EDA and the Shift-Left with Mike Ellow – Part 2 – Transcript
In our first conversation of the latest season of the Industry Forward Podcast, Dale and I were joined by Mike Ellow, CEO of Siemens EDA! Mike holds 30 years of executive sales and technical management experience in Electronic Design Automation (EDA). We invited Mike on to gain his insights on the transition to software-defined products and systems, the shift-left of engineering processes, and especially the increasing need for the integration of EDA and semiconductor development into the creation of complex systems like airplanes, cars, and more.
Listen to the second episode and read a transcript of the conversation below!
Conor Peick: So Mike I was, you know, we’ve been talking about this transition towards the software defined product and it got me thinking about the kind of meshing together of the software, the electronics and the mechanical, maybe the more traditional sort of subsystems. And particularly kind of the meshing together of the digital ecosystems, I guess you could say. You know, does this create a big need for these domains to talk more closely or integrate more closely I guess you could say? And if so, what do you think companies can do to sort of enhance that integration? To bring them closer together and be more effective?
Mike Ellow: Well that’s a interesting and loaded question there Conor, because, if you take a look at where we are with this idea of software is the primary differentiator now, and so the system integrator, if you will, is looking at software being the primary driver, so that dictates into the entire connected design community, that ecosystem, what all the deliverables are. And you can imagine in a world where you’ve got specific software workloads that you’re trying to optimize. And you’ve got a configuration of all of the swim lanes, and this includes the semiconductor as placed on a PCB that sits inside of an enclosure that then has, back to an automotive example, you know, a specified set of requirements around, let’s use an EV, for example the range associated with it. So that’s got weight requirements, that’s got then implications on what your engine looks like, and your battery configurations and everything else.
Now all of a sudden we make a change in software, because some evolving requirements associated with that, that changes some of the power dynamics. You know the compute platform needs to draw more power because there’s an enhanced set of capabilities required by software. All of a sudden that has an impact on the existing battery ideas and range. So now we’ve got to contemplate, well, if I’ve got to change the battery, well then maybe there’s a configuration change that’s happened, so it has an impact on the body. If the weight changes, well, maybe that has an impact on horsepower for the engine, which then has an impact on the braking system. You can imagine how changes that you’re making where you’re optimizing for differentiation has a ripple effect throughout the entire set of swim lanes connected to the development of the system. Well, how do you effectively communicate that, then, through the entire ecosystem so that you’ve got real time information that specifications are changing and because of that we need to modify what we’re delivering back to this system. Well, that’s a verification digital thread that has to now exist from what’s happening with the owner of the overall set of specifications associated with this complex system-of-systems. How they are then looking at the feedback loops with changes that then have to be proliferated back into the ecosystem and then as part of that, make sure that as each of these entities are going down their specific development path that they also are providing realistic feedback into the system, verifying that they can meet all of these requirements ’cause if they can’t, then that may have a change someplace else.
And this is, where, you know, you start taking a look at where we’ve been in the past versus where we need to go into the future. Based on the complexity of what we are contemplating here, solving the connectivity with, you know, verification capture points, for example. You know, how we’re starting thinking about the digital threads that go through all of these swim lanes where we can take a look at the outputs of each one of these activities against the requirements and verify, yes, we are meeting those or no, we’re not meeting those and then, if we can’t meet those, what the implications are somewhere else. This is how we enhance that predictability of on time delivery at the level of functionality required with the proper cost profile. Without that you know, you see too many examples today, especially in aerospace and defense, where you’ve got years of elongation of design activity and cost overruns because the complexity of what is being contemplated changes. And then if you’ve got a three-year development cycle and all of a sudden you’re 10 years into the program and you’re just starting to ship first articles, you know that you’ve had issues there with what you contemplated at the system level versus how you just punched that out to the entire ecosystem to deliver against.
Dale Tutt: I think about all of the challenges with the software and electronics that that you know that I’ve seen in the past. That companies, yeah, as a as a chief engineer, we had figured out, you know, a lot of the mechanical elements and mechanical systems. But, you know, we’d see the delays of the software being, you know, being slow to get developed, it was out of configuration, out of sync. But you know, I think when we start talking about these digital ecosystems that it’s no longer just like one company worrying about all of these mechanical parts coming together. But it’s all the mechanical parts coming together, and all the electrical parts and all the electronics, and then the software and all of it has to work together and stay in sync. So the ability to control that, that configuration it’s, at some level, it’s a bill-of-materials solution, or a need, that I understand all of the configurations and how do they all tie together to be able to sync, make sure that everything stays in sync. And so when we talk. About you know, how do you make better decisions? You have changes coming from over here, maybe not meeting a requirement. It’s gonna affect other systems on an airplane or the car. It’s gonna affect how you’re gonna do your verification. And really, having that digital twin of those systems and all the different elements and being connected together, knowing that they’re all in sync is really, I think one of the tools that companies are starting to use to, you know, to really manage their processes.
So it is a challenge, but I’m gonna maybe shift gears just a little bit on your Mike. You know, as we continue to see an uptick in how all aircraft, automotive, everyone’s using more and more semiconductors and you started to touch them a little bit when you’re talking about some of the making sure that the semiconductor was set up for the future. There’s a lot of discussion about getting into the advanced nodes and things like, you know, 3D IC and so you know, as you think about where these industries are heading, what do you see as some of the motivations? And how, you know, some of these companies are adopting, you know, more advanced semiconductor technology to make sure that they’re ready to go in the future and manage that power, manage that bandwidth, that everything else that they need to worry about with the design of electronics?
Mike Ellow: What’s the change going on as far as this idea of a software defined product? I think that the fundamental underpinning of why customers are now rapidly reconsidering how they drive differentiation with software is really wrapped around what’s happening with heterogeneous integration, or 3D IC, or chiplets. I mean, you can call what’s happening out there any one of these names. But basically what it says is inside a package that a semiconductor used to be a monolithic piece of silicon in. Now you’ve got the ability to change the number of processor cores, change the amount of memory, change what kind of interfaces you have in there, consider AI accelerators. Such that you can more closely now match a silicon platform to your software workloads.
And because of this recent development, which is all enabled by advanced technology, I mean this means the smaller design nodes associated with what’s happening with transistors, it allows companies to say hey, “You know there is a efficient way both from the capabilities of the silicon platform, the power consumption, the cost of the devices that I can really optimize now my software based on how I imagine what a silicon architecture underneath it looks like by varying the number of cores, varying the amount of memory on chip and things like this, which didn’t exist in the past.” And so that’s why I think you’ve got this reimagining now of what the world looks like into the future. And many customers now saying, you know what for me to be successful, I have to consider doing my own development or having a very close silicon provider that is tightly linked to my program driving my specific requirements so that I get this as optimized as I can and don’t have to make the compromises in software. And that’s the uniqueness now. And the anticipated starting point for how this explodes into the future.
Dale Tutt: Yeah, you know, and I want to pull on that thread just a little bit more. A lot of companies, when they start a new program, they’re thinking about how do they manage their risks and they look at their development schedule. And I think sometimes at least in the past and maybe the old paradigm, was the development of a new semiconductor or new chip would be considered as a high risk, they, you know it’s a bit of an unknown because it’s outside of their expertise, so they would select an old chip, an off the shelf chip. And I always kind of joked that by the time our airplanes would hit, you know, would actually start being delivered, you know the electronics that were in ‘em were already 8-10 years old. And so you know, you just you were that much closer to your obsolescence problem and you also had limits on what you could do with upgrades.
So it is a paradigm shift for a lot of these companies to think about, OK, now, in addition to doing all of these other things that I’m developing around it, I’m also going to be developing either by myself or with a partner or semiconductor company, a new chip and the software. And so, you know, what’s the challenge for, but obviously it’s a challenge, I mean, but how are these companies now bringing that together that they can still deliver on that schedule that they said they would sign up for that they’re not signing up for an old schedule? I mean like you say, you know you reference, you know, some of the defense programs or some of the aircraft programs that are 10 years late because they’re still struggling with getting the software all aligned. How do you see that coming together? It feels like a much different paradigm than, say, 10 years ago?
Mike Ellow: I think what’s interesting for the semiconductor industry is that they also are looking at this as an opportunity for how they provide more value. So they’re looking at what their design cycle time requirements are, you know, and putting pressure on the EDA industry in order to compress those things into more predictability around 12 to 18 months versus maybe 18 to 36 months for some of the developments. In order to give you know a tighter time scale between availability of that silicon and what’s happening with the integration into a broader program. But also the virtualization that’s happening with the architecture itself such that you can continue software development in virtual environments as you’re developing the semiconductor device over time, getting better fidelity as the models start moving from a virtual world into more physical representations so that you’ve got, you know, a tighter linkage between understanding actual performance characteristics of the silicon, as enabled by the software over time, to the point that when you get silicon on dock, you know, you’ve got a high degree of certainty that the integration point of the software to that silicon is much less than waiting for silicon to come back and then you start a massive software development.
Because in that world you know you’re scaling up resources dramatically at a point in time, which is hard for a lot of customers because they don’t have infinite software developers available. Having an environment where you can start your development in a virtual world where we’ve got some interesting technology with what we have done to speed up that virtual environment to allow the individual design environments, or the IDEs, of the software developers to run at speed. Because a lot of the reasons why virtual environments in the past for software development have failed is that they are too darn slow. But we’ve got some enabling technology that allows these things to run at speed, so the software developers can continue in the virtual world.
And as the hardware platforms are silicon start getting better fidelity as far as actual representation of silicon and they’re modifying and understanding the, again, the performance characteristics against this evolving platform such that the marrying of software and silicon at the point in time of parts coming back from the silicon, semiconductor manufacturing process, you’ve got better clarity, right? Better fidelity. So it’s a new paradigm also with how we take a look at the new starting points for software growth so that you don’t end up with too much activity at a point in time that you can’t scale. And then all of a sudden, you realize that things aren’t working the way you anticipated because the physical reality of what you got back from the fab was mismatched with where you thought your software was headed.
Dale Tutt: Yeah, you know, and it’s a really good point and I wouldn’t say it always would just you know you said it was too darn slow on the software development. Think sometimes it was just. It was. We were always waiting. Many companies are waiting for everything else to get settled down enough that they can actually start their software development and actually go through their verification processes. And so it really does demand more virtual verification earlier throughout the entire design process. It does really, I think change how companies design and maybe you rely a lot less on physical, or your prototypes and you’re relying more on your virtual prototypes as you go.
Siemens Digital Industries Software helps organizations of all sizes digitally transform using software, hardware and services from the Siemens Xcelerator business platform. Siemens’ software and the comprehensive digital twin enable companies to optimize their design, engineering and manufacturing processes to turn today’s ideas into the sustainable products of the future. From chips to entire systems, from product to process, across all industries. Siemens Digital Industries Software – Accelerating transformation.
