Warren Savage: Grand Challenges in IP

This is the third in a four-part series showcasing Grand Challenges in IP. The first two conversations were with Sonics CEO Grant Pierce, and CAST Board Chair Hal Barbour.

This week’s dialog is with Warren Savage, founder and CEO of IPextreme. The company was purchased by Silvaco in 2016, where Savage now serves as GM of the IP division.

Warren Savage has been an energetic leader in the IP community, working to get companies in the industry to link arms, address common concerns, and give greater visibility to the importance of their products in the global semiconductor supply chain.

When we spoke on May 24th, Savage began by addressing my question about protection for IP blocks as they move down the manufacturing chain

************************
Ecosystem for IP Protection & Standards

“There are a number of companies working on protection,” he said, “but I am skeptical.

“Not so much from the technology point of view, but because it requires close cooperation with the EDA industry. These are issues where all three of the major EDA players need to cooperate to make something possible in the technology.”

“The problem with the protection stuff,” Savage added, “it’s difficult to ascribe a monetary value to it in order to make it a worthwhile [target] for the EDA industry.”

I referenced my two earlier conversations about Grand Challenges in IP, published in the last several weeks. In those dialogs, Grant Pierce focused on the technical challenges in IP and Hal Barbour on the business challenges.

Savage responded: “I agree that there are technical aspects, as Grant said, and business aspects, as Hal said.

“But the real challenge in IP is in the ecosystem. That requires something like a dance between the buyers and suppliers to keep the ecosystem in balance, so one company doesn’t dominate the industry.

“For instance, with the SystemVerilog impact on traditional flows like verification, the EDA companies can now see how to make money out of investing through cooperation – something that is particularly important with the consolidation going on in EDA today.

“It’s not like the 1990’s, where there was a mix and match of tools. Today you’ve got a Gotham-like situation [where the big players totally dominate] and customers get locked into one company’s individual flow.”

So there’s not enough competition between the major players to drive innovation?

Savage agreed, but was still optimistic: “As soon as someone figures that out, that lack of competition, there will be important new players [developing] something that no one can actually predict at this point. Something that no one saw coming – one company or one technologist that comes up with an innovation that tilts the playing field towards change.”

Do young technologists wanting to break into the industry need patience to wait out the cartel currently running the EDA industry?

Savage refuted the waiting game theory: “Actually, what’s really unusual about the semiconductor industry is the pace of change. It’s very fast, and can be unexpected.

“And that’s one reason standards are so difficult to achieve in some areas. The pace of change is so fast, by the time a standard works its way through the relevant [organizations] it’s a lot less relevant, if relevant at all.”

************************
Customization Demands

My own list of Grand Challenges in IP includes a way to guarantee that a block of IP will function as predicted when introduced into a new design, but Savage disagreed.

“That’s an old-fashioned way of looking at things,” he said. “Today, it’s the market that will tell us when that happens, when something doesn’t function [as promised] when used in a design.

“Basically, the most successful IP companies have worked to design once and sell many times. Starting in 1995 or so, the big push was to create something that could be sold many times, to get the right amount of profit off of your original IP block design investment. But that is all changing.

“Today I’m seeing more interest from customers in buying things that are a bit differentiated or customized. We get requests from customers to customize a piece of IP, so they don’t have the same thing [in their designs] that their competition has, or might have a license to use.

“Now up to 80-percent of the chip is internally developed IP or third-party IP, but if everyone is using the same IP there isn’t enough differentiation between the designs.

“Everybody does not want to have the same IP. Today the IP suppliers are under pressure to make special changes in their [blocks] to satisfy the customers’ demands.”

************************
IP as Product Business or Consultancy

Does all of this customization mean IP suppliers are morphing into design consultants more than product vendors?

“There are technically sound off-the-shelf products,” Savage said. “For instance, the UK company enSilica has an IP base, but they do a lot of services on top of that.

“And that’s especially true for some of the vertical standards like audio and power. These are specialized technologies that [require a lot of services].

“The industry is just at the beginning of these changes, so it is something that we want to watch closely.”

************************
Yield Targets & Moore’s Law

How do you know when a block can be generalized and made commercially viable as third-party IP?

“There’s the glib answer,” Savage chuckled. “If there’s a customer, there’s a market.

“But at IPextreme, now part of Silvaco, we talk to the big semiconductor companies about the IP they are using internally.

“If we find certain blocks that could be designed once and reused many times – with only small adjustments between projects – these are the yield targets [the company] can take one notch higher for use and sale outside of the company.

“The poof point is always: Has anyone used it twice?”

“There are a whole bunch of technical things about the design of an IP block,” Savage added, “and a focus on how it is architected.

“Was it customized for one particular chip? More importantly, is it possible to recover the big up-front design costs, recouping them over the next 5 years?

“Reusable blocks today are artifacts of design, and are following Moore’s Law. We are seeing a 10x increase in the size of IP blocks every 6 years or so, with the average size of blocks going up because conglomerates of IP are being put together as sub-systems.

“Today chip designers are becoming system integrators as the size of the blocks are increasing, the average complexity and gate count of the blocks moving up [with Moore’s Law].

************************
Grand Challenges to Worry & Inspire

Beyond issued mentioned so far – IP ecosystem, security and standards, increased demand for customization, knowing when to commercialize a reusable block – are there other Grand Challenges that keep IP vendors awake at night?

Savage chuckled and said, “I’m closer to Hal Barbour on that question. The business of IP is difficult, and [made more so] by consolidation.

“What I’m seeing happening now in IP, and we have already seen it in EDA, is the large semiconductor companies – instead of selling on a technical basis – are going in and carving out a piece of the customer’s design budget.

“With the big players able to command a bigger and bigger piece of that budget, it just leaves bread crumbs for the smaller guys in the industry. The small companies are then left picking up whatever remains [in the budget for IP], which hurts innovation because the smaller companies cannot then invest as much in R&D.”

“The bigger consolidated IP companies are also not investing as much in R&D,” Savage noted, “because they need to return more of their company profits to the shareholders.

“These companies have less bandwidth for innovation – have less need, for instance, to do a new USB because there’s already an established market there and no one is going to take a risk.

“Instead they are trying to grab things like the 10-nanometer finFET market – things they know are coming – rather than investing in NRAM or novel RF technologies.”

Despite these worrisome trends in the industry, Savage ended our conversation on an optimistic note: “I’ve been somewhat heartened over the last 20 years that I’ve been doing this, because there are more IP companies starting everyday than you can shake a stick at.

“In the old days, you used to see design teams spin out of the big companies to create EDA companies, but that is no longer the case.

“Now you see innovators starting IP companies. Despite the headwinds they face, people are excited to be coming into this industry.”

************************

Tags: CAST IP, Grant Pierce, Hal Barbour, IPextreme, Moore's Law, Silvaco, Sonics, Warren Savage

This entry was posted on Thursday, June 1st, 2017 at 2:57 pm. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.