What Would Joe Do?
Peggy Aycinena is a freelance journalist and Editor of EDA Confidential at www.aycinena.com. She can be reached at peggy at aycinena dot com.
Teklatech: Work smart, Not hard
October 19th, 2017 by Peggy Aycinena
The discussion then moved to the business challenges within EDA, a topic that Bjerregaard said he always enjoys discussing. As he feels that Teklatech is providing critical solutions for today’s semiconductor designs, it’s not surprising that his enthusiasm here is palpable, for the industry and the trends that drive it.
WWJD: How are things going
Tobias Bjerregaard: Things are going very well, thank you.
First, I will tell you a little bit about Teklatech. This is an EDA company, a niche company focusing on the specifics of power integrity at the back-end of IC design. Power integrity has had a lot of attention in recent years; engineers think it is a very important.
As is typical in engineering, a problem is recognized and then analysis is done, which is exactly what the RedHawk analysis tool from Apache [Ansys] does at the full-chip level. We also understand the problem, and have optimized [a solution] for on-chip design.
We have been around for 10 years, and have focused on this power integrity optimization technology for about 8 years, which means we have a lot of know-how in this specific challenge related to power integrity.
If you go back to the ‘childhood’ of EDA, with the first synthesis and place-and-route flows in the 90’s becoming fully automated, power was only an afterthought. You placed and routed your design, and then did the power analysis.
But now with scaling, the power density is going up and up and the metal sheet resistance is going up and up, and this has caused the dynamic power density to increase.
What we seeing at 16 nanometers, and particularly at 10 and 7 nanometers where it gets really bad, the designs are routing constrained. To explore the full benefits of scaling in terms of area, the limiting factor is routability which means you need metal, but not all of the metal is available.
What’s happening now in the scaling of technology, the challenge of power integrity is becoming increasingly costly. We are seeing where power integrity is now a limiting factor toward profitability in the semiconductor business model.
In terms of Teklatech, this means we are now working with some of the biggest companies around. The problems we are addressing are really costly for these companies, and optimizing is their key to remaining competitive at the most advanced nodes. We are now working at 10 nanometers for production for some customers, and are moving to 7 nanometers.
WWJD: For Teklatech, it sounds like your time has arrived.
Tobias Bjerregaard: Exactly. We were just a bit early starting out [with this technology], to put it mildly. But the good news, we’ve stayed around and built our know-how, and have a very mature product ready for offering at the right time.
These are very exciting times for us. We have doubled our revenue over the last few years, and expect to maintain that trajectory going forward.
WWJD: I’m guessing, you are not feeling the pain of the slowing of Moore’s Law.
Tobias Bjerregaard: [laughing] No, we are the solution to that.
Solving the power integrity challenge is a major part of addressing “the end” of Moore’s Law so I’m glad you asked this because this is very dear to my heart. After all, what is Moore’s Law?
WWJD: Thank you, because I didn’t want to ask the obvious question.
Tobias Bjerregaard: [laughing] We see it as scaling of transistors, but that is only a proxy for applying of the concept.
What’s really happening is also the scaling of power, area, and performance. So how do we exploit the resources available for scaling? At 7 nanometers, it gets really crazy. Routability is a major constraint factor, because of all of the design rules that make it impossible to add flexibility to your routing. It’s becoming a huge issue, which could mean an end to Moore’s Law as we know it.
It may be costly to scale transistors, but Moore’s Law is not just a physical barrier. It’s more of an economic barrier, because it becomes more costly to go to the next node.
Since 28 nanometers, the cost of transistors has not gone down. But that ‘going down’ is an integral part of the semiconductor business model, so in that way the semiconductor business model is broken.
I believe we are in an era where we need to work smarter, not harder. We need to find that profitability elsewhere. We need to optimize our designs better, and shouldn’t be leaving anything on the table.
With Teklatech, we let designers optimize their designs in a smarter way, rather than just through brute force. We’ll go to the next node, make the power go down, but the performance will be up. Although, of course, it is not as easy any more as it once was.
WWJD: Your arguments seem compelling. Do you still have to knock on people’s doors to get attention?
Tobias Bjerregaard: Now we are starting to see the first people knocking on our door.
Yes, selling EDA is hard – there’s no question about that. Even at big companies like Intel, Samsung, Qualcomm, there are only a handful of people, the stakeholders, who really understand these problems and are able to see us as the wise strategic choice.
And there is no question that EDA is a global business. Even for a small company like us, we need to be globally present and need to know exactly the right people. But that can be quite difficult, so word of mouth is a big part of it, things like this article about us that you are writing.
We try to target our messaging through very special communication outlets and media, but we also rely to a huge degree on personal connections, mapping out an organization so we can understand who should [see] the direct evidence of our results.
This has been very successful. We have never had anyone who wasn’t awestruck by our technology, because they didn’t know until then that it’s really possibly to solve these problems.
So our job, in marketing this kind of technology, is first to make people aware that they can do something about it. They think power integrity is in the physical domain, handled through ECOs, but ECOs have never been good for convergence of the design flow. If you change something at the physical level, a new problem pops up.
You really want to go correct-by-construction by actually prepping the design to behave more properly, so there are not so many problems to fix in the end.
It’s all about design closure at the optimum design and performance level. And that message requires a very specific audience [at the customer] that has a specific strategic role in the company.
Again, there are only a few of them in each company, and they are hard to reach. But we have a good name in the right circles and, of course, are always trying to spread our message. This is important to us and to the industry as a whole.
Also, we are seeing a shift [with regards to] timing closure. Although power was important, timing has always been king. But that is changing. The old battle between AMD and Intel, the battle for megahertz, that’s not as important as it used to be.
In the same way, at the back-end, timing closure was once the biggest hurdle to design closure and everything else was a second priority. EDA now has a pretty good handle on timing closure, with designers using cell replacement and other [strategies].
So power integrity is the next big bad thing. You don’t know when the problem will pop out, so priming the circuit at an early stage is a huge, critical challenge in the semiconductor industry that will show profitability in the next 10 years.
WWJD: How does your tool work?
Tobias Bjerregaard: RedHawk does the sign-off analysis all the way at the end to check how things went. Then the designers ask: Do we need to go back and fix something [through] an ECO?
Instead, our FloorDirector is really an implementation technology. It fits closely into place-and-route, just after placement, because our tool does an early stage, post-placement power analysis – a dynamic power analysis with power-wave forms.
At an early stage, power and routing are not complete. Maybe the design hasn’t even been routed yet. But we are able to do dynamic routing analysis to get a qualitative feel of where in the design there will be hot spots, and so on.
We are able to treat a number of parameters – dealing with dynamic power, and static, and scheduling, and even improve timing. Doing all of that at an early stage has a number of benefits. It basically provides guidelines for the rest of the implementation, so the problems are already fixed.
Also with our technology, there is capacity for dynamic voltage drop analysis at an early stage, so designers can make changes and do what-if analysis and actually fix problems. Later on, if you rip up one part of the design, you cause other problems. But this is not at issue at an early stage because you haven’t locked in the design, so you still have flexibility.
What differentiates our tools, they are very robust [in the face of] incomplete design data, where sign-off analysis relies on everything being there.
WWJD: These seem like critical solutions, so why hasn’t a big EDA company bought you yet?
Tobias Bjerrregaard: [laughing] We haven’t put up a ‘For Sale’ sign.
The big companies have had a lot of focus on timing closure until recently, and they’re very good at it, but it’s funny: We can actually improve timing over what they are doing with what we do.
The reason that we haven’t been made part of a bigger flow is we’re doing quite well and the time hasn’t come yet. Or maybe it will not come, because ours is an agnostic tool that fits into all of the design flows.
And once you’re in a larger company, your market is limited. So we haven’t seen a good reason [to be acquired]. We have not seen how that would be of benefit to us or to our customers.
WWJD: Do you have to travel a lot to see your customers?
Tobias Bjerregaard: Yes, everybody in EDA needs to travel a lot.
There are concentrations of customers in Silicon Valley, in Southern California, in Texas, and there are a lot of other places in America, in Taiwan, in Korea. It’s a very global industry, so you can’t be a local player in EDA. Wherever you are, whether in Europe, the US, or Asia, it’s all the same.
And it’s important to be where the talent is. We have R&D in Copenhagen, a subsidiary in California, and have agents in Korea and India where we have users of our tools. We are spread out [around the world].
Being in Denmark is actually a very good thing for an EDA company. First, there are very talented engineers here. We have a great educational system producing software designers who are creative, who take ownership of their work, and bring structured thinking [to problems].
Also, compared to Silicon Valley, the competition [for employees] is lower here. If an EDA company trains an AE, it won’t be long before the customer hires them away. But there are not a lot of semiconductor companies here in Denmark, so we have loyal people who stay a long time.
[laughing] And there are direct flights from Copenhagen to San Francisco and Oakland, so that is not an issue for reaching Silicon Valley.
WWJD: I’m a big fan of flying non-stop whenever possible.
Tobias Bjerregaard: [laughing] Yes, but when you are booking only a few days in advance that is not always possible. And the London connection to San Francisco is not too bad.
WWJD: Why belong to the ESD Alliance?
Tobias Bjerregaard: I am a firm believer that an industry as a whole, and as individuals, can benefit from these types of alliances, the meeting up, networking, and sharing market information.
A big part of the benefit of the ESD Alliance is that they collect market information, and distribute it in an anonymous fashion. You can’t do that without a third-party alliance.
Also, for the collecting of general industry trends, technology topics like OS use, and for creating get-togethers at DAC. This is all very important to have.
WWJD: What are the important conference for Teklatech.
Tobias Bjerregaard: Definitely DAC. It is hard to see the others as important.
Although DesignCon is possibly one, I was on a panel at DesignCon. But it addresses a different type of engineer – system-level engineers, not so much chip designers – so DAC is the only one that really makes a difference.
One big conference a year is enough, a place where people can meet with you and ask questions. But a conference must wait for that particular date when everyone can be together. It used to be that a conference was where you met the customers, but they are becoming less and less important.
Instead now, you have your website, you meet with customers online and in web meetings, and email has made communication much easier. So with online presentations and video conferences, the need for conferences is really lessening.
WWJD: Online meetings also make it possible to coordinate a company’s team when it is distributed geographically. Is it still important for everyone to be in one place regularly to talk over a beer?
Tobias Bjerregaard: This is a very relevant question, because we are still human.
We are tied to this body, we need to be friends, to read the emotions on faces. Talking on Skype makes that easier, but having that beer is also important so you can know each other.
For Teklatech, being spread out across the world, we have our weekly online meetings where everyone is present, hearing the same thing. With any new release, all of our AEs across the world dial in, ask questions and hear the same answers.
But we also make a big effort, myself, our VP of Engineering, as well as our VP of Sales. We travel around a lot to meet our agents and our AEs. We meet with them in person several times a year.
And once a year, everybody flies into Denmark. Yearly, the whole company gathers all together for a week. We do presentations, the AEs talk about the customers, R&D talks about new things coming up. We all have that beer and really get to know each other, which makes working together during the rest of the year so much easier.
WWJD: What are some funny things that have happened during your career?
Tobias Bjerregaard: [laughing] When we crashed the TSMC party at DAC. And when my first sales director decided that we should brand me as the Bad Boy of EDA, that I should only wear furs and go out with super models.
WWJD: And how did that work out for you?
Tobias Bjerregaard: [laughing] It was a great idea, but was maybe meant for another time. We did not pursue it.
WWJD: Do you get ribbing that some of your old code is still around?
Tobias Bjerregaard: [laughing] Our employees do make jokes about my old code. Sometimes when they’re having trouble, they say: That must be Old Tobias’ code.
But I say: If it is still there, it must still work.
Dr. Tobias Bjerregaard is Founder and CEO of Teklatech A/S. Under his leadership, Teklatech introduced the first commercial version of FloorDirector in 2008. Bjerregaard has spent more than a decade with a focus on leading-edge micro-electronics, and has invented and patented several semiconductor innovations in the areas of power integrity and clock distribution networks.
Bjerregaard has received numerous awards for his individual achievements and industry contributions, has written and presented a range of papers for leading industry publications and conferences, and has been recognized worldwide for his contributions and achievements. He has an MSEE and PhD in micro-electronics.