Why would you spend tens of thousands of dollars on commercial electronic design automation (EDA) tools for your design team when they can get a decent suite of tools almost for free from their FPGA vendor? This question is probably asked daily in design teams around the world. As you might expect, it also comes up every now and then in the strategic planning rooms at EDA companies. The EDA company version of the question has a slight twist, however: “Should we invest millions to develop FPGA tools if FPGA vendors may be giving away similar tools almost for free before we have a chance to recover our development costs?” For many EDA companies, the answer is a resounding “no.”
A few, however, don’t like the implications of that answer. Around the world, electronics companies are shifting development investments away from ASIC technologies to alternatives such as FPGA and structured ASIC. As the money moves away from ASIC, the traditional ASIC-based EDA ecosystem is shrinking. With the other EDA segments relatively flat, the EDA industry needs to find a new market for growth over the next decade. Despite the apparent iron grip of the FPGA vendors, many in EDA feel that FPGA and structured ASIC are the two areas most likely to experience explosive growth in design tools.
If FPGA is the most likely high-growth tool market and is already dominated by the FPGA vendors themselves, where does commercial EDA stake its claim? How do companies trying to make their living from tool revenue alone compete with FPGA vendors who can afford to give tools away almost for free in order to lock in lucrative silicon sales? Obviously commercial EDA has to find a way to justify the additional cost of their tools when compared with the silicon-subsidized alternatives.
It turns out that each EDA company has crafted their own answer to this conundrum. Some, like synthesis leader Synplicity, stake their bet on running faster than the competition. “We are quite used to competing with tools from the FPGA vendors as they have been providing them to customers at little or no cost either through EDA vendor OEMs or directly for many years now” said Jeff Garrison, director of marketing for FPGA products at Synplicity. “We have a solid track record of providing products to our customers that deliver the absolute best quality of results when compared to others including those from the FPGA vendors themselves. By delivering tools that emphasize performance and area optimization and are technology independent, we are able to provide significant value to our customers making it a solid business decision for them to justify the higher price for the design tools.”
If your company plans to buy more than a few FPGAs, a tool investment can pay for itself in one or two designs if you save a speed grade or squeeze into a smaller device. Even though commercial EDA tools for FPGA are significantly higher than the vendor-subsidized tools, they are still very inexpensive by ASIC tool standards. Even a slight improvement in quality of results (QoR) or a small speedup in time to market can offset the difference in tool price many times over.
Mentor Graphics leverages the breadth of their product offering to get potential customers to look at the big picture instead of focusing exclusively on the FPGA piece. With products ranging from electronic system-level (ESL) tools for capturing system specifications and architecture, to printed circuit board layout, signal integrity, timing analysis, IP, and embedded software tools, Mentor is cutting a wider swath through the system design field than any FPGA-focused company would attempt.
In the FPGA vendors’ camps, life is not all peaches and cream either. Five years ago, when a “complete” design tool suite meant HDL simulation, synthesis, and place-and-route, and most FPGA design was simple bridging circuitry to match up ASSPs or ASICs with incompatible standards, maintaining a solid tool offering was a manageable task. Today, however, with complete system designs built around FPGAs including digital signal processing (DSP); embedded processors, software and peripherals; high-speed serial I/O; and complex memory interfaces; the tool requirements have exploded. Any FPGA vendor that wants to maintain a complete and competitive tool suite developed in-house is looking at hiring hundreds of software engineers and spending tens of millions per year in development costs.
In today’s reality, only two FPGA vendors have the chutzpah to challenge commercial EDA with their own proprietary tools. Xilinx and Altera each maintain tool development groups that are comparable in size to their core silicon design teams. They reason that tools are a source of potential competitive advantage, and they want to capture every advantage they can get. Also, historically EDA companies have not shown the level of commitment to FPGA that they made to ASIC tools. The largest two FPGA companies needed their own tool programs to insure their long term survival.
Other vendors like Actel, Lattice, and QuickLogic rely on commercial EDA companies to provide tools and technology that they redistribute to their customers. These OEM licensing agreements actually provide an additional revenue stream and distribution channel for commercial EDA companies trying to make money in the FPGA market. “Actel can offer its customers a one-stop-shop tool suite that includes the industry’s best third-party tools, as well as our own optimized implementation tools,” says Saloni Howard-Sarin, director of antifuse and tools marketing at Actel. “We understand that third-party vendors, who live and breathe EDA design, are going to reign superior to the FPGA vendor, especially in areas like synthesis, formal verification and physical synthesis. Rather than try and compete in this area, Actel has chosen to focus its expertise on back-end tools, where extensive knowledge of chip architecture is necessary, and join forces with leading-edge EDA tool vendors who possess the expertise and technology to provide first-in-class front-end tools.”
Even the hardiest FPGA companies are hedging their future bets on an all-inclusive tool offering. Xilinx announced this week that it is creating a new $100 Million “ecosystem” venture fund to fuel programmable system innovation. This is a notable deviation from the company’s established path of a proprietary, one-stop-shopping model for programmable logic products and tools. The reality is that probably no single company can maintain leadership in every single design dimension as programmable technology explodes into numerous new markets and applications. Funding and encouraging independent development of new technologies to support the progress of programmable systems design is a wise move on Xilinx’s part, even if it makes the company’s tool strategy appear somewhat schizophrenic for a short time.
The real, long-term battle for control of the tool environment will probably be over the issue of stickiness. As we discussed previously in our “Sticky Business” article, FPGA vendors invest heavily in developing IP cores that they provide almost for free in an effort to lock design teams into their proprietary technology. The motivation behind their design tool programs is much the same. If a customer locks his design into the FPGA vendor-provided tool suite, migrating it to a competing FPGA technology is extremely unlikely. So, while breadth and depth of design tool offering are reasonable places for EDA companies to try to differentiate themselves, their real, long-term sustainable advantage is in vendor independence.
A small electronics company that does only a few designs per year may not mind being locked into a single-vendor flow. It may even provide them a temporary advantage as their engineers only have to familiarize themselves with a single company’s technology and methodology. For larger companies, however, maintaining multiple, single-vendor, proprietary tool suites and methodologies can be a nightmare. There is significant benefit to be gained from standardizing on a single design environment that works across multiple vendors, and building internal experience around that one methodology.
The IP picture is affected by this decision as well. “The only IP that has competitive value is the IP you develop yourself,” says John Hengeveld, Group Marketing Director at Mentor Graphics. “Commodity IP is useful, but everyone has similar access to it, so there’s no real competitive benefit. A tool suite that protects your investment in your own IP, and makes that investment portable across various target technologies has significant value.” This argument also makes the most sense when applied to well-funded organizations that are interested in developing a portfolio of design technologies that will give them competitive advantage over a number of projects and products. The single-product startup doesn’t stand to gain as much if anything.
What we may see in the future, then, is a scenario where the FPGA vendors’ tool suites are increasingly used as the “introductory” flow that gets new teams through their first few programmable logic projects, and the EDA vendor-supplied tools are used as the “production” vehicles that companies rely on for sustained, long-term development. If this turns out to be true, the EDA companies and startups with the foresight to stake positions now in the FPGA market may be the big winners in their own space over the next few years. Even now, while FPGA vendors may have a dominant position in terms of the total number of seats, their top-tier accounts and accounts doing the most complex FPGAs primarily use EDA-provided tools for better results and vendor independence.
What’s best for your design team in the meantime? If you’re just doing a one-off project and want to get going as quickly as possible with the least cost and effort, the FPGA-vendor tools are clearly the place to start. You get one-stop shopping and support, and the cost/capability ratio is the best you’ll find. If, on the other hand, you’re doing a high-volume design or if performance is critical to you, you can probably optimize those last few percent more effectively with a set of third-party tools, particularly on mission-critical steps such as synthesis or physical synthesis. Also, if you’re part of a larger organization that will be doing designs with a number of different FPGA companies, and want to build up a library of proprietary IP to leverage across multiple projects, a vendor-independent tool flow that captures your entire design process is probably preferable to a vendor-locked suite.