Not sure if it has been posted, below is the Spotlight Session by Patrick Little, CEO & Chairman,
@SiFive TOPIC: "The Future of RISC-V " on Day 2 of
@SixFiveSummit. Published on 07/06/2022. I have also attached the transcript. Some interesting comments regards commercial traction as highlighted in the transcript.
0:11 Hello, and welcome back to "The Six Five Summit".
0:13 I'm Shelly Kramer, one of the founding partners
0:15 and a principal analyst here, at Futurum Research,
0:17 and on behalf of my team at Futurum
0:19 and the team at Moor Insights & Strategy, welcome.
0:21 We're glad to have you.
0:23 In this spotlight session, Futurum's Daniel Newman sits down
0:26 with SiFive CEO and Chairman, Patrick Little,
0:29 for a conversation about all the buzz surrounding RISC-V,
0:33 and what the future holds for this exciting technology.
0:36 Let's dive in.
0:37 We want to thank Accenture
0:39 for sponsoring the Semiconductor Track
0:42 of this year's Six Five Summit.
0:53 Patrick Little, CEO, SiFive,
0:57 welcome to the 2022 Six Five Summit.
1:00 So excited to have you here.
1:02 Yeah, really excited to be here, Daniel.
1:03 Looking forward to it. Yeah, it's great
1:06 to have the opportunity to, you know, sit down,
1:09 have a conversation.
1:11 So much going on in the semiconductor space.
1:13 It's a really interesting time in the market
1:16 and the economy.
1:17 So, I'm gonna pick your brain about a lot of things.
1:18 Hopefully, everyone out there in the audience can walk away
1:23 knowing a little bit more about SiFive,
1:25 a little bit more about RISC-V,
1:26 and maybe even learn a little bit more
1:28 about what's going on in the world of technology
1:30 and how semiconductors are continuing to change the world,
1:34 or as I like to say, eat the world.
1:36 That type of things, right? Absolutely.
1:39 So, let's start with RISC-V.
1:43 Hearing a lot of buzz about it.
1:44 We've been covering it, talking about it,
1:47 but it's not necessarily household name yet.
1:51 So for those in the audience that are, you know,
1:55 hearing about RISC-V, or maybe you've heard about it
1:57 but don't really know much about it,
1:59 how do you explain, define, introduce RISC-V
2:02 to that audience?
2:03 Yeah, I only do this like 20 times a day.
2:05 So, happy to kind of, lay the foundation.
2:08 At first, I have to giggle what you said
2:10 about the changes in tech.
2:11 You know, 10 years ago, if you would've mention hardware,
2:14 there'd be a mad rush for the door,
2:16 and now it seems like hardware's really kind of,
2:17 having a renaissance,
2:19 and this computing supercycle is just profoundly strong.
2:22 It'll outlive all of us, but RISC-V.
2:25 So, let me take you back, I don't know,
2:28 when Dave Patterson first started,
2:30 even coined the term RISC,
2:32 and really started working
2:33 on reduced instruction set computer's architectures
2:38 at Berkeley.
2:39 In fact, when I was going through college,
2:40 I used Dave's textbooks.
2:43 And so, I kind of view him
2:44 is the father of the entire thing,
2:49 the entire RISC architecture.
2:51 But then you kind of fast forward to about 2010
2:54 and we have Krste Asanović, who's a professor at Berkeley,
2:58 who has a research program and he needs to solve...
3:01 You know, he's looking
3:02 for a new instruction set architecture
3:04 that gives him and his team the ability
3:07 to collaborate across projects,
3:09 to collaborate even across universities,
3:12 but something that is clean slate, modern, scalable,
3:16 and he couldn't find anything out there.
3:18 And it was not only the commercial pieces
3:20 of not being able to access the technology free
3:22 and openly and collaborate openly,
3:24 but it was also just the tech itself.
3:27 He wanted something small.
3:28 He and his team wanted something small and very modular,
3:31 but also scalable through extensions.
3:33 And so, they sat down for a brief summer session
3:37 to try to really design something
3:39 and they came out with the RISC-V architecture,
3:42 which I will tell you, is gonna have profound repercussions
3:45 across this industry for not five years or 10 years,
3:49 but 50 years.
3:50 And so, the idea behind it was,
3:53 let's design something that is open for collaboration
3:56 across academia, not just at Berkeley,
3:58 but across all of academia initially,
4:00 but something where anyone can pick it up
4:03 without an upfront fee, without the encumbrances up front,
4:06 and work with it and scale it
4:08 and devise their own computing architecture
4:11 around the RISC-V core.
4:12 And that was kind of the essence of it,
4:14 but really, at the very heart of it, collaboration.
4:17 And so they were quite successful in doing that.
4:19 And some five years later,
4:21 it started to go into not only other universities,
4:24 but other countries, very quickly spanned into India
4:27 and China and Europe and many other territories.
4:30 And so, that what he found is that by 2015,
4:33 the entire architecture was put into something
4:36 called the RISC-V Foundation
4:38 to ensure that it remained open and accessible
4:41 to really everyone.
4:42 And so, all of that IPL that was generated
4:44 at Berkeley and beyond was donated to the RISC-V Foundation,
4:48 which now has become RISC-V International.
4:51 And another thing occurred kind of,
4:53 coincident to that in 2015 is
4:55 that team from Berkeley went often started a new company
4:58 because they saw the commercial aspects of the technology,
5:02 really with the same fundamental principles in mind,
5:05 collaboration across partnerships,
5:07 collaboration across a community of participants,
5:11 and also collaboration with customers.
5:13 And so, that was kind of,
5:15 the nucleus in the starting of SiFive back in 2015.
5:19 But I think it began with David Patterson in 1980,
5:23 and it just progressively RISC-I,
5:25 all the way through RISC-V, gotten stronger and stronger.
5:29 And I think, it sets a brand new model in computing
5:33 where things are open and available,
5:35 and it just makes it a lot easier for smaller companies
5:37 or universities, or now very large companies
5:40 to innovate around the computer architecture.
5:43 So I think it's a, not only a new architecture,
5:46 and there have been many,
5:47 but I think it's a new model for collaboration
5:50 on computing platforms.
5:52 Most of the computing platforms
5:54 or the instruction set architectures that have been here
5:57 through history have been but from a single company,
6:00 and so most the failures of those have been
6:03 that sponsor company failing financially.
6:06 And so, I think, what a lot of our customers see is
6:08 since the community now owns this,
6:10 literally thousands of companies, universities,
6:14 software companies, you name the type of tech company,
6:17 are collaborating, it's protected.
6:20 And I wanna say, recently we closed a very large design win
6:24 with a very well known space and aeronautical company,
6:29 and one of their primary reasons
6:30 for choosing it is they felt that this is a standard
6:33 that's gonna last for decades and decades into the future,
6:36 and that the software they write today will be scalable
6:39 into the future, and this platform will continue
6:41 to have strength and existence over the coming decades.
6:44 And so, that's what started in Berkeley some decades ago
6:47 and primarily 12 years ago for RISC-V,
6:50 and incredible legs on it.
6:52 Frankly, I'm very proud to be part of it
6:54 because it's something that's going to be going on.
6:56 My children will be learning RISC-V in college,
6:59 like many do, and so it's great to be part of something
7:02 that's gonna outlive all of us.
7:04 But that's kind of the nucleus and the intent
7:05 behind the RISC-V architecture.
7:08 Absolutely, and you touched on a lot of things there,
7:10 and I'm gonna want dig in a little bit more
7:11 into some of the, you know, differentiations
7:14 and the verticals in which you're serving.
7:16 We'll hit that in a minute.
7:16 But SiFive, so one of the well known participants,
7:23 you know, in the RISC-V arena
7:26 but you started to kind of allude a little bit to the,
7:29 you know, the finding of the company
7:31 and what was the basis of SiFive.
7:34 But give us just a little more on that
7:36 because, you know, I'm interested.
7:38 You know, talk about the mission of the company.
7:41 You mentioned all these different participants.
7:44 What's the SiFive story?
7:46 What is your role to play?
7:4 Yeah, great question, Dan.
7:49 So, I think first and foremost,
7:51 our role, I believe, is the brand standard.
7:54 So, the same team that founded RISC-V
7:56 at Berkeley is the team that founded SiFive, as I mentioned.
7:59 And so we feel a responsibility
8:01 to keep the virtuous essence of the architecture around,
8:06 for the collaboration, for the scalability,
8:09 and to try to protect and grow it.
8:11 And not only as SiFive, but also as a community member
8:14 in the RISC-V Foundation.
8:16 So first and foremost, we feel kind of responsibility
8:19 to extend the seed that began at Berkeley some 12 years ago.
8:24 And then the rest of our mission today,
8:26 it's gonna sound a little corny,
8:28 but our mission today is all wrapped
8:29 around the success of our customers.
8:31 When we talk about success,
8:32 we don't talk about a design win or a product release,
8:35 we talk about what can we do to perpetuate the success
8:39 of our customers to deliver best in class products
8:43 in their end markets.
8:44 And so, that's what we're focused on.
8:45 We're very right to left.
8:47 If you walk the halls of SiFive,
8:48 you'll hear the term "right to left" an awful lot.
8:51 And so, we believe that if our customers...
8:54 We want RISC-V and SiFive to do amazing things this year
8:58 and over the decades.
8:59 To do that, we really need to empower our customers success.
9:02 So we focus on that.
9:04 Their success always translate,
9:06 the transitivity from their success to our success,
9:08 and frankly, down to the success of RISC-V is highly linear.
9:13 And so, that's what we do is we focus on collaborating
9:16 with partners, working with them to drive their innovations,
9:19 frankly, in almost any market and application you can name,
9:23 and drive that innovation all the way through SiFive
9:26 and back into the standard.
9:27 And I think, this is one of the reasons
9:29 why the standard will last for many decades.
9:32 The architecture was well defined in the RISC-V standard,
9:37 but not the implementation,
9:38 and so, the micro architecture
9:39 or the implementation of the standard is
9:42 left open intentionally for interpretation.
9:45 And so, one of the things we do at SiFive is we interpret
9:48 that in a way that benefits our customers.
9:51 And so, one of the areas
9:53 that I think, that we're very unique
9:55 and the RISC-V standard is very unique is
9:57 that we always knew that it was gonna be clean.
10:00 We always knew that it was gonna be modular and scalable
10:03 and therefore, tight and efficient.
10:05 But one of the things we're really finding out...
10:07 You mentioned, how we differentiate ourselves.
10:10 One of the things that we're finding out now
10:12 that we're actually taking this
10:13 into real world applications is
10:15 that our performance per watt is profoundly better
10:18 than anything out there,
10:20 and it has to do with the focus from the very beginning
10:22 and the cleanliness of the architecture.
10:24 And so, we have,...
10:26 And, I mean anything from
Edge AI
10:28 all the way to the cloud implementations,
10:30 our customers are coming back to us saying,
10:32 "I don't know if you guys really know this,
10:34 in the lab and in our products, we're starting
10:36 to see some outrageous performance per watt advantage."
10:40 And at the
Edge, it's very obvious
10:42 that customers would need that power efficiency,
10:46 but now, Daniel, when you talk to any customer
10:48 in any vertical, any application,
10:50 they're trying to get the maximum performance they can
10:53 in a power envelope, very obvious in mobile,
10:56 first order obvious in mobile,
10:57 but same thing for automotive
10:59 and same thing for the data center
11:00 and many, many other applications.D
11:02 So, in addition to the openness of the architecture,
11:05 which we believe is foundationally unique for this ISA,
11:10 we believe that on its own right,
11:12 technically speaking, it's super low power,
11:15 super area efficient.
11:17 And then I now think, another thing that's very important is
11:19 it's highly scalable from the smallest micro controller
11:23 to the largest, very parallel processing monster
11:26 that we design and one set of software works across
11:31 that entire continuum.
11:32 So, one set of software binaries can work
11:35 on any RISC-V machine that's compatible,
11:38 and that's whether it's from SiFive or another vendor.
11:40 And so, these are some of the attributes that I think,
11:43 do differentiate SiFive, and we do view ourselves
11:46 as the brand standard and the founder of RISC-V,
11:48 but we also think that the value that this ISAs bring
11:52 to our customers is profoundly different
11:55 and differentiated and very powerful,
11:57 and this is why I think we'll be around for decades to come.
12:01 And then I should say something about the culture,
12:02 it's wacky, super...
12:04 A lot of Birkenstocks, as you might guess,
12:07 but great group of people, incredibly humble,
12:11 incredibly focused on not this year or this quarter,
12:15 but on really bringing this ISA and changing the industry,
12:19 and driving low power and collaboration.
12:22 You know, community collaboration is top of mind for us.
12:24 Every time we think about something for our company,
12:26 we think about, well, how can we also move the community?
12:30 So it's wonderful mission that we're on.D
12:32 And that's why, you see, we're starting
12:34 to really build our engineering team super quickly,
12:37 and we're getting a lot of very excited candidates knocking
12:40 on that virtual front door.
12:41 So, Patrick, I think you can read my mind a little bit,
12:43 'cause I was gonna hit you up.
12:44 I was gonna say, "Hey, talk a little bit
12:45 about the differentiation", but you kind of nailed it,
12:48 talking about, you know, everything from low power
12:50 to open and, you know, you mentioned several other.
12:54 So, I'm gonna kind of jump ahead a little bit,
12:56 'cause I'm gonna talk a little bit
12:57 about the traction in the market,
12:59 because I think that's a lot of time
13:00 the rubber meets the road, proverbially speaking.
13:05 Right. Right?
13:05 When customers start to utilize and buy.
13:07 Now again, you go from...
13:09 You mentioned Berkeley and you mentioned Birkenstocks.
13:12 I did often wonder if there was any relationship.
13:14 I don't think there, there actually is,
13:16 but if you walk around Berkeley,
13:18 you'll also see a lot of Birkenstocks, so.
13:20 By the way, that was just a little interruptive humor
13:23 to the interview,
13:24 (Patrick chuckles)
13:25 if anybody out there just need to take a break
13:26 from all this technical content.
13:29 I'll throw that one to the team
13:30 and we'll see what they come back with,
13:32 but, I think, with that correlation.
13:34 Yeah, throw that out there, let me know,
13:36 where were Birkenstock invented?
13:38 Maybe that'll be the title of this session.
13:40 No, let's keep focus.
13:42 Patrick, don't let me go too far off.
13:46 You did mention when we talked a bunch of applications
13:48 though, you talked about the
edge a few times,
13:50 you talked about the data center.
13:53 So these are, you know,
Edge, data center, AI, automotive,
13:56 these are a number of the areas
13:58 where these new architectures are being looked at,
14:01 new competition entering markets.
14:03 You're seeing hyperscalers are building home grown.
14:05 You guys are obviously identifying opportunities
14:10 in a big market and a big market opportunity.
14:13 Where are you seeing SiFive
14:15 and RISC-V gain commercial traction?
14:19 Right, great question.
14:20 So, the company actually had two phases.
14:23 The first phase was to prove out the architecture,
14:26 build the brand and we did that
14:28 through a series of very scalable embedded products
14:33 in very established markets.
14:34 And so, we did that for the first four to five years
14:38 of the company.
14:38 We offered the very smallest course
14:40 and the very largest embedded course,
14:42 but because we focused on embedded
14:44 our application applicability was extremely broad.
14:47 We were in 5G infrastructure.
14:49 We were in mobile handsets.
14:51 We were in wifi gear.
14:54 We were in networking.
14:56 If you name it, we were in those applications,
14:59 over hundred customers,
15:00 hundreds of different design wins and applications.
15:03 And so, the early adoption
15:05 of the company's tech was very, very broad,
15:08 and frankly, really enabled the company
15:11 to make a name for itself
15:12 and prove that we could deliver the quality
15:14 at a commercial and enterprise level.
15:16 And so, we proved ourselves in those early days.
15:19 Since coming on board, about two years ago,
15:22 we pivoted to higher performance applications.
15:26 And so, our objective is,
15:28 if you can get it from any other vendor
15:29 at any level of performance,
15:31 you'll be able to get that same thing from SiFive.
15:34 We're probably about a year and a half away
15:35 from that intercept point
15:37 where you can look across all of the existing ISAs
15:40 and say, SiFive has best in class raw performance
15:43 in every category.
15:45 But what we look at now is the applications
15:48 that we're servicing now are still quite broad
15:51 because the nature of the ISA.
15:52 We are in quite a few of the Edge AI solutions
15:56 with our vector products.
15:58 For example, we have one customer
16:00 who is building security cameras,
16:04 and it is originally an embedded device that was IoT,
16:07 so it was connected, but it kept sending you notifications
16:10 to your phone when the trees blew.
16:11 And so what they wanted,
16:12 and this is a very well known company in high volume,
16:15 what they wanted was some intelligence at the Edge,
16:18 so where you don't get a notification at work
16:20 when trees are blowing,
16:21 but only when someone with a black mask is
16:23 at your front gate.
16:24 And so, that's one example of something very small
16:27 that's battery powered that we're servicing
16:30 with our new products.
16:32 Another example was we also have
16:34 a brand new data center customer
16:36 who's developing multi-core, multi-cluster,
16:39 extremely wide compute solutions,
16:43 and we're sitting alongside in cooperation
16:46 with their computing pieces.
16:48 And so, you really do see this very broad applicability
16:51 of the RISC-V or the SiFive architecture.
16:55 And so, we're enjoying all of the applicability
16:57 across the board.
16:59 We do have some areas of focus,
17:00 I think, in the immediate term
17:01 that make a me lot of sense for us
17:03 because we feel like they're underserved markets
17:06 in very exciting applications.
17:08 One of them in particular is automotive.
17:11 I'm from automotive,
17:12 I recently left Qualcomm where I ran automotive
17:15 to come here, and so those applications,
17:18 those customers, that industry is close to my heart.
17:21 And so, there's a lot of intelligence going
17:24 into those designs, there's a lot of compute going
17:26 into the car where there wasn't so much even 10 years ago.
17:29 And so, we're partnering quite publicly
17:32 with quite a few of the automotive customers.
17:34 We feel like we're a great fit.
17:35 They love our outright performance.
17:37 They love the openness of the architecture.
17:41 So that's one category.
17:43 Another category where I feel like
17:44 we're really getting a sharp rising
Edge on the uptake is
17:48 around vectors, and so, those customers that wanna add AI
17:52 to their existing solutions,
17:54 and so, a lot of them will need vector processing
17:57 alongside of their kind of, traditional scaler processing.
18:00 And so, what we're finding is many customers don't want
18:03 to be captive to an existing proprietary solution
18:07 in a new category.
18:08 So, if they're moving to a new category,
18:09 they're thinking, "Everything's wide open.
18:12 The software's not really established yet.
18:14 The ecosystem's not really established yet.
18:16 Let's go with an architecture that will take us 20 years
18:19 in the future where we won't be captive,
18:21 where we'll be open to innovate with a partner,
18:24 without a partner."
18:27 And I think Intel is a perfect example
18:29 of leadership and maturity.
18:32 They are a partner of ours.
18:33 They're a customer of ours.
18:35 Many people might look at that and say,
18:36 "Why wouldn't you be natural competitors?",
18:38 but the truth is they're one of our best customers
18:40 and one of our best collaborators,
18:42 because they look at the x86 architecture,
18:44 then they look at the RISC-V architecture with SiFive,
18:47 and they see an ability to achieve different things
18:50 at different design points.
18:51 And so, they're an investor in our company.
18:54 They recently joined RISC-V International,
18:57 which I feel is a historic moment
18:59 that was overlooked by a lot of people.
19:02 Everyone should have turned their head and said,
19:03 "Wow, the maturity there is pretty phenomenal."
19:07 And I think that Intel's joining RISC-V International is
19:10 really a tipping point for the entire foundation
19:13 and entire ecosystem, because they're doing so much,
19:16 they're bringing so much tech in cooperation with us
19:19 and the rest of the foundation.
19:20 It's pretty amazing what's happened there
19:23 And I think that we'll end up doing some phenomenal designs,
19:25 you know, hybrid designs with them.
19:28 Yeah, it's a little bit of a sneak peek into the future.
19:30 And, you know, we have a minute or two left
19:32 in this conversation, Patrick,
19:34 and I, you know, wanna thank you
19:36 because it's been great insights.
19:37 I think anyone out there that didn't know about RISC-V
19:40 or wasn't overly familiar, has to know quite a bit more now.
19:44 So that's 20 minutes well spent.
19:45 As I like to say, it's a Ted talk for RISC-V.
19:49 But you say something,
19:50 or I've heard you say something in the past
19:53 along the lines of "RISC-V has no limits".
19:57 Bold, it's a bold statement.
19:59 I mean, do you believe that?
20:00 What do you mean by that?
20:02 And you know, what does that mean
20:03 for the future of SiFive and RISC-V?
20:07 Yeah, the reason why I say that is,
20:10 first and foremost, we're living
20:12 in the computing supercycle.
20:13 So, I mean, just timing is everything.
20:16 We're so lucky with our backgrounds
20:19 to work into an envelope in time
20:22 where computing is so profoundly important to the world,
20:25 and this is from medical to communications to data center,
20:29 so important to the world.
20:30 And so, when I look with objective eyes
20:34 at the RISC-V architecture and the difference it can make
20:37 in the trends that are going on in the world,
20:39 I see such a incredibly good match.
20:43 And then when I layer on top of it,
20:44 kind of, back to the virtue of
20:46 why this thing was even started in the first place,
20:49 let's make it open, let's make it clean,
20:51 let's make it scalable, and most importantly,
20:54 let's make it collaborative where the best idea wins,
20:58 whether it's Intel's or ours
21:00 or somebody else on the RISC- V,
21:01 or even a competitor's idea if they put it
21:04 to the RISC- V International pot of ideas.
21:07 I just think that it's the computing supercycle
21:11 really taking this world to a whole new level,
21:14 and RISC-V kind of, coming off the on ramp
21:16 in the perfect time to be able to deliver a lot
21:20 of the attributes that are necessary.
21:22 And so, we're seeing it live.
21:24 It's incredibly exciting at the company.
21:26 We're in handsets, we're in 5G infrastructure,
21:29 we're in art, we're in cars, we're going into data centers,
21:32 we'll soon be in client computing.
21:34 We have not seen anything about the architecture
21:36 that would limit it from servicing with very high impact
21:40 into almost any application.
21:42 And so, it's super excited about...
21:44 I'm personally excited about being part of SiFive
21:46 and incredibly excited about just being part
21:48 of the RISC-V movement.
21:50 It will make a difference in this industry
21:53 that's gonna outlast both of us,
21:55 and we'll be standing on the hill next to our customers
21:57 and next to our partners, and next to academia,
21:59 'cause we'll all be collaborating.
22:02 Super exciting times.
22:04 Well, you totally spoiled my self belief
22:06 that I'm gonna live forever there,
22:07 now that I know this is gonna outlast me.
22:10 But in all serious, if you've been out there,
22:13 you can read some of the analysis that I've written
22:15 after some of the recent moves
22:17 that you've made, Patrick, as CEO of SiFive,
22:20 and I definitely see a lot of potential.
22:23 We are definitely in that supercycle,
22:25 as I often say and continue to say,
22:28 that semiconductors will change and eat the world
22:31 because you can't run all this software on nothing.
22:33 It can't run on air.
22:35 It's gotta be built on a world of chips,
22:38 and if the shortages that we faced
22:40 during this pandemic didn't make people just acutely aware
22:44 of how important semiconductors are,
22:46 then I don't know what would.
22:47 But Patrick Little, SiFive, congratulations.
22:51 Thank you so much for joining me
22:53 at this year's Six Five Summit.
22:55 We can't wait to hear more,
22:56 to watch and see how SiFive and RISC-V continue
23:00 to change all of the areas in which you're participating.
23:04 So we'll see soon, Patrick. Thanks, Dan.
23:06 Exciting times ahead and future is bright.
23:08 Appreciate it. Enjoyed it.