CadenceLIVE 2020: As It Happened

CadenceLIVE 2020 Americas took place virtually earlier this week, spread across Tuesday, Wednesday, and Thursday. First, there were close to 2,000 attendees. Since this was a virtual conference, it's not fair to compare that number to last year when it was in-person. But any way you look at it, it was a big success with everything up and to the right. People attended 67 sessions, and 5 keynotes.

Today I'll look at those keynotes, although I'm going to give a couple their own posts next week. Everyone has to talk about data, artificial intelligence, and security in a keynote these days, and I think almost everyone did!

As a general comment, I thought the keynotes came across really well. When the Olympics are in an inconvenient timezone, NBC broadcasts them as "plausibly live" so that (if you avoid the temptation to discover the result on the net) it seems as if it is happening right there. These keynotes came across like that. I know they were pre-recorded (I'd seen some of them in advance) but they came across as being delivered live. The fact that they were not going to be available for replay later meant you had to be there (we actually relented on the replay thing and they are available until the end of this week aka today).

Lip-Bu Tan

Lip-Bu, Cadence's CEO, opened the conference with a keynote Fueling the Data-Centric Revolution. He started by talking about the five drivers of the industry, which I wrote about recently in my post The Five Waves: AI, 5G, Cars, Clouds, IoT.

He spent most of the time on what is currently the biggest driver of all, the huge amount of capital flowing into data center investment. The reason that this is so important is that automotive is challenged right now, since we're all working from home and not buying new cars; mobile is big but not growing fast. In any case, a lot of what is driving data center (and cars and mobile) is AI.

He dug into the technology innovations and semiconductor opportunities in today’s environment, and how those are driven by the restructuring of the ecosystem around the hyperscale data center. He also showed all the new products that Cadence has brought to market in the last few years, what Lip-Bu calls the pipeline of Innovation. Note that these are our organically developed products. We have made some additional acquisitions, such as AWR and Integrand earlier this year.

AWS

Nafea Bshara of AWS told the fascinating story of how Annapurna was created, originally using their own servers. Then they switched all EDA to the cloud. This made for a huge efficiency gain and cost-saving. He finished with some advice on how to re-structure your organization when you move EDA and design into the cloud. I'll give that its own blog post sometime next week.

Facebook

Some of the most interesting things that Vijay Rao of Facebook presented was on the scale at which they operate. He actually opened with a fascinating video (you had to be there though) on what it takes to build a Facebook data center, from pouring concrete, to electrical, to installing servers.

Here's a picture of their Singapore data center. Land is limited so they had to go up instead of doing the usual four football fields long style of data center. If you want to see more pictures and some details on how they build them, Facebook has a whole data center video page.

Facebook operates at scale: Their monthly-active-users (MAU in the jargon) are:

• 2.5B for Facebook itself
• 1.5B for Messenger
• 1B for Instagram
• 2B for WhatsApp

But more impressive still is the scale at which Facebook operates. Those big numbers are the monthly active users (MAU in the jargon) for Facebook itself, Messenger, Instagram, and WhatsApp. Some of them are the same people, of course, so it doesn't make too much sense just to add them up...especially as that exceeds the world population.

He moved on to talk about AI and how that is used, and, in particular how they are investing heavily in purpose-built hardware. They do a lot. They have 3.5B images that have had hashtags added automatically. They do over 6B translations per day.

For more details on their strategy, see this post on Facebook's blog Accelerating Facebook’s Infrastructure with Application-Specific Hardware.

They have built it to be scalable with common ASICs that fill modules that fill a chassis, as in the picture above. The systems are very complex and "We have collaborated with Cadence and many other partners." And to prove it, he had lots of photos of Palladium.

Anirudh Devgan

On Wednesday, the keynote was by Anirudh, Cadence's President. One of the things he did was to announce Xcelium ML which I already covered in my post Xcelium ML: Black-Belt Verification Engineer in a Tool that went live just as Anirudh took the stage. I'll talk about the rest of what he said next week.

Alberto

The keynote yesterday was by founder and long-time board member of Cadence, Alberto Sangiovanni-Vincentelli. The title was Cadence and Academia: From Transistors to Systems with Computational Software. I will cover that in more detail next week, but in the meantime, he started by likening innovation to truffles. Nobody really knows how to cultivate either.

So instead of a cat video, I'll leave you with a truffle dog.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email