-Level Design sat down with Simon Bloch, vice president and general manager of ESL/HDL Design and Synthesis at Mentor Graphics; Mike Gianfagna, vice president of marketing at Atrenta; and Jim Hogan, a private investor. What follows are excerpts of a lively, often contentious two-hour conversation.



SLD: Where does the consolidation happen in the chip design world?

Gianfagna: It looks like an upside down triangle. At the top are the systems guys. They will differentiate, if not in hardware then in software. The guys in the middle who are creating the silicon building blocks and building the platform—that’s where the consolidation will happen. Below them in the foundry area there will be even more consolidation. But as you go up that inverted pyramid, now there are more users for new cell phones, netbooks and strange devices we haven’t thought about. Maybe there’s hope that as you serve that top part of the inverted pyramid that EDA can provide enabling value to bring those platforms to a broader range of users. There’s new money to be made.

Hogan: I’ve been scratching my head for weeks about what does (Intel’s) Atom being made by TSMC mean. Microsoft is Intel’s biggest partner, so all those applications run on Intel or AMD. ARM runs Linux. So now that Intel goes to TSMC at 28nm and you have a bootable Microsoft address sitting on the Atom, do you use ARM or Atom? Which market do you go after? That gets interesting.


SLD: It’s a war, no doubt. Is that good?

Hogan: We’re going to build a model for Atom and a model for ARM Cortex12 and the DDR model, a simulator and a cockpit to manage all this. We want the battle to happen. Having a monolithic model is bad. TSMC logic is SoC. That’s been ARM alone up until now. Now it’s Intel and ARM SoCs. The world has gotten a lot more interesting.

Gianfagna: But tell me who is more responsive in listening to their customers and adding their input into designs, Intel or ARM? I’ll give you one guess.

Hogan: I agree with you. ARM has the ecosystem. But Intel doesn’t care about ARM. They only care about ARM because it’s an enabler for Qualcomm, Broadcom and Samsung. They build billions of units. What does Intel want to do? Build billions of units. This is all about ensuring they don’t lose the netbook market to ARM. The SoC is the way do it. That’s a huge opportunity for us.

Bloch: I think it commoditizes the chip implementation. It’s a block. It’s a very complex block, but it’s still a block.

Hogan: Let’s say it’s a hard block from Intel. We know that TSMC isn’t going to create that. We know Intel well enough to say that will never happen. So it gets laid out on Synopsys’ tools. Synopsys already has done a deal with Intel. There are no more incremental dollars for place and route because Intel has already paid for that. So the money that’s available to us is in the TLM space.

Bloch: Implementation is difficult, but it will be done. What’s important for managers is that these end products must be different from a competitor’s product. If the products have similar features, you compete on price. People are paying more and more attention to the system-level design. The level of worries is moving up to the system-level design rather than implementation.


SLD: So what do you need to make that happen?

Bloch: Architectural design, virtual prototyping of software, verifying the system fast and designing your new functionality quickly through high-level synthesis.


SLD: How much of a factor is time to market now? We hear a lot of stories about missing deadlines and market windows.

Hogan: The Chinese semiconductor companies don’t have back ends. They just have front ends.  What I think will happen is they’re going to go to eSilicon and have them do all of that stuff for them. They’re going to hit it 80% or 90% first-time silicon because they have the methodology to do that.

Bloch: These misses happen when companies do a napkin design instead of a system-level design. They say, ‘This is what we’re going to build, this is the functionality and this is the power consumption.’ Then they go and build it and it doesn’t do what they expect.

Gianfagna: But people don’t admit it, and you’ll never hear about it. Time-to-market pressures are alive and well. Most of the market is consumer-driven and the consumer market will always be fast-paced.

Hogan: That back-end semiconductor process is a fixed time. It will take you nine months from logic to tapeout, if you’re really good, and then 12 weeks for manufacturing. You can’t get around that. The fact that you do it right once and get to market, that’s great. If you don’t do it right, that’s where you’re going to burn through a lot of money.

Gianfagna: We can talk about the sophisticated things like IP. But there are fundamental issues. If you have a complex project and you do a poor job of planning, implementation is a mess.

Hogan: This is what Nokia is really good at. Their platform is two generations and they spend about two years getting there. They have a virtual platform that they do the software development on and they can hand it off to NXP. You’re ready to ship 20 million units in two years, and they do it time after time.


SLD: That’s compared to the old days when they used to do 20 platforms in 1 year.

Hogan: They can’t do that anymore. And the guys right behind them are Samsung.


SLD: Will we start seeing consolidation of virtual platforms?

Hogan: If you start with cell phones, there are about six major players. At least one will go out in the next year. And of those who remain, how many have the money to invent a new platform every two years? Nokia will have three phones—high end, midrange and low-end. So yes, there will be fewer platforms.


SLD: As business decisions increasingly drive technology, do we see less innovation?

Bloch: I don’t think so. People are less patient and more interested in faster cycles of innovation. Product cycles are shrinking.

Gianfagna: In a vibrant economy, there will be new markets for new applications. Below that, which includes the silicon version and building blocks, there will be consolidation.

Hogan: And if you have a really strong balance sheet you begin buying up companies, which is what Intel is doing.


SLD: Are we still talking about EDA companies, or are these now broader in focus?

Hogan: No, it’s now system design automation.

Gianfagna: There are new markets above the traditional EDA market. It will be healthy for some suppliers. Above that, there is opportunity. Does that become traditional EDA or is it vertical-market focused? I’m not sure.

Hogan: If I’m going to bet on one area, it’s going to be the mobile/wireless market.

Bloch: Computing, wireless, imaging and video are all converging.


SLD: Let’s go back here. Does the end market drive the technology or does the technology drive the market?

Gianfagna: And that will slow down innovation. If you think about whether technology is driving the market or the market is driving technology, I think it’s a hybrid. What if Atom had 1,000 embedded processors in it, would that drive cell phone use? Absolutely. But it isn’t one thing driving the other. There’s a symbiotic relationship. As long as both are advancing in lock step, we have a healthy business.

Bloch: It’s what you want versus what you need. The price points have gotten to the point where what you want may not be what you need. If design costs skyrocket, these prices will go up.

Hogan: Would you ever have dreamed a decade ago that there would be 10 million text messages an hour in the United States?