Getting More than You Pay For - Part I
Someone once told me that a good journalist goes out to cover a story free of any pre-conceived notions. That's the only way, I was told, that you can really hear what people are saying.
This little story is about the FGPA tools used for design, and I'll admit up front that I broke that rule. I went into this series of interviews with a distinctly pre-conceived set of notions. I expected to learn why the big EDA vendors, with the exception of Mentor, seemed fairly uninterested in an aggressive expansion into the FPGA space.
I expected to hear that the big EDA guys couldn't/wouldn't see that reprogrammable, reconfigurable everything is the thing of the future. I expected to be told that the major vendors are ignoring FPGAs to try and shore up their established customer base, those who are doing more traditional cell-based design. I expected to hear from industry prognosticators that the EDA guys who refused to read the (FPGA) writing on the wall were going to end up losing dominant market position.
As it turns out, however, it's not like that at all - at least from what I heard from various people in the industry. The problem isn't with the EDA vendors at all. It's with the FPGA vendors.
The FPGA guys are giving away the tools for free - or essentially, for free. So, compared to, say, $75,000 or $175,000 for a traditional EDA tool, an FPGA tool from an FPGA vendor is only costing you $1,500 or $2,000. A complete flow will run you as little as $2,500. In other words, the tools are essentially free.
And, it's not like the EDA guys are at liberty to say to potential customers, “Watch out, because you only get what you pay for.”
Because the “free” design tools from the FPGA guys are apparently excellent, well designed, and well supported. The EDA vendors can't compete. And, for now, they're not going to try. At least, that's my conclusion. This week, we'll hear from a variety of people - none of whom are FPGA vendors. Next week, we'll hear from the FPGA vendors themselves, and from the leader in the FPGA synthesis market. After you've read it all, you can draw your own conclusions.
So, let's get a quick start out of the gate with a comment from industry oracle, Gary Smith.
Gary Smith, Chief EDA Analyst at Datequest/GartnerGroup - “The FPGA market is a dumb market to be in for any EDA vendor. Because the minute you mention FPGAs [to a potential customer], the price you can ask for the tool goes down by a factor of 10.”
“Right now, the FPGA vendors have really fouled things, as far as the tools are concerned. If they would just quit giving away the tools - something they should have stopped doing in about 1995 - they would have help from the major EDA vendors to support large designs. The ASIC vendors figured that out in the mid-80s. [It remains to be seen], when the FPGA vendors will figure it out.”
Ping Chao, Executive Vice President and General Manager at Cadence Design Systems, Inc. - “The FPGA segment obviously is getting a lot of attention these days. But, it's always a challenge for the EDA industry to figure out how to make money out of it. Strictly speaking, we're not participating in the market. We're continuing to [concentrate] on ASIC and SoC-type design.”
“We do participate somewhat in the FPGA verification market, but we don't have offerings on the synthesis side. We work with Xilinx and partner with them, and that's the extent of our participation in the market at this time.”
“There's at least one angle, however, where we do have an advantage. That's in the structured ASIC market. Recently, we've been working with NEC and other people in that area, but it's not really the FPGA space. It's the gate-array space. So there may be some strategic investments there, but mainly we're just watching the market for the moment.”
Jeff Jussel, Vice President of Marketing at Celoxica Ltd. - “We're sponsoring a Tuesday afternoon panel at DATE on exactly this topic - should FPGAs be the focus for EDA tools? We set up the topic based on a survey we did in the last couple of months of 2003. We found from that survey that the number of FPGA design starts are now about three times the number of ASIC starts, from 1800 design starts last year, to 2700 this year - a 50 percent increase.”
“We asked a bunch of designers about the role of ASICs versus FPGAs. Our argument said, if design starts are all going to FPGAs, per data from Gartner and our own data, and EDA tools traditionally follow the design starts because that's where you're selling seats, and if EDA makes money on those design starts, then shouldn't EDA be spending more money on FPGAs?”
“The topic may not be as sexy as talking about deep submicron, or the effects of 90-nanometer, or timing closure, but we still ask - if that's where the design starts are, isn't that where the EDA focus should be? The DATE panel will address increasing the functionality and decreasing time to market for FPGAs.”
“At Celoxica, we're focused on system-level design for FPGAs. Our tools allow a C-based design to be compiled directly to an FPGA. In that regard, we see a couple of things. The solution we provide is a unique flow for customers of FPGA vendors. Customers who need to get to system FPGAs need these kind of system-level design tools. For customers who might not get into hardware design, we provide a path for them to implement their design in an FPGA.”
“Typical markets might include data security, for example, where complex algorithms are needed to get to gates. If the customer doesn't want to go through gates, they can use our tools. In addition, we ourselves are a customer of the FPGA vendors, because we sell FPGAs on our boards.”
“Clearly, the FPGA vendors are important partners for us - and investors as well. They sell silicon, not software tools. No one at Altera or Xilinx is going to make money on tools. But traditionally, they've had to provide the tools because the [third-party] vendors just weren't there. When the FPGA vendors first started, there were only RTL tools available. The only way [the FPGA vendors were going to succeed] was to develop their own tools because EDA wasn't supporting FPGAs. Now that we're getting into the system space, the FPGA vendors need to get on with [partnering] with third-party vendors.”
“[From the perspective of the EDA vendors], I think we'll continue to see the traditionalists saying, 'No, we need to continue to focus on ASICs.' In fact, that's because we're all semiconductor guys and we're very used to the market being a certain way. But, if this thing is truly what's happening and we're moving to another level above the silicon, then it does change the game. I can't fault Mentor [for their commitment to FPGAs], but even their focus has been on the traditional side with simulation and synthesis, not on the system side.”
“The FPGA vendors have had a vested interest from the beginning in having a viable RTL flow in-house. The system-level space, however, presents a new area that they're attacking by partnering with third-party vendors. The FPGA vendors actually provide an important channel for us in that their customers are the people we're trying to reach, as well. They also help us out as technology partners. In fact, Xilinx is an investor in Celoxica. They ship our tools.”
“Relationships are very important [in all of this]. Both Xilinx and Altera are important to us - we put all of their products on our boards. As EDA vendors, we need to [protect and promote] these relationships as they are critical for our success.”
“Right now, Xilinx and Altera are taking over the commodity part of the FPGA business. At the high end, it's IBM where they've got a monopoly - highest performance, highest volume. They're the ones who can afford to put out millions of dollars for the newest fabs. In the middle, you've got people who can't afford the high-end, but who also can't compete with Xilinx and Altera.”
“Also what we're finding is that the design platform - an overused term - with a processor, maybe a DSP, maybe some customer hard IP, plus some blocks of reconfigurable logic, is being used to create a design for a particular niche. It's the hard IP and reconfigurable logic, which allows them to differentiate their product. Those are the people looking for system-level design tools that target the problem. So, that's what we're seeing, a [set of] mid-level customers attacking various market niches by way of application specific reconfigurable something-or-other. As an industry, we'll have to come up with a term for this device eventually. But for now, there really isn't a consistent term being used here.”
|
