Gary Meyers: FPGAs versus ASICs
by Peggy Aycinena
March 25, 2009
The principle subject of the DVCon 2009 Executive Panel in San Jose in February was "EDA: Dead or Alive?"
The seven speakers on the panel were unequivocal in the answer: EDA is definitely alive, because it’s the single most important pre-requsitive for progress in electronics. The electronics industry will never die and, therefore, neither will EDA.
In the weeks after the event, I spoke at length with all seven panelists individually to follow up on various topics discussed at DVCon. This article recaps my conversation with Synopsys executive, Gary Meyers. Links to the rest of the conversations in the series can be found at the end of this article.
FPGAs versus ASICs ...
Q – Can you distinguish between an FPGA and an ASIC?
Gary Meyers – An FPGA is field programmable, as the name implies. Customers can buy the product off the shelf and program it on their own premises, where as an ASIC is manufactured to a customer’s specification. The distinction has not changed since the dawn of both industries.
I come at this from the perspective that there’s room for both technologies. If you look at the FPGA industry, it represents 15% of the ASIC industry in terms of size and semiconductor revenue. It’s certainly a good business, but considerably smaller than the ASIC industry.
From a growth rate point of view, FPGAs are set to grow slightly faster than ASIC revenue. Even if you were to project that forward, it’s still hard to project a scenario where FPGAs would overtake ASICs. Due to the architecture of FPGAs versus ASICs, in an FPGA you’re giving up efficiency for programmability.
ASICs still offer higher performance, higher capacity, lower power, mixed-signal integration, and more efficiency. Most importantly, they offers lower unit cost over volume than an FPGA.
If you look at low power, there’s a huge disparity between FPGAs and ASICs. Certainly there are now some very small FPGAs manufactured with much lower power, but there are no large, complex FPGAs that would, in fact, displace ASICs in volume – particularly in consumer applications where power is paramount.
I can’t see a time where an SoC in a cell phone is displaced by the equivalent FPGA.
Q – So, it’s the power issue?
Gary Meyers – Yes, but there’s also the efficiency of the architecture. The massive routing grids in FPGAs present an area penalty. FPGAs, even manufactured in volume, can be considerably more expensive for large designers. If you try to look several generations ahead, you could possibly see a cost equivalent, but that’s not really a fair comparison.
So, cost, power and performance are all issues. You have devices and PCs running at several Gigaherz, but FPGAs run in the low Megaherz, and part of that is power related. Part is, the fabric in FPGAs prevents them from getting the performance of an ASIC device.
It’s an interesting discussion to compare the two technologies, but truly it’s the application that is the final [differentiator].
Q – Does a generic hardware platform need to be an ASIC?
Gary Meyers – Yes. There’s more differentiation in the software, which is why we see so much growth in rapid prototyping solutions. But hardware can also be a source of differentiation.
Let’s not forget that companies develop custom chips for a reason, so to say that everything is a standard chip and the differentiation is in the software, is too extreme a viewpoint. Companies are realizing that a balance is important. Some features are in the software and some in the hardware.
Q – Have you found the growth of the FGPA market to be remarkable?
Gary Meyers – I always thought FPGAs would be very viable, and I saw terrific opportunities for EDA tools to be sold into the FPGA industry. Every generation, more applications that "couldn’t be in an FPGA" move to an FPGA, nonetheless. But they are still at modest, modest volumes and have not moved the revenue needle that much.
It’s about 5% annual growth for FGPAs, and 2% for ASICS, so from a revenue perspective, there’s not that much difference going forward. You might say there are more designs being down in FPGAs, but there will still be markets where FPGAs can’t penetrate and those are the markets with the bulk of the semiconductor revenue. It’s been the history of FPGAs that as newer generations of the technology come out and get bigger, it becomes attractive to larger segments. But the largest markets still stay unavailable to FPGAs.
From my perspective, I just don’t see FPGAs taking over volume markets.
Q – How have the two worlds blended with the Synplicity/Synopsys acquisition, the FPGA world of Synplicity and the ASIC world of Synopsys? Particularly given the historical difference in price between FPGA design tools, including the free tools from Xilinx et al, and ASIC design tools.
Gary Meyers – There has been a blending of the two attitudes. There is an opportunity for FPGAs tools that are free, provided by the FPGA vendors, but there’s increasing demand for the kinds of tools that EDA providers produce – low-power, IP, synthesis tools. These are all increasingly important to the FPGA industry.
At Synplicity, we built a profitable public company based on FPGA technology, but I still don’t buy the premise that the tools are free.
There is an opportunity now for both technologies to thrive. Synopsys saw that opportunity, and not just to strengthen their FPGA business, but to tackle rapid prototyping in the ASIC market, as well. This is not an either/or proposition, FPGAs or ASICs. Synopsys is the leader in ASIC design and verification, and now we are the leader in FPGAs as well.
Q – Sill, I wonder why Xilinx didn’t buy Synplicity. Seems like a better fit than with Synopsys.
Gary Meyers – I disagree, and so does Synopsys. They were interested in our core business, and the opportunity for rapid prototyping – platform plus tools. To develop a really strong prototyping platform, you need both the hardware and the software.
Q – Do you promote the use of FPGAs in end products?
Gary Meyers – We have a good business in FPGA tools for those who use FPGAs for production purposes, vendor independent tools which are important for a large number of companies. We continue to see growth in this segment every year.
That’s the principle benefit of our merger with Synopsys. They were always connected to the ASIC customers, and Synplicity was always connected to the system houses. Over time, as FPGAs have gotten bigger and more capable as platforms for ASIC customers – in particular, with the growth of embedded software and rapid prototyping – we now have the advantage of being part of the substantial Synopsys sales channel. We see the merger as being greater than just 1+1, greater than 2, the sum of the parts.
Q – What constitutes your principle bucket of customers?
Gary Meyers – It’s a split. We address both the FPGA product companies – communication accounts and mil/aero – and IC companies doing prototyping. By being part of Synopsys, the part that sells to the IC houses will grow faster. Of course, we continue to focus on production FGPA customers, and virtual platforms on FPGAs. By connecting our tools with the existing tools at Synopsys, we can provide a more complete solution.
Another example – with Synopsys VCS, we’re now connecting that into our prototyping offering. This is an opportunity for us to provide more solutions to our customers.
Q – Do you think the trajectory of the ASIC/FPGA paradigm has changed due to the merger?
Gary Meyers – Clearly there is now a recognition that FPGA platforms are essential to modern ASIC design. However, the fundamentals haven’t changed. It’s a rare case when a design manager struggles between implementing on an ASIC versus an FPGA. When it’s about cost and volume, the choice is always the ASIC. And yes, these days it’s not uncommon to hear FPGA executives say that more and more designs are being implemented in FPGAs. That’s definitely been the trend for 10 years now, but it still hasn’t moved the revenue needle in the overall semiconductor industry.
Q – Has the merger been a smooth one between Synplicity and Synopsys?
Gary Meyers – It’s been very smooth. Synopsys has been very welcoming, and we continue to invest in both FPGA and rapid prototyping tools. This year, we’ll have 40-nanometer FPGAs that can support several million ASIC gates. The same old challenges exist here, but we have an opportunity to meet those challenges with new technologies.
Q – Do you consider the Synplicity Business Unit to be an EDA supplier?
Gary Meyers – Absolutely, we are an EDA provider! EDA is all about electronic design automation. We just happen to be focused on a different segment of the semiconductor market. But, we provide tools that help designers get their jobs done more quickly and with better results. We’ve always been focused on that. The fact that we are now art of Synopsys does not change that focus at all. Rapid prototyping is now part of EDA.____________________________________________________
Gary Meyers is currently Vice President & General Manager of the Synplicity Business Unit within Synopsys. Meyers joined Synopsys in May 2008 with the acquisition of Synplicity and is currently the Vice President and General Manager of the Synplicity Business Group, where he is responsible for FPGA implementation and DSP synthesis tools, and the Confirma rapid prototyping solution. Prior to Synopsys, he served as President and CEO of Synplicity and held senior sales and marketing management roles at LSI Logic Corp. after having started his EDA career as an ASIC designer at TRW, Inc.
The DVCon 2009 Panel Series
*Atrenta’s Ajoy Bose & Mike Gianfagna:
Recrafting the concept of EDA
*Javelin Design’s Diana Feng Raggett:
Government funding for R&D in EDA
*Synopsys’ Gary Meyers:
FPGAs versus ASICs
*EVE’s Lauro Rizzatti:
EDA in Europe
*Berkeley Design Automation’s Ravi Subramanian:
Setting the EDA roadmap
*SpringSoft’s Scott Sandler:
EDA dead or alive?
*Calypto’s Tom Sandoval:
Reaching the EDA customer
Peggy Aycinena owns and operates EDA Confidential:
Copyright (c) 2009, Peggy Aycinena. All rights reserved.