Supply Chain Adjusts To Design At The System Level

By Ann Steffora Mutschler
System-level design is impacting the supply chain at many levels. Software suppliers, IP providers, semiconductor companies, system integrators and OEMs are challenged to work ever more closely together and find a new balance of power for who controls what in the content of an SoC.

“We see more and more the design chain driving how our tools work together,” Frank Schirrmeister, director of product marketing for system level solutions at Synopsys said. “The semiconductor house works with us on how to integrate the IP they use to make the chip development, and the representation of the chip to be used as a prototype, more efficient. They work with us to interact with their customers — the integrators and system houses – to take prototypes like virtual prototypes and send that to the system house to do software development and hardware/software integration on it, long before the chip is available.”

In the other direction coming down from the design chain, the system houses use tools from the system level to articulate requirements to their suppliers, namely the semiconductor companies and integrators, in a way that is much more crisp and precise and not open to interpretation. “Instead of pushing down a paper spec and having 10 meetings to identify what the spec means and talk about it they provide an executable prototype as a virtual prototype to their suppliers and tell them, ‘This is what I want, can you deliver on it?’ Then you have something much more precise to interact on within the design chain,” Schirrmeister said.

What this reflects is the notion of how the power structure within the design changes, which is becoming very application-specific. In wireless, a lot of vendors are pushing virtual platforms because they need to make sure Android, Windows Mobile 7, and the Apple iPhone OS all work on it when the chip comes back. In automotive, it’s much more about precision of the specification so it sometimes more the other way around, where the system house articulates downstream.

“That’s an interesting thing to watch as we basically go from IP to subsystems, from subsystems to more complex subsystems in semiconductor houses, and then to the system houses gaining more importance in specifying things,” he said.

Similarly, Mark Throndson, product marketing director for technologies at MIPS, also has seen a lot of change and an increasing demand for system-level solutions. “There is a software aspect, as well, which ties together everything. That’s a heavy amount of the investment MIPS has made, from our Android activities to work with Adobe, and a variety of major open source initiatives.”

The system implementations for new products, such as TVs that connect to the Internet, require a whole new suite of capabilities and services to enable that experience effectively, which is a more complicated development effort. “Our customers and our customer’s customers look to us to provide more of the solution than just licensing an IP core,” Throndson said.

“More of the solution” may include software that doesn’t necessarily just depend on the microprocessor IP. It may have some dependencies on what graphics support there is, the type of audio and video processing capabilities, security, specific interfaces and the like. He noted that MIPS leverages its relationships with complementary IP third parties to create platforms with software that can support the different pieces of IP.

Seeing is believing
From the perspective of network-on-chip supplier Arteris, the shift to system-level design is causing changes in the design chain represented by semiconductor companies finally starting to ask for models, noted Kurt Shuler, Arteris’ director of marketing. “Since I joined Arteris in August I have seen five prospects ask for SystemC TLM models of the interconnect, either at the cycle accurate or approximately timed levels of abstraction.”

The NoC IP configuration can be used to interface these models with the SoC system models.

Mentor Graphics also sees a lot more interaction happening between the designers and the rest of the supply chain. John Isaac, director of market development for Mentor’s Systems Design Division, said it is being driven by “increasing pressure on the OEMs to get the most competitive product to market faster. For example, starting at the very beginning of the design process the designer is communicating the parts that have been chosen for the product to procurement and manufacturing. This will enable them to identify any of those chosen parts that may affect their ability to get the product to target volume production on time, or components that may require special manufacturing and thus increase product cost and time,” he said.

“We see additional communication between the various disciplines in the product development process again during the entire design process rather than at the end. An example is between electrical and mechanical domains. The mechanical designers have a view of the entire product, including the enclosure, and they are in the best position to analyze if the power being generated by the ICs on the PCB will be dissipated properly or if junction temperatures will be too high and affect product reliability,” Isaac said.

Supply chain changes driven by rise of ODM model
With the very strong emergence of ODMs, the role of the OEM has pretty much disappeared said Giddy Intrater, vice president of product marketing at MIPS. “Rather than the OEM building the system, the ODM actually expects to get the complete system done by the semiconductor company so the new system companies are the semiconductor companies rather than what used to be the big OEMs of the world.”

He noted that OEMs such as Apple still do complete system design, but the bulk of consumer electronic products are built by ODMs that actually expect to get complete solutions from their chip vendors. “Now these chip vendors are turning back and actually wanting to reduce the amount of work that they need to do and go back to their IP suppliers and expect the IP suppliers to do as much of the system-level design for them,” Intrater said.

As a result of these shifts in the market, he said, MIPS’ customers supply the complete system to their customers, which means MIPS is responsible to supply the complete subsystems. “We typically supply the CPU subsystem, which includes the CPU and major pieces of software that go with it. In the past we would just offer a compiler and an operating system and our customers would write their code on top of it. Now we have to come with complete software systems that run on top of our solutions and Android or other software systems that we’re working on,” he continued.

Michal Siwinski, Cadence’s group director for product management and system realization, agrees a major shift is taking place with OEMs today. “The idea of a traditional OEM and the demarcation line between an OEM and a semiconductor company are shifting, and will continue to shift. I would argue that over time what semis have traditionally done is going to change pretty drastically, and a smaller number of semis in the classical sense will remain. Some will basically evolve much more toward being system providers and we see that already with a few companies that identify themselves as semis—they are basically behaving like OEMs already.”

What exactly will they become?

“I don’t think that most of the semis or the system houses today know where they will actually end up,” Siwinski said. “I think a lot of them are trying new things because the time pressures and the cost pressures are just prohibitive. Some are exiting businesses, some are taking bold, new steps—and all of them are looking for different, more productive, more efficient ways to get things done.”

What the shift to system-level design means
The industry has been talking about system-level design taking off for quite a number of years, noted Mike Gianfagna, vice president of marketing at Atrenta. “It’s finally happening, but not the way a lot of people envisioned it. Sure, high-level synthesis is becoming more of a mainstream tool, but that’s a small part of the story. The real story is about platform-based design, IP reuse and software differentiation. Advanced chip design is now all about finding the right collection of pre-existing blocks, stitching them together in the most efficient way, and adding differentiating software to the design in a predictable manner. It used to be about who came up with the best original circuit design. These days, too much original design is likely to spell long time to market and a loss of market share.”

These trends are creating new markets and new supply chain configurations. IP providers have to prove their product is robust, complete and integration ready. “EDA can help with that, creating a new market. Tools to help assemble the design are also a new area for EDA, and this one has a potentially very interesting twist. The semi suppliers are focusing more and more on providing complete platforms, software stack and all, as opposed to just a bare chip. Putting that platform together requires a lot of EDA assistance. Could EDA become a business enabler for the semi companies, and not just a cost of doing business? Maybe, and that would be a new model for the supply chain,” Gianfagna said. “If we take a look at what’s going on with 3D stacked die design, the supply chain will take on a new look, as well. Who is going to assemble a stack of silicon from multiple vendors? Someone has to take the yield and inventory risk. Who will that be? It could be a new breed of company.”