On this episode of The Six Five – On The Road, hosts Daniel Newman and Patrick Moorhead welcome Lenovo’s Scott Tease, Vice President, GM High Performance Computing and AI, and Imperial College London’s Andrew Richards, Director of Research Computing Services for a conversation on Lenovo’s partner ecosystem, HPC and AI research during SC23 in Denver, Colorado.
Their discussion covers:
- Lenovo’s vast partner ecosystem in the public sector and the goal for the industry that drives these collaborations
- The current collaboration between Lenovo and Imperial College London
- An overview of Lenovo Neptune™ Water Cooling Technology and how it can help organizations reach their own sustainability goals
- How this technology will support High Performance Computing (HPC) and AI researchers in the future, given the tremendous potential impact and focus on empowering future generations of researchers and students
Be sure to subscribe to The Six Five Webcast, so you never miss an episode.
Watch the video here:
Or listen to the audio here:
Disclaimer: The Six Five webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded, and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors, and we ask that you do not treat us as such.
Patrick Moorhead: The Six Five is on the road in Denver, Colorado. It’s Supercomputing 2023. We are in the Lenovo booth talking about supercomputing. We’re talking about flops to–
Daniel Newman: To tops.
Patrick Moorhead: There we go.
Daniel Newman: Flops to tops.
Patrick Moorhead: We almost got that right here. I got to tell you, I’m going to give credit to Daniel even though it’s very hard here. But the amount of technology and the amount of innovation is incredible here. I think the last Supercomputing I went to was five or six years ago, and it is just bursting from the seams. In fact, it was hard to get in. There were 10,000 people I think waiting to see the great booths everywhere. The drinks, the food are maybe a combination of all three.
Daniel Newman: I don’t know, AI is hot, it is red-hot. And the feeling of excitement of energy that’s going on around the event, it’s palpable. And we are nerds, so we can say this in good company, but this event used to be kind of a place for us. And now it feels like AI is cool. So tell all those athletes to go home, the real athletes are here making world changing technologies. And Pat, I couldn’t be more excited to be here.
Patrick Moorhead: Probably one infrastructure vendor that has really made a run here and a name for themselves is Lenovo. And it’s pretty incredible. And the only thing better than talking to Lenovo is talking to Lenovo and one of their customers because we always say customers are that grand purifier, right? They align analyst with vendor, with the customer. So let’s see where this goes. Scott, great to see you again.
Scott Tease: Nice to See you too. Thanks guys for being here with us.
Patrick Moorhead: Absolutely. Andrew, welcome to the Six-Five, first time and hopefully not the last time.
Andrew Richards: I hope so.
Daniel Newman: Yeah. Scott, you are an alumnus. You are back. Welcome to the show, great to have you here. Let’s start out talking a little bit about the vast public sector ecosystem at Lenovo. What’s the driving force behind the investment and why is this part of the business growing so quickly?
Scott Tease: We’ve been committed to helping our customers solve humanity’s greatest challenges for a long time. We’ve been talking about that for a long time, Pat.
Patrick Moorhead: Yes, very consistent.
Scott Tease: Very consistent.
Patrick Moorhead: Consistency is good.
Scott Tease: Yeah, it is. Agreed. So where does a lot of that solving humanity’s greatest challenges happen? It happens out in the public sector at our big HPC sites, at places like Imperial. So we’re committed to try to provide the best possible technology to those users so they can go solve those problems, make our lives better, make our healthcare better, understand the plan, and understand the atmosphere, the climate, what’s going to happen in five years. So we’re committed to that and partnering with people like Imperial and the rest of our customers around the world. We are the number one provider of supercomputers in the top 500. The reason I’m proud of that is not the number of listings, but it means we’re putting HPC in the hands of more users than anyone else out there and that’s what we’re proud of.
Patrick Moorhead: So Andrew, can you talk a little bit about what you do and what Imperial College does, but also talk a little bit about the engagement that you have with Lenovo?
Andrew Richards: Sure. I’m Director of Research Computing at Imperial College in London. We are a very STEM-heavy university. We have a business school as well, but we primarily focus on the STEM subjects and we have users across the whole university who are big HPC users who need the services that we provide to underpin their research activities. So in order to do that, it’s key that we can provide platforms, systems that are really cutting edge, taking things forward. Last year we partnered with Lenovo so that we can go on this journey, not just to providing next-generation platforms across all the research areas, but also to adopt sustainable technology to do that as well.
Patrick Moorhead: I love it. So the best bits are kept inside of the college, or do you give access outside to all this supercomputing goodness?
Andrew Richards: So primarily it’s just for the university. It’s within the college. We do have collaborations. Academia by its very nature is collaborative. So we’re collaborating with other universities with industry and you will have people who will use that but first and foremost it’s for the university.
Daniel Newman: Yeah, I remember in the early 2020 during the really rough period of the P word, I don’t know if we can say it on YouTube or what will happen, but I do remember actually reading a lot of studies. I imagine it was from Imperial.
Andrew Richards: That’s right.
Daniel Newman: You had a very well, well-regarded, respected model that was being shared, that was trying to understand the likely impact of what was going to happen. So I remember Imperial, your data, your computing was behind that. And so it was very interesting to watch Academia supercomputing. And for us that was really a horrible time, but it was very interesting to see what you said about collaboration, tech companies, universities, national laboratories, working together so profoundly to basically try to help people understand what was going on.
Andrew Richards: Absolutely, and our role in central research computing is to help people do their research without having to think about the technology, as much as we have big central systems, there are lots of other systems within the university as well. And part of this partnership with Lenovo now is to help us build new systems in new data centres using the best water cooling technology so that we can bring other people in the university into using this technology for all that research as well.
Daniel Newman: Let me double down on that. You started talking about Neptune and the technology. Imperial has sustainability ambitions. Lenovo of course wants to offer greater levels of sustainability. Talk about the process of learning more about the Neptune technology, how it helped you meet your needs, and then maybe you can talk a little bit about how Neptune is evolving as well more broadly, but with you Andrew.
Andrew Richards: So we reached a point last year where we were at the limits of what we could do with our existing data centre, with our existing systems. So we wanted to work with somebody who could take us on this water-cool journey. Imperial has an ambition to be net-zero by 2040. So how do we help with that? How do we go on that journey? It’s not just that the technology providers are pushing us in this direction because we can’t do things air-cooled anymore. We need to do water cooling. We want to do that so we can get access to the best technology that we need. So it was really about, in order to do that, we need to get all the various pieces in place. We need a data centre, we need a water cooling technology. We need to work with a partner who understands what we need to do so that we can have sensible conversations with data centre providers so we can put it all together and then build the systems for the next five, 10 years that Imperial needs.
Patrick Moorhead: Scott.
Scott Tease: Very cool.
Patrick Moorhead: Yeah, give us a little broad here.
Scott Tease: Can you believe we’ve been doing water cooling for over 10 years? We shipped our very first one back in 2012. It was 9,700 units of warm water cold supercomputing back when the thought of mixing computing and water was kind of like everybody was freaking out. So 10 years on. I’m telling you, man, 10 years on, we have got it going all over the world. Love seeing it go into Imperial. They’re one of our newer customers on it, but it’s all over the planet. And the reason for it is multifold. One is you need water to unlock these new high-end GPUs. The amount of power they consume means there’s going to be a lot of heat given off as they do their job. Dealing with that heat is not easy liquid cooling, makes it far more easy. Liquid’s really good at extracting heat out of the systems. So we’re going to save money on our power bill, but as Andy said, it’s also going to have a direct sustainability impact because the less power you use, the less CO2 we’re going to admit into the environment, the less water we’re going to use in the data centre. We’re not having to chill, we’re not having to do evaporative cooling. It just keeps on unstacking on itself. That’s why it’s, again, it’s growing everywhere around the globe.
Patrick Moorhead: Yeah, industry analysts need to be very careful about when they see firsts and things like that. But Lenovo and before that, IBM were clearly first in this technology and what’s funny is everybody’s trying to get on the bandwagon. But what I have found in talking to end customers is that every vendor solutions are not the same. I’ve heard talks about leaks and quite frankly, leaks and semiconductors and electricity is not good and there is a science to water cooling. You can’t just check the box and make it happen. And I’m really excited about what the future holds and look forward to Lenovo upping their game now that it’s right. People are jumping into this left and right.
Scott Tease: People want us to hear about it. It’s beautiful. It’s great because more people never are asking about liquid. The robustness of the design is one of the things that I think makes us different. The other thing is we’re using warm water. That is a real differentiator. No one else is driving that agenda. We don’t ever want to have to chill that water and spend the power to chill it. We also want to recycle the heat that comes out of that system. You got water coming from these systems, it’s over 50 degrees Celsius. It’s about 120 degrees Fahrenheit, something like that. We can recycle that heat. There’s value in that energy. So we’re doing that and we’re trying to do a hundred percent heat capture. That’s the other thing that makes it a little bit unique. So really exciting.
Patrick Moorhead: So, Andrew, I wanted to shift gears just a bit. I mean, for decades it was always about the most flops and then it was flops per energy. And then this AI thing comes in and it’s not necessarily a driver of flops. It’s a different type
Daniel Newman: Tops.
Patrick Moorhead: Tops. Excuse me.
Daniel Newman: Flops to tops.
Patrick Moorhead: Flops to tops. Yes. By the way, all credit goes to Daniel here.
Daniel Newman: I figured that out today. I love it.
Patrick Moorhead: But I wanted to ask you, what do those workloads look like? How are you enabling both high performance computing, classic workloads, let’s say governed by flops and these AI workloads?
Andrew Richards: So It’s a very good question. I mean, to first order, we have people who are using our traditional HPC systems to do AI to do some of the work that they want. But increasingly what we’re seeing is people want access to more interactive platforms. They wanted a different approach to doing the work that they want to do. So we have a big initiative in Imperial called I-X. It used to be called Imperial X. It was shortened to just I-X, and it’s all about AI and machine learning and that community is really helping us to think about what we do in providing services because they want things that are more interactive. They want access to the GPUs, they want access to other technologies as well, but they wanted in a more interactive way. So we’re having to rethink how we provide services to perhaps be a little bit more cloud life on premise to do that.
Patrick Moorhead: That’s cool. What is your thinking when it comes to maybe GPUs or ASICS, or does it even matter at this point?
Andrew Richards: So at the moment, a lot of people are wedded to using GPUs because of the whole ecosystem around GPUs. For a while, people have got used to using that technology. So that’s the kind of safe technology to go to for a lot of people, they kind of feel comfortable knowing how it works.
Patrick Moorhead: Well, very programmable, not as programmable as A CPU, but more programmable than an ASIC or an FPGA.
Andrew Richards: Exactly. But increasingly, we do have some areas where people are looking at what can we do with things like FPGAs. So we’ve got people right now asking us for access to FPGAs. We’re looking at things like DPU as well, how we use those and we want to bring other things into the ecosystem, but underpinning all this is we need a data centre, water cooling, et cetera, to bring the platforms into deliver it.
Patrick Moorhead: I love to hear that you’re exploring all those options. Quite frankly, everybody is, because the great part of ramping up the performance curve is we can do so many other things, but boy does it take a lot of electricity and a lot of resources and need a lot of availability of these things that I always like to say, given enough time, most things get sucked into under the main processing unit. I mean, whether it was the floating point unit vector instructions, I mean, it’s just a matter of time. And that’s never to say that GPUs whatever go away. It just says there’s going to be something that’s going to layer on top to keep delivering this. But I’m really glad to see you’re researching it. I shouldn’t be surprised, but I’m glad to see. Sorry Dan.
Daniel Newman: No, absolutely. Scott, as we wrap this up and wind this down, we see some great innovation through this partnership. I’d love to just get your take how you see these technologies, whether it’s water cool, all the way to the next generation high-performance systems. How do you see this enabling more innovation across public sector?
Scott Tease: Yeah, so Dan, we understand the technology at a very deep level. And it’s not just the CPU U at a GPU, it’s all these other things that Pat talked about. Understanding how to apply those is something we need to partner with him and his researchers and with our ecosystem. So what we’re doing with Imperial is pretty cool. So we actually gave it a name we call it Icicle and it’s Imperial College, IC, Intel Corporation, IC in Lenovo, Le, Icicle. And it just shows it’s the power of the ecosystem. It’s working with all our partners to put our knowledge behind what these researchers trying to do, together we do some pretty amazing things. We want to do that all over the planet.
Daniel Newman: Well, I really appreciate both of you, Andy. Scott, thanks so much for joining us here on The Six Five.
Scott Tease: Thanks guys. Nice seeing you.
Andrew Richards: Appreciate it.
Daniel Newman: All right, and that’s a wrap here at SC23 in Denver, Colorado. Supercomputing is hot. It is back from flops to tops. It’s time to say goodbye but hit that subscribe button. We’ll see y’all really soon.