A good discussion raises tough questions, and that was definitely the case at a "fireside chat" at the ARM Technology Conference (ARM Techcon) Nov. 9. The chat was a lively discussion between John Bruggeman (below, right), Cadence chief marketing officer, and Simon Segars (left), general manager of the Physical IP Group at ARM.
Since my colleague Steve Leibson already posted a great writeup on the 45-minute discussion, I'll focus on a single question that I found particularly vexing. It's a complex question, but it basically boils down to this - in a world gone crazy about software applications, how do hardware developers differentiate their hardware? Are we heading for an era in which all the innovation and differentiation will be in software that runs on "least common denominator" hardware?
You may recall that the EDA360 vision is about the creation of hardware/software platforms optimized for applications deployment. But the idea is that both the hardware and software will be optimized, and the software will have enough visibility and control to exploit the unique features of the hardware. Software that blindly runs on top of undifferentiated commodity hardware is not what this vision is about, and it will not lead to revitalized EDA and semiconductor industries.
It's Not Your Teenager's Fault
To grapple with the question posed above, we need to first look at the explosion of software applications or "apps." Many presentations have talked about how teenagers are driving the applications market. Not so, Bruggeman said - it's not really driven by end users at all. It's driven by suppliers who have found a new revenue model.
Suppose you want to buy a plasma flat-panel TV like the one used to simulate a fireplace during the fireside chat. You pay for it once, walk out the store, and you don't come back until it's time to buy a new one, probably many years later. The new applications-driven model, in contrast, gives electronics OEMs a constant revenue stream. Most iPhone users, for example, continue to purchase apps after they buy the phone.
"These new business models are driving complexity, which is rolling down to everybody in this room," Bruggeman said. "Device manufacturers are moving quickly to embrace this new model." Segars commented that the advent of downloadable software has driven the need for more and more IC performance. "As much as we can deliver, someone will figure out what to do with it," he said.
The Android Example: A "Massive Risk"
So far, the advent of "apps" sounds like a win for everybody. But there's a dark side, a "massive risk," as Bruggeman put it, and the Android platform provides an example. (Keep in mind that this is not a specific critique of Android, but a warning in which Android serves as one illustration).
Bruggeman noted that Android provides a software platform for applications that have not been conceived or developed, yet Android has to provide predictability. The APIs that touch the application are thus fixed. "Android can't take advantage of hardware differentiation unless it is brutally hard coded," Bruggeman said. "So it plays to the least common denominator and says, ‘we'll write to that and make the software really spectacular.'"
"So the semiconductor company sits there and says, how can I highlight the differentiation of what I do if the software plays to the least common denominator? How do we, in the semiconductor industry, in the systems industry, differentiate ourselves?" Bruggeman asked.
"Least common denominator hardware is not what we want to achieve," Segars said. "We want to make sure we can have innovation going on in the hardware, and not kill that by the cost of building software."
The Answer Requires Collaboration
So how do we foster hardware innovation and differentiation? No one company can accomplish this alone - it will take close collaboration among IP providers, foundries, EDA vendors, and customers, Bruggeman said. One example of such collaboration, he noted, is the work undertaken by ARM, Cadence, and Texas Instruments to develop an EDA reference flow in time for the first Cortex-A15 implementation (I wrote a recent blog post about this initiative).
Presenting another example of collaboration, Bruggeman identified Linaro as one of the most important new trends in the electronics industry. This open-source Linux initiative is backed by ARM, IBM, Freescale, Samsung, STMicroelectronics, and Texas Instruments. "If we can all work together to support the intentions of Linaro, it gives us the opportunity to make semiconductor performance great, make power low, make differentiation clear, and allow an Android or Linux application to work really smoothly," he said.
Segars explained Linaro this way: "Lots of people were optimizing Linux kernels for ARM, exactly the same work was being done in lots of different companies, and it was a complete waste of time." He noted that "Linaro is a great example of where the industry has come together to solve a problem."
During the fireside chat, both Bruggeman and Segars talked about the "inefficiency" in the supply chain that results when companies duplicate their efforts. "If we can stop doing duplicate work and start collaborating on a whole different plane, we can strip some of that efficiency out," Bruggeman said.
At the conclusion of the chat, Bruggeman lamented the reluctance of EDA vendors to come together to solve common problems. The focus, he said, should be on industry problems such as cutting the costs of SoC development and shortening chip development schedules. "Cadence can't do this alone," he said. "It will take the cooperation of an ecosystem, the cooperation of fierce competitors, and the cooperation of our customers working with us."
Bruggeman's concluding pitch to the audience: "I hope you're hearing there's a new candor and a new openness to work differently than we have in the past. We need you to push us to continue the forward momentum that we have."
Photo by Joe Hupcey III