SiliconBlue: Closing the Application Processor Gap
Date: Wednesday , December 31, 2008
There are two reasons as to why a startup, in this case SiliconBlue Technologies, can manage to win over 80 customers in a span of five months: first, it has a disruptive technology that prospective clients had been scouting for a long time; second, it has an expansive sales network that enables it to tap a huge number of customers on a daily basis.
But neither of these reasons applies in its entirety to SiliconBlue Technology, pioneer of an Ultra-Low-Power 65-nm FPGA (Field Programmable Gate Array) technology for consumer applications, says Kapil Shankar, CEO and founder of the company.
“Rather,” says Shankar, “we focus on giving solutions that clients are looking for. In that sense, we allow them to play on their own terms.”
Consider ‘on their own terms’ and you could be looking at 80 different sets of terms for 80 different customers. For a startup to build that range of solutions might have taken a long time, but not for SiliconBlue, founded in late 2005 operating in the immensely dynamic consumer applications market. What, then, lies at the core of its exploits?
On Customers’ Terms
The answer lies in the fact that the company builds FPGA devices that are used by customers to design ‘what they want’. For instance, Fusao Saito of Tomen Electronics says, “We have designed the iCE65 device (using SiliconBlue’s 65 nanometer FPGA product line) into an application for our leading-edge embedded solution - something that could not have been done with a traditional FPGA.”
SiliconBlue has built from the ground up the first single-chip, reconfigurable, ultra-low power FPGA optimized for handheld, mobile applications. Next, SiliconBlue created the largest FPGA IP offering targeting smartphones, PMPs (portable media players), DSCs (digital still cameras), MIDs (mobile Internet devices), PNDs (personal navigation devices), handheld POS (point of sale) systems, medical instruments, Edu-toys, and flash camcorders among others. Broadly, the company focuses on two themes, as Shankar calls them: the handset market and the non-handset consumer applications market.
The market that the company addresses is huge. According to Semico Research, handheld applications are a rapidly growing market for programmable logic as sales in these applications are forecast to be over $650 million by 2010.
SiliconBlue also offers programmable solutions for ASSPs. Given the fact that application processors and peripherals evolve independently, a gap is created by rapidly changing customer needs: Whereas application processors are relatively constant, (adoption 2-3 years), new peripherals and their combinations are needed more often in less than a year’s time.
If one has to mount new functions on the proven application processor, it would require lots of time and money. This would, in turn, delay the customers’ time to market. SiliconBlue’s ultra-low power low-cost FPGA plays to the customers’ terms in this too. All an ASSP vendor would need to do is to stack the application processor with SiliconBlue’s FPGA; this will help reduce cost, risk, and time to market by many factors.
The question that arises, especially in relation to handsets, is why would a startup design programmable logic into this rapidly evolving, dynamic, and cost-sensitive product when in emerging markets like India and China the product lifecycle ranges just six to eight months? The question becomes more pointed when one considers the power-hungriness of erstwhile FPGAs; they were known to reduce battery standby time by almost half. The answer to the first part of the query, says Denny Steele, Director, Marketing and Applications, SiliconBlue, lies in the handset designer’s environment and challenges. “Handsets have experienced extreme levels of convergence. It is a phone, a PDA, a camera, an MP3 player, a GPS unit, an email conduit, and an Internet browser, all at the same time. The process of merging all of these functions, into products that can work within the disparate phone standards of multiple countries, is full of ambiguity.” All handset designers rely on ASSP chipsets for building their products. “They build their design around the least number of components, write the necessary software, encapsulate it in plastic, add the logo, and ship it,” says Steele. However, these off-the-shelf chipsets are readily available to multiple competing companies and if everyone uses the same chipsets, all products would begin to look alike.
Given the fact that price cannot continue to be a long-term differentiator, handset designers, many of whom are SiliconBlue’s customers, have realized that programmable logic can be used to customize a standard chipset. For instance, in China, some low-end phones have blinking LEDs that function like equalizers while playing music. It is here that SiliconBlue’s devices come in handy, in that they help factor in the added functionality of the LED in to the basic chip.
“Some might argue that the same could have been done by building an ASIC,” says Steele. “It might have been feasible if,” he says, “the product lifecycle of devices had not been only 18 months, and at times even less.
“A further challenge for ASIC development is that the features may not be well defined; for instance, interfacing two devices not meant to work together. This creates a situation that is not conducive to effective ASIC development,” says Steele. These make FPGAs a viable alternative.
Power of 3 Lows
SiliconBlue’s ground up design for ultra low power answers the power challenge. But there are a couple of other issues one need ask of SiliconBlue, especially given the fact that it addresses the largely dynamic, price-sensitive consumer applications market. What about SiliconBlue's cost of its FPGAs and the space they occupy?
In fact, as Rich Wawrzyniak, a senior analyst at Semico, says, “Power and cost have been issues from very early on in the programmable logic industry. Despite the efforts to improve in these areas, traditional FPGA solutions, for the most part, do not meet the handheld application’s requirements for power or cost.”
It is here that SiliconBlue’s technology comes up trumps, in the form of the power of ‘three lows’. Shankar, the CEO, is certain that his FPGAs will be, if not are, the ‘lowest-power consuming, lowest-priced, and lowest space-consuming’ in the market.
His bullishness about his company’s products is endorsed by industry analysts.
Says Wawrzyniak, “SiliconBlue’s technology positions them to deliver an unprecedented combination of power, price, and performance to designers that want to use programmable logic but have not because power consumption was too high or performance too low.”
Shankar, well aware of this when he started his company in late 2005, says, “We began with the aim of building the lowest-power FPGA.” And along with his team he created an NVCM (Non-Volatile Configuration Memory) technology based on the leading edge 65nm process node. “Since new process nodes with lower capacitance and voltage consume roughly one-third to one-fifth of the power of older process nodes, the power consumption was straightaway cut. Using this process node, we built a brand new FPGA fabric,” explains Shankar.
On the way, they sacrificed some speed for low power. “We kept in mind that we’re not going to power a server, but only a small handset; speed needed to be optimized to that end,” he recalls.
Whereas SiliconBlue’s devices can run at 150 MHz, most handsets only use 25-50 MHz. That leaves a lot of leeway for future requirements.
Also, its non-volatile and reconfigurable FPGAs feature: operating power as low as 5 uW (typical); logic cell gate complexity unparalleled in existing low power PLDs; and sizes as small as 3mm by 4mm.
Competition comes from the traditional CPLD players as well as incumbents in the FPGA space, according to Shankar. “They are making efforts, but at best, those are all partial,” says Shankar. Most of them concentrate on bringing embedded flash technology to a single-chip solution to achieve what SiliconBlue does. This, says Shankar, is tantamount to going back two to three generations. The most advanced technology used in such cases is on 0.13 micron, which at worst goes up to 0.18 or 0.25 micron. On the other hand, SiliconBlue’s products, based on the 65 nanometer technology, are much better prepared to address the future. Also, the company has tested its solutions on 40 nanometer, and there have been no major problems.
“Most of our customers were using CPLD technology earlier. They were very tiny solutions and we offered them a much better solution in terms of functionality,” claims Shankar.
“In crunch situations, our ability to touch upon customer requirements and deliver soft IP comes in handy,” he notes, claiming that his company has the broadest IP portfolio in the market, three times as many as compared to the nearest competitor. Also, the fact that SiliconBlue offers turnkey as well as custom designed solutions keeps it ahead of competition.
Power of Partnerships
SiliconBlue has struck several partnerships in its three years of existence. The most recent, and formally announced, among them was the technical partnership with Magma Design Automation. The agreement between the companies facilitates collaboration and co-development “to provide a comprehensive FPGA design environment for SiliconBlue’s iCE65 FPGA product line.”
Magma’s tools have now been fully integrated into SiliconBlue’s iCEcube tool flow.
“Our partnership with Magma gives us and our customers access to proven, state-of-the-art logic and physical synthesis tools,” says Shankar, and he goes on to add, “Tools are what a client sees first. In a sense, they are our face to customers.” Shankar is certain that the partnership, announced in November, will help push the adoption rate of SiliconBlue solutions.
The company’s other formally announced partnerships are on the manufacturing side. SiliconBlue has tied up with TSMC and ASE in Taiwan for wafer foundry services and assembly. The manufacturing pipeline is fast, the current product lead-time being around 12 weeks ARO. In addition, the process is scalable, and capacity is expected to exceed 10Mu/month by the second half of 2009.
In addition to these partnerships, SiliconBlue has partnered closely with various vendors on a number of designs. In fact, as Shankar informs, the company has drawn up specific roadmaps with a few customers for whom it has done three to four products as well.
Like any entrepreneur, Shankar went through the proverbial ups and downs at the start of his venture. He had spent about 11 years at Xilinx prior to starting SiliconBlue, eventually parting ways to focus on the “stacked market”.
In low-power FPGAs, he saw an immense market opportunity and was backed by BlueRun Ventures and Crosslink Capital, which pooled $16 million into the company in August 2006.
Initially, when he went to prospective clients and started talking about stacking solutions, what they had in mind were large expansive and expensive products. “They were shocked when I told them that it was small and would cost $1-2, at times even less.” Scouting around for better features and technology, he found his eureka moment with NVCM. Today, SiliconBlue offers on-chip NVCM for free. This removes the need for separate configuration using Flash. NVCM technology also enables silicon products to have the best data retention and security.
The company — which has 40 patents under its belt, some of which are still being filed — received its second round of funding ($24 million) this October. The round was led by NEA (New Enterprise Associates), with the previous investors BlueRun Ventures and Crosslink Capital also participating. Speaking at the announcement of the funding, Brooke Seawell, venture partner at NEA said, “SiliconBlue has successfully introduced the first new FPGA technology in ten years, which is aimed at rapidly growing, battery powered handheld applications. We are impressed by the effective execution of an incredibly skilled SiliconBlue team of PLD experts. As the world goes ‘green’, this battery powered market is expected to exceed $650 million by 2010.”
SiliconBlue plans to use the funding to accelerate the next phase of the company’s growth by expanding sales operations globally to support new customers. It already has a sizeable channel in Japan and Taiwan, and is building a sales team in China and Korea. In addition, John Birkner, VP - Strategic Marketing, says that going ahead, SiliconBlue will enhance its presence in the North American market where many smartphones are designed, and Europe, where additional smartphone vendors are located. All this, in a bid to move closer to prospective customers.
“Our vision is crisp, but is fraught with challenges,” says a cautious Shankar. The non-plugged market, and medical devices and instrumentation market are areas where the company must increase its presence, he says.
Teamwork, the leitmotif of the organization, will help carry through the technological advantage. The company has around 75-80 people on board across the world. Given the recent round of funding, the global economic downturn is not expected to hit the company much. In fact, as Shankar points out, the company has enough runway for the next two years.
Design and related issues intensify and undergo a lot of disruption in times of downturn, because existing products based on existing designs don’t sell. The fact that SiliconBlue’s requirement for revenue for the next couple of years is low will help it concentrate on the design aspect and take the lead in newer platforms. Once the downturn reverses, FPGAs could be the gateway to dominance.