Welcome to 2008. At this writing, the global markets are things of terrible beauty – worldwide, they’re down, down, down, with the Dow Jones this morning hitting a 52-week low, along with the NASDAQ. After spending yesterday’s Martin Luther King holiday watching the rest of the world’s indices get trashed, the NYSE dropped a ferocious 400+ points this morning, just past the opening bell, but then rebounded somewhat on news of an emergency rate cut from the Fed.
Meanwhile, anybody interested in EDA this morning would see that CDNS is mimicking the DJI and trading at a 52-week low, as is SNPS and MENT. Not so, however, for LAVA. What’s up with that? Why is it, that despite overwhelmingly bad economic news, Magma seems to be laughing in the face of danger? Hopefully, I’ll have a chance to track down the LAVA folks sometime in the next day or so for a chat about their take on that situation. I look forward to exploring the Mystery of Magma. [Editor’s update: The LAVA folks aren’t able to discuss such things, apparently, because they’re on the verge of their quarterly financial update.]
Although the two conversations with these gentlemen spanned both space and time, through the magic of modern computing it will appear the discussions took place at the very same hour, next to a roaring fire in a cozy, paneled library. Since it’s colder than the surface of Mars outside pretty much everywhere in North America at this moment, keeping that fireside-chat image in mind during the writing/reading of this piece is a definite attraction. So, put on your smoking jacket and pull up your big wingback chair. I think you’ll enjoy listening in.
I’m grateful for Nanette Collins for arranging my chat with Rick Lucier. I’m grateful to Ed Lee for arranging the chat with Jim McCanny.
Fireside chat: McCanny & Lucier
Question – Are you enjoying your job?
Jim McCanny – I’ve been here just over 2 years and, yes, definitely – I’m enjoying my job. But that doesn’t mean it wasn’t tough [to get the business] started. But, we feel that even in our short life, Altos has proven what we’re doing has value. We’ve got a good growth path and we feel optimistic about what we’re doing. Hopefully, others feel the same.
What we’re doing is a heck of a lot harder than working at a big company and we do keep long hours at times, but it’s a market that takes a long time to grow. We try to [balance] the grind with the fun.
Rick Lucier – Yes, I am. I’ve been at Carbon for 18 months now, and we’ve really clarified our position in that time. And, we’ve added many key customers and partners. Yes, it’s fun – not necessarily easy, but fun.
Question – What’s up with the market volatility today – does it impact how you do business?
Rick Lucier – No, people panic when things go up and down, but it’s too early to see what’s going to happens – how much of it is emotional. For us at Carbon at this point, it doesn’t really impact us. We’re going to continue to attack our markets and stay on the path we’ve planned for ourselves. Nothing’s going to change that.
Question – Give me the elevator pitch – what does Carbon do?
Rick Lucier – We automatically create system-level models that can be deployed for system-level design. We take Verilog and RTL and automatically develop a model that can be deployed in a virtual system prototype, which can be used by system architects and for pre-silicon software development. We’re attacking the primary bottleneck in EDA and design today – software development.
Question – Jim, what does Altos Design Automation do?
Jim McCanny – What we’re involved in, is taking the models created for devices and converting those to models at the cell level. Those cell-level models can be used for timing signoff, and can be used in the design flow for [looking at] signal integrity.
The old way [design was done], the designer would give the design to the characterization guy, and he would create the models. That was the Black Box theory. But with the Altos tools, we don’t require the designer to tell us anything because our tools can understand the functionality and electrical characteristics of the paths in the circuit.
Question – Is the famous whack-a-mole model too simple to describe the design optimization problem?
Jim McCanny – It may be too simplistic, but in the past designers could make tradeoffs between various design targets [and still achieve design closure]. You could look at the variables and treat them as independent factors [in moving to design closure].
But, when you get down to 90 nanometers and are working to get more components on the chip – working to make things smaller and more portable – you start worrying about leakage power and signal integrity and trying to add in optimization on top of that. You struggle to optimize even those two variables, when there were actually now five variables that need to be optimized – area, timing, power, leakage, and signal integrity.
Basically, optimization and place & route systems focus on making sure you get the timing that you need – but we’re seeing that this methodology no longer works. You can fix the timing, but blow the power. You can optimize for one or two variable, but only get [best guess] success with the others. You have to build in so much margin for your variables [that you don’t get closure anywhere].
Rick Lucier – At Carbon, we take a more holistic look at system-level design, not a traditional look. If you look at traditional hardware companies today versus how they looked, say, 20 years ago – they’ve got 10x-to-20x more software people than hardware people compared to 20 years ago, because that’s where the fundamental bottleneck is today. You look at the high-level problems [being solved] and you see really big gains in shrinking the overall design cycle by attacking the software problems. Without a doubt, those problems are hard to solve, but that’s how we’ve improved the chances of our customers getting to market faster.
So, rather than a whack-a-mole game, I would say that EDA has always been a game of bottlenecks. You look at where there’s a bottleneck in the design and you work to solve it. A successful EDA company solves a bottleneck and puts the solution into an existing design flow. How much you’re able to shrink the design cycle [defines] how successful you are. And that’s the way it’s always been in EDA.
Again, I would say that rather than whack-a-mole, it’s a question of the bottleneck problem. It used to be the hardware, and now it’s the software. By creating models like we do, that can be used at the system-level, we allow for earlier architectural and software tradeoffs to be made – and when you start doing that, 1) you start collapsing the design time, and 2) you becomes more efficient at testing the software and the hardware.
Question – Respectfully, Rick, is that really a new idea?
Rick Lucier – Well, definitely people have been striving for this for a long time. Maybe I’ve been in EDA too long, but it is something that’s been around for a long time. Now, however, the tools are ready to address it and the market is ready to invest – hardware/software co-design. Now there’s a whole host of vendors who are providing virtual prototypes, but the tough part [remains the] modeling. How are you going to model existing and third party IP and accelerate them? Since hardware is becoming more commoditized, more product differentiation is being done in the software, so it’s critical to get an earlier start on the software side of the equation.