Exploring the 5G + AI + IoT Wave with Rene Haas of Arm

Spela upp

In this episode of The IoT Unicorn Podcast, Rene Haas, President Intellectual Property Group at Arm, discusses the development of edge devices and the 5G wave. Download Transcript Here 00:00 PETE BERNARD: Rene, thanks again for joining us here on the IoT Unicorn. I was trying to remember the last time actually we saw each other face-to-face. That's something that we do these days. I think it was Barcelona 2019 or something. It was a while ago. But again, thanks for joining us today. 00:23 RENE HAAS: You are welcome. I wasn't sure if it was CES of 2020, but... 00:28 PETE BERNARD: It could be. 00:28 RENE HAAS: Gosh, you might be right. Barcelona, 2019. My gosh, over 18 months ago. 00:32 PETE BERNARD: Yeah, that was a long time ago. Well, CES 2020 was our last... It was kind of the last hurrah for events, although going to Vegas always has its potential infection rates of all sorts of things going on there, but... Not in that case, but... Cool, yeah, no, it's good to see you again, and we've known each other for a little while and worked on some interesting projects, so it was great to have you on the show, and obviously very timely with the DevSummit coming up and some recent news that we'll talk about as well. But maybe you can give us and the listeners a little background on your journey to where you're at as President of Arm IP. 02:07 RENE HAAS: So my role at Arm is I run the IP products group. Our acronym is IPG, Intellectual Property Products Group, and that's the sales marketing development of all of our products, GPUs, CPUs, NPUs for the markets that we serve, the client market, infrastructure market, automotive autonomous and IoT. I am in the Bay Area now, but I've had a fun journey at Arm. I have spent seven years at Arm, but only a few years in the Bay Area. I was in Shanghai, China for two years, and I was in the UK for three, living in London, commuting to Cambridge. And I just came back to the Bay Area at the beginning of 2020, and... 02:50 PETE BERNARD: Are you an original California person or what's your... Where is your home base? 02:54 RENE HAAS: I'm originally from Upstate New York. Yeah, I'm, originally from Upstate New York. 02:58 PETE BERNARD: Wow, cool. 02:58 RENE HAAS: My dad was a Xerox guy, so I was a son of a Xerox guy working in... He was working in Rochester, New York, which is where I grew up. And then I came out to California in the mid-1990s, and I've been here ever since. 03:12 PETE BERNARD: I'm a New Jersey person myself, so that's something we have in common, the Tri-state area. Although Rochester is pretty far up state there. 03:21 RENE HAAS: Serious snow country. 03:22 PETE BERNARD: Serious, yes. Good, good. Excellent. So you've been at Arm for a while then, and you also spent a little bit of time at Nvidia. 03:31 RENE HAAS: I did, I did. I'm gonna pre-fetch probably your next set of questions, but before I spent... 03:37 PETE BERNARD: No pun intended. 03:38 RENE HAAS: Seven years at Arm I was with Nvidia for seven years doing a number of different roles there, but primarily in the notebook graphics space, GPUs, as well as Arm-based CPUs that went into all different types of laptops including the very first Surface that was running Windows 8 on Arm. 04:00 PETE BERNARD: Yes, those were the days. I had one of those. A lot of us up in Redmond had one of those. [chuckle] Unfortunately, not a lot of the other people had them. That was the problem. [chuckle] But, so cool. So now sort of full circle, just to touch upon that topic, Nvidia and Arm. For you, it's kind of break out the old badge, I guess... 04:20 RENE HAAS: Yeah. It's something that came live last Monday. Obviously, the rumors had been out for a number of weeks, so some people were surprised, but some people were not so surprised when it finally was announced to everyone actually last Sunday. It was supposed to be on Monday, and then we pulled it forward to Sunday. We're actually very excited about it at Arm, we think it's a really, really amazing opportunity. Nvidia is an amazing company, has done some fantastic things over the years obviously. And Arm efforts around client and data center, autonomous and such. When we think about what's going on in the next wave of computing where everything is gonna be touching something that is around artificial intelligence, I think the opportunities for the two companies to be a combined entity in this new area of computing, the opportunities are somewhat limitless. 05:17 RENE HAAS: So we're quite excited. Me, on a personal level, sometimes when these M&A things [05:21] ____ talking to the company on either side, there's a lot of questions of, "Do I know these folks? And can we really understand what their language is?" But for me, having spent equal amount of time in both places, I feel very fortunate to be in a position to be where we are on this, and it should be very exciting. And someone over there even pinged me not long after the announcement and said, "Hey, your email address is still available." So it's interesting how things circle back. 05:55 PETE BERNARD: Yeah, yeah, I wonder if you get credited those seven years at Nvidia as part of your Arm tenure. So how that works I'm not sure. 06:00 RENE HAAS: You know what, that's a really good question. I haven't... 06:03 PETE BERNARD: You might get a double hit on that one. 06:06 RENE HAAS: Yeah. In fact [06:08] ____ Pete, that was not on the FAQ. That's a good one. I'm gonna go check on that. 06:14 PETE BERNARD: Well, one of the things that's happened over the past number of years, what's been super exciting working with Arm is kind of the proliferation of where Arm is, the Arm silicon showing up. And you mentioned the early experiments, early efforts I should say, on Windows on Arm, but we had kind of a relaunch or a re-emergence of that tech a couple years ago, and I know I had the pleasure of working with you guys on that. So Windows on Arm, Windows on Snapdragon and all that stuff, it seems to be kind of a resurgence now on that as well. So what are your thoughts there? 06:50 RENE HAAS: Oh gosh. And as I mentioned, the history with working with Nvidia and Arm and Microsoft for me goes way back. And having worked on the original Surface product, that was basically what we called [07:06] ____ back in the day. And if I just think back to the value proposition we were hoping to get from those systems, it was really around extended battery life, always on, always connected, things like that. But you go back those years, there was no connectivity story, so those were just obviously purely WiFi devices. And the app story was really, really incomplete. I remember meeting with analysts early on and one of the biggest questions that I got asked when we were going to press reviews was, "Will it run iTunes?" And the answer to that question at the time was, "No." And that was a bit of a killer, if you just think about how people were getting access to music back and when these products came out. Fast-forward to now, the landscape is so different when you just think about, A, how many of our applications exist in the Cloud? B, the devices that have been introduced by third-party OEMs and as well as Microsoft. You have these amazing connectivity type of solutions that are brought forward by Snapdragon, so there's a great story in terms of connectivity. There's a great story in terms of app compatibilities on Windows 10 with everything running across. So we... 08:19 PETE BERNARD: Including iTunes, by the way. So iTunes now runs on that. 08:23 RENE HAAS: ITunes runs. And I bet you if I went through and asked that analyst and told them that iTunes ran successfully on these Windows devices, he would not care. But yeah, the experience is great. We use a lot of them inside of Arm. In fact, when I was living in the UK, I used to use it all the time on the train because the WiFi was actually spotty on the train and the cellular worked pretty good, and it was a great device to use. And not the least of which, I would literally leave my power supply back in the flat during the day. I wouldn't bring it with me, wouldn't need it. And so the devices have really, really advanced, and then there's just more great things to come. 09:04 PETE BERNARD: Yeah, fantastic. I use the Galaxy Book as my main PC and yeah, it's a game changer. When you don't have to worry about power and connectivity, all of a sudden, it's like a behavioral change in how you use a PC, so it's pretty cool stuff. And then I guess the other big thing where you're making a lot of headway with partners is in the Cloud and sort of bringing a lot of low-power. A lot of times, people think of low-power as battery life, but it's not just battery life, it's just low-power, a greener, more smarter consumption of power, overall, especially in a big data center. 09:42 RENE HAAS: Yeah, no, that's exactly right. Arm has been working on products for the data center for actually a long time. Even from back in the time when I was at Nvidia, Arm was working with early partners around SSEs for the data center and such. Like everything else, over 10 years a lot of things have changed. Confluence of a lot of work being done on the engineering side to get great products. We've gone from 32-bit to 64-bit. The performance has increased. Geometries have also gone in such a way that you've gone from 10 to seven to five nanometre type of technologies now, so you can get some really, really powerful type of processing. And then just again, like any technology trend, you need a confluence of a number of things to take place. 10:32 RENE HAAS: 10 years ago, we were thinking largely about the enterprise; we weren't thinking as much about the Cloud. And what has happened with everything moving towards the Cloud, to your point, it's put such a premium on data efficiency, on power. These Cloud data centers typically have a very, very fixed power budget and a very fixed area where they put the compute capacity. So efficiency really, really matters, it's really, really important. And we continue to innovate in this area. We've introduced some new products. Our Neoverse V1, which has scalar vector processing for HPC and high-end computing. Our N2 platform, which is 40% more efficient than our N1 platforms. And we've seen some of the large hyperscalers including AWS who have announced products based upon our N1 with their Graviton2 processor. And they've talked very publicly about a 40% power advantage at the same performance level versus the competition. So yeah, it's very real and people might think, "Oh, my gosh, it's happened overnight." And you've been in this industry a long time, you know it doesn't. 11:46 PETE BERNARD: That's right. 11:47 RENE HAAS: It's a long, long effort by a lot of partners and a lot of people inside of Arm. But yeah, now I think confluence of a lot of things in the marketplace, it's really starting to take off. 11:56 PETE BERNARD: Yeah, it's true. For a lot of things, it's a matter of the right time and the right tech and the right need for it to all come together. Actually, interesting anecdote, just to circle back to the PC discussion. We were first working on the Windows on Snapdragon PCs, we had a big beta test inside of Microsoft and we handed them out to all of our engineering managers and stuff. And we started to get bug reports that the battery meter was not working right because it was just always full. And it turned out the battery meter was working fine, it's just people weren't used to the fact that this thing would last for whatever, 20 hours. And so it was an interesting discussion with folks that that's actually how it's supposed to work. 12:38 RENE HAAS: Which is game changer, like you said. 12:41 PETE BERNARD: Yeah, yeah. So let's get to IoT. This is called the IoT Unicorn, so we might as well dig into that. Probably the real fascinating things happening on the edge, the far edge, the near edge. The definition of the edge depends on where you're standing, I guess. But Arm at the edge and things that are happening out there, what do you see as disruptions that we should be expecting beyond the incremental things getting faster and less power, but what's the view there? One of the interesting things for our listeners that aren't aware is an IP license is like pretty far up the food chain. So you get probably one of the best long-term views of what's happening in the business over the next, whatever, five years. But be curious on the IoT and edge side, where do you see things heading? 13:30 RENE HAAS: Yeah, no, it's a great question. And that area is evolving fast. Even over the last number of years, we've seen a real acceleration of activity, innovation in that space. And particularly around the area of that these edge devices are increasingly becoming small computers in of themselves. When IoT kicked off with VIGOR inside of Arm, we were talking to companies about this. It included a small microcontroller with potentially a sensor and a Bluetooth connector that could send the data back somewhere. Now you're talking about a heavy degree of compute power, you're talking about machine learning at the edge. Increasingly, we have partners who are looking to not only use our micro-controllers that have extensions for machine learning, but even tiny MPUs, tiny ML doing some level of inference at the edge. 14:24 RENE HAAS: And with that, you have a much different requirement for security because now these devices are small computers, they're dealing with a tremendous amount of data, the data needs to be protected, you need to ensure that you have an architecture that will keep the data secure. So we've done a lot of work with our partners around an innovation that we call the platform security architecture, which does a number of things. We've done a lot of work over the years around Root of Trust and things at that nature. With this platform security architecture, we actually allow for third parties to certify the devices that will essentially assure a level of data encryption and security going up the line. And with that, I think it just all feeds onto itself relative to... These are small computers, these small computers are doing more and more compute intensive tasks, they're sending more and more data through the Cloud, you then have 5g that is also adding more bandwidth and more compute capability. So what that basically means is you just start pushing from the data center to the edge, the amount of compute capacity is going up exponentially. 15:41 RENE HAAS: And I think over the next number of years, these edge devices are gonna become even more powerful and more sophisticated in terms of their capability. And you'll have a very interesting trade-off between the applications that run with that edge device at the node next to it, things that are cloud-native where the app can be running in a number of different spots. I think also you're gonna see huge innovation. And that's gonna mean certain things like autonomous entities. Not necessarily cars. Obviously cars are the most popular areas that get a lot of attention, but drones and robotics and things that can run at a much more sophisticated way, factory floor robotics, all kinds of things around managing warehousing, things of that nature. All of this is gonna become much more intelligent and much more sophisticated. 16:27 RENE HAAS: And then, back to the Nvidia/Arm potential about the edge of AI, these devices will learn, they'll get smarter. And as they get smarter, that again builds on having the compute capabilities. I know it sounds a bit of a cliche, and I've been around the industry probably to see at least a number of these waves of computing, but we're definitely into another very large one. And 5g, because of the additional bandwidth, is gonna be able to enable a lot of that. 16:55 PETE BERNARD: Yeah. I think I had this discussion with Rob Tiffany from Ericsson on the last episode or two episodes ago, but we were talking about the confluence of 5g, AI and IoT, sort of three, these... It's like peanut butter, chocolate and whatever the third thing is. But I haven't... The metaphor breaks down after that. It's like you get these ultra low latency, high performance networks combined with AI, which you could either do at the edge or the cloud or somewhere in between, with the concept of Internet of Things, which is just things connected to the Cloud and sending intelligent data back and forth and actuating in real-time. And then all of a sudden, you've got some really potential transformative scenarios there, right? 17:34 RENE HAAS: Yeah. 17:36 PETE BERNARD: And so I think... So it's sort of like... And I've had Qualcomm on the show before and other folks, and we talk about IoT being a team sport, that that statement of 5g, AI and IoT is an interesting example 'cause you need lots of different companies to come to the table to work together on behalf of a customer problem, 'cause it all starts with a customer having a problem that they need solved. And, yeah, I agree with you. You mentioned also about the fact that we're bringing AI horsepower into MCU devices or really tiny edge devices that previously were controlling a light switch are now going to be smart, and be able to learn and execute AI models. And I think that's fascinating. 18:22 RENE HAAS: Yeah. And you still have to get into... And by the way, I like that peanut butter and chocolate analogy, which are two of my favorite ingredients on [18:28] ____. You just need a third, but... 18:29 PETE BERNARD: [chuckle] Peanut butter, chocolate and more chocolate, I don't know if that's fair or not. 18:32 RENE HAAS: But similar to... One of the stories I like to talk about is a bit of what these new waves of technology enable. When we went from 3g to 4g, and I know you and I both were around for that, people were not talking about the fact that 3g to 4g was going to enable a brand new ride sharing capability, and it was gonna be able to enable people to rent their homes for vacations and such. Yet Airbnb, and Uber, and location bearing apps and things you can do on a smartphone all came through with that. I think the same thing is true for 5g and IoT. It's a little hard to completely imagine all of the possibilities that can happen. There's a lot of smart people and, as you said it, it takes a village of a combination of chip people and OEMs and software and makers to come up with a lot of ideas to advance this. But it will be there because there's such a profound shift of compute power that's gonna exist in these edge devices that is going to allow for a lot of really, really interesting potential. So it's gonna be really exciting to see. 19:37 PETE BERNARD: Let me kind of cut into one blurb here around AI Toolchain, because I believe one of the things we've done with Arm and I think should be announced for DevSummit, if not, we'll edit it out, but we've come to some agreement with you, I believe, to integrate your AI Toolchain into Azure. 19:56 RENE HAAS: Yeah. 19:56 PETE BERNARD: One of the things is around... ML Ops is a kind of a hot term, but how do you leverage a hyperscaler cloud to develop and train models and then manage those models across the edge to the cloud securely on updating these edge devices with new AI capabilities or models or trainings and tunings? And so your Toolchain's kind of at the core of a lot of that for a lot of Silicon partners, so the ability to sort of integrate that Toolchain into Azure for our customers should be a big deal, right? 20:26 RENE HAAS: Oh, it's a really, really huge opportunity. We're actually quite excited about it. We do a lot of work on the Toolchain with Compute Libraries and frameworks and different things to allow folks to develop solutions for ML at the edge, and I think we probably have as many people in our ML group doing hardware MPUs and also are doing the software libraries and frameworks. So it's really, really large. And you're reaching a brand new set of developers, if you will, and think about a Raspberry Pi or an Arduino-like platform for people who are developing things for the edge. If you can now allow those to integrate, upscale into the Azure cloud framework, because all of this tiny data becomes big data in the cloud, and then ultimately it can get serviced in such a way that end users can benefit. It's actually a really exciting thing and we've been partners with Microsoft for such a long time in a broad set of areas. I'm very excited to be involved here as well. 21:27 PETE BERNARD: Yeah, that'd be great. Hey, so DevSummit. We're on the eve of DevSummit or the day one of DevSummit. I'm not sure what the publication timeline is here, but it's a big deal. It's very exciting. Obviously this year kind of highly virtualized, but still exciting. Do you have any kind of words of wisdom if you're an attendee for DevSummit? What are some of the things you wanna look for or try to get out of? And maybe first time visitors or whoever, how do people really grok the scene? 21:57 RENE HAAS: It's a big change on a couple fronts. Obviously, first off, it's virtual. It's not live. So that's for starters. So go to your favorite search engine and search for DevSummit and you get all the details about registering and such, but we have moved it to a virtual event. For those of who are saying, "Okay, it's virtual, I get it, but I've never heard of DevSummit. Tell me what DevSummit is," DevSummit is the re-branded name of a show we used to call TechCon. And so, TechCon was the show we had every fall. And it used to be in Santa Clara for many years and we moved it to San Jose the last couple of years. So, what's new is old, what's old is new. It's the TechCon show that we're now targeting really more towards... Broadly towards developers, although I would say we think 60% of the folks who have registered are self-proclaimed or self-identified software types, versus about 40% hardware types. 22:54 RENE HAAS: We've got about 4000, 5000 people already pre-registered. We think we'll have a bit more when the time comes. It will be very broad, as Arm typically is in nature. We'll be talking about things like cloud native, chip design, autonomous vehicles. It will run the gamut of all the areas that we're involved in, relative to what it takes to integrate Arm IP and an SoC and what do you need to know about hardware libraries and partners in that space, versus everything around open source software and popular development tools and operating environments that we just talked about on the software space. There will be a lot of emphasis around autonomous, which is a pretty hot area. A lot of areas also around cloud native. You'll see the typical key notes from Simon, myself and some of the other leaders inside of Arm. I would also encourage folks to tune in because there will be some special surprise guests. I won't... 23:56 PETE BERNARD: I can imagine. 23:57 RENE HAAS: Give that away at this point of time, but it should be a very, very interesting and fun event. We have our annual Arm partner meeting every August. I think you've been to it. It's not a public event, it's an NDA event. But I bring that up just in the context of... We've had one rodeo with doing this thing virtually. So I'd like to think we've got some good practice in terms of things that... The dos and don't-'s in terms of doing something from a virtual standpoint. But yeah, it should be very, very good. We're looking forward to it. 24:26 PETE BERNARD: Cool. Yeah, it's interesting, Microsoft's done a number of events now virtual and I don't think we published the data but my understanding is the engagement we get because it's virtualized, we get so much broader engagement, we get so many more people quote, unquote, "attending" and engaged in the content than you would if it was only a... You had to get on a plane and go somewhere. So I think one of the nice by-products, if there is a nice by-product out of all this craziness, is we are all building more muscle about how to enable people to be more engaged regardless of where they are. And especially when you talk about developers, developers everywhere in the world and there should be. And now to be able to enable them to plug in and get educated and learn some new things, that's a fantastic by-product. 25:13 RENE HAAS: Yeah, yeah. No, you're completely right. We'd love to do these events live versus virtual, but when I think about the size of the developer community that exists... Arm is a fairly broad platform, as you know, and it would be really hard to figure out events that could bring all the potential developers who work on Arm... And it's all over the place. There are apps developers, there are kernel developers, there are people who do open source software, it's a broad, broad community. So we're actually kind of excited to do this thing virtually. It'll be a bit of a lab test to see how that works in terms of reaching the development community in a virtual way, but we're looking forward to it. 25:53 PETE BERNARD: Cool, awesome. Well, lots of stuff going on at Arm these days. And so it was great again to connect, Rene. I think hopefully we'll keep in touch here as things transform into Nvidia landscape. Maybe you'll get those extra years on your seniority. [chuckle] But that would be great. 26:14 RENE HAAS: I should get some credit somehow for that. I am going to talk to Jensen about that next time I have our consultation with him. 26:21 PETE BERNARD: Yeah. Cool. Well, good. Any last closing thoughts? It sounds like we've really covered [26:27] ____ here today. 26:29 RENE HAAS: [26:29] ____ I appreciate it and [26:29] ____ as I mentioned, I was listening to some of the podcasts you had done prior and I really enjoyed them and I'm very, very honored on behalf of Arm to join you and be part of what you're building here. It's really cool. 26:43 PETE BERNARD: Sounds good. Alright, Rene. Well, take care and I'm sure our paths will cross again. 26:48 RENE HAAS: Alright, great. Thanks. 26:50 PETE BERNARD: Alright, take care. Thanks.