All Things To All Customers

As FPGAs push past traditional boundaries and into new markets such as mobile computing, power management is a bigger concern than ever.

popularity

By Ann Steffora Mutschler
Low-Power High-Performance Engineering recently spoke with Suresh Menon, VP of systems development at Lattice Semiconductor, about the challenges of directing the development of power-sensitive FPGAs from architectural decisions to identifying the target applications. What follows are excerpts of that discussion.

LPHP: When you look at the products that Lattice is working on today how do you direct the engineering teams in terms of the low-power aspects of the design?
Menon: First and foremost Lattice is focused on what I would consider low-density and ultra-low-density marketplaces. In those two areas, power is a very important factor in the requirements. Typically in the FPGA marketplace you achieve power by the scaling and technology, and pushing down the technology scaling, but that only goes so far. In recent generations, particularly since about 90nm, what you get when you scale down does not meet your requirements so you have to find other ways to achieve it. What happens essentially is, number one, you have to understand first and foremost what customers need. To me that’s actually not as simple as that because in an ASIC and ASSP space when you’re designing fixed function devices, it’s easy to figure out what customers need because it’s a specific thing you are doing.

For an FPGA typically we are in several market marketplaces—about three being the most important, and two or three more being secondary marketplaces—so you are addressing quite a number of markets there. The challenge is how to figure out what the key applications you want to go after are and deliver, and then focus your specs in those areas and what you can do there. It comes down to several things. One is to deliver power reduction you have to look at it holistically. You have the circuit side tied to the tools and what you need in the tools, and then also what you need in the IP, and you have to tie all that together in order to deliver a solution.

LPHP: So you have to have very highly skilled engineering teams and engineering management to guide those teams and make sure that they are doing everything possible to manage the power?
Menon: Correct. And you have to also understand the applications to do that. I mean applications as in all the different applications. You have to pick the set you are primarily going after. You have the mobile market, you have the consumer space, you have the wireless space and you have to look at the different ones and decide what the key things that you’re going to go after. We can’t solve everybody’s problem, but we want to do enough to get that 80% because that’s what you really need to do. But you can’t solve everything so you have to be selective.

LPHP: Once you have the application set and you know what marketplace you are going after, from an architectural perspective how do you make the tradeoffs to make sure that the design is optimal from a power perspective?
Menon: The first step is understanding the use model. Figure out what pieces and changes you need to make in the hardware and then what support you need in the tools to support the changes in the hardware—and what key IPs you need to deliver that customers need in those marketplaces.

LPHP: How much IP does Lattice use that is either left over from a previous design or acquired from third parties?
Menon: From the basic building blocks of the FPGA—and it varies from block to block—you can leverage anywhere from about 50% to 80% from previous generations. It just depends on which block and which IP block.

For the basic building blocks [you can leverage] probably a higher percentage—IP blocks like the high-speed serial. You leverage less as you go up in performance. But all-in-all, bear in mind that the challenge in an FPGA is when you are developing new IP. It’s not the same as the ASIC and ASSP world because you can’t just leverage what a third-party uses directly. You have to take what they have and modify it to provide the flexibility from the programmability standpoint that customers want. We always have to bear in mind that an FPGA is all things to all customers because they think they can do just about anything with it, so we have to be very careful with how we handle it.

LPHP: Are there specific challenges in regard to power when you are integrating IP from third parties?
Menon: There are challenges in achieving power within an FPGA because you start without having a specific application in mind. There is a general fabric, so you have a challenge in power. And then when you take a third-party IP, you have the challenge in how you achieve power in there and provide the flexibility at the same time. That flexibility gets in the way of delivering power reduction at the same time.

LPHP: How do you account for that?
Menon: You essentially harden what most of your customers would use, and then you provide the flexibility only in the areas where customers want to create their secret sauce because it is different for different customers.

LPHP: What is the most exciting development that we are seeing right now in the FPGA world?
Menon: The exciting thing here would be that we are entering the mobile marketplace and that’s a very exciting thing for FPGAs because FPGAs have largely been in the high-end comms market spaces and have taken a long time to move down into the low-density area. We are starting to see that across the board. In the case of Lattice we’ve penetrated into the mobile marketplace and I think that’s an exciting area because that’s where a lot of the growth in the next 10 to 15 years will be and this gives us a huge opportunity to find the same sort of applications that originally the high end FPGAs did in the communications marketplace.

LPHP: Why have FPGAs been able to move into the mobile space?
Menon: A large part of the reason is that the mobile space is now being challenged the same way that the high end used be challenged, which is now integrating a lot of functionality in a very, very short space of time. Therefore, there are no off-the-shelf chips that can solve your problem in that short space of time. So the FPGA has now become the tool to integrate these things …and have been able to penetrate that space.

LPHP: Looking ahead, how do you see new opportunities evolving for FPGAs beyond mobile?
Menon: Beyond mobile, it becomes the Internet of Things. What the mobile market space, or if you like the cellular/wireless area, is producing now is all these little gadgets that go into all the things we live with and depend on. That will help grow the FPGA space because as such you start to leverage a wider market area that’s growing….The million-dollar thing in my mind comes back to, people want to differentiate themselves from their competition.

For example, if you are building a cell phone—say, a higher high-end cell phone particularly—and you are building it with chips from Qualcomm and ARM, every cell phone maker can get those chips but they don’t differentiate them from each other. The FPGA allows you to add your little piece of differentiation and give you that value that customers will pay for.

LPHP: What is the design cycle like today for an FPGA going into the mobile space compared to a comparable approach in an SoC?
Menon: From a customer standpoint I believe that you can easily turn around a design in a quarter, and maybe in a couple of quarters you can prototype it and have something going. It’s a fairly short space of time, whereas if you have to do an ASIC or an ASSP that’s easily a year to 15 months.

LPHP: How is the price differential today? There’s always been the argument against FPGAs because of the cost.
Menon: Ultimately it becomes what value the FPGA provides. Typically in any mobile space it’s $1 to $1.50. What can you provide in $1 to $1.50? As the FPGA follows the technology shrink we are able to provide more and more value in that $1 to $1.50 for the customer.