On this episode of The Six Five – On The Road, hosts Daniel Newman and Patrick Moorhead welcome Lenovo’s Giovanni Di Filippo, President of EMEA ISG and Dieter Kranzlmueller, Chair of the Board of Directors of the Leibniz Supercomputing Centre (LRZ) for a conversation on Lenovo’s partnership with the LRZ and how they are driving innovation forward in the High Performance Computing (HPC) space, together.
Their discussion covers:
- The Leibniz Supercomputing Centre’s (LRZ) role and objectives in advancing supercomputers and how the partnership with Lenovo helps drive strategy
- The advancements made with Lenovo’s partnership with the LRZ, driving innovation forward in the High Performance Computing (HPC) space together
- How Lenovo’s partnership with LRZ has evolved over this past year, and the milestones marking the progression of their partnership
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 here at Supercomputing 2023 in Denver, Colorado. We are currently in the Lenovo booth where you can hear it is rocking and rolling around us. There’s a ton of people. Dan, I may or may not have been stampeded last night waiting to get in this event. I mean, what’s going on here is supercomputing and AI, just that cool or popular?
Daniel Newman: I think it is. I know the word cool and AI and accelerated computing and supercomputing probably haven’t been put together in the same sentence in a long time, but I did stand outside, I was looking down on a crowded street here in Denver in front of the convention center of people trying to get into the building for the opening reception, anything for some sweet potato fries.
Patrick Moorhead: Exactly.
Daniel Newman: But in all serious Pat, AI and accelerated computing is in vogue, it’s been a massive year for this technology. And this is where so much of the foundation of these LLMs and these next generation workloads, and of course the advancements of space and all the things starts here.
Patrick Moorhead: Yeah, it is amazing too. And I think I went to my first Spercomputer conference about a decade ago, and it was all about flops, right? A certain type of compute and then it shifted to, not the flops is going away, but this combination of FLOPS and AI, right? And that’s I think why this has got bigger. And we’ve seen high performance computing, which was typically national labs and university labs being embraced by energy companies, being embraced by pharmaceutical companies, being embraced by design companies to be able to do their work better and faster. So without-
Daniel Newman: In…
Patrick Moorhead: …further ado… Oh, you were about to do this.
Daniel Newman: No, I was just saying in the AI era, you got flops and tops.
Patrick Moorhead: Oh my gosh.
Daniel Newman: We got flops and tops.
Patrick Moorhead: You’re always coming up with a new one.
Daniel Newman: I just wanted to say-
Patrick Moorhead: And along the way we are in the Lenovo booth- is a market share leader in high performance computing, and I want to introduce two our guests, Giovanni with Lenovo and Dieter with LRZ. Welcome to The Six Five, first time guest, but we’re thrilled to have you here.
Giovanni Di Filippo: Excellent. Thank you very much for having us. Excellent, yeah.
Dieter Kranzlmueller: Pleasure to be here.
Daniel Newman: So let’s start off talking about LRZ roles and objectives, all about advancing supercomputing. Can you talk a little bit about the partnership with Lenovo and how it’s going to help drive the next generation of supercomputing?
Dieter Kranzlmueller: Sure. So LRZ was founded 1962 and the goal was to run big computers at that time, we’re still running big computers and we are doing it together with Lenovo and Lenovo is building our leadership class systems, so to say. The goal of LRZ is we are a partner for the digital transformation, so we do anything for scientists and we help them to do their daily work. And that really starts from the desktop to what we have as a big system and how to use it. And the real intention is… Our metric is time to science, so the goal is that we really enable these scientists to get their leadership, their enlightenment, their scientific breakthroughs through computing.
Patrick Moorhead: So who can get access to the LRZ Supercomputer services, is it folks just inside of Germany, is it EU, is it global?
Dieter Kranzlmueller: So LRZ, is it German national HPC center, one of three and we are providing access to these systems through your peer review process. So a scientist submits a proposal which is then evaluated on the technical merits, and depending on how good the science is that the scientists want to do, he gets a certain allocation on time on the system.
Patrick Moorhead: And these are what things that you would expect, drug discovery, weather patterns, design, things we haven’t even thought through yet?
Dieter Kranzlmueller: That’s a very good question. In fact, we are very proud because we have all scientific domains, so we really go from astrophysics to zoology kind of. And the important thing is it’s also not only the scientists that know exactly on how to use these systems, we have those that we call the clickers, which use the mouse for their scientific research to those, I don’t know, high energy physics people who take every bit and know what the processor is doing, you know?
Patrick Moorhead: Cool. That’s cool.
Daniel Newman: So talk a little bit about the partnership, Giovanni, it’s really great to learn from Dieter here what they’re focused on but of course there are many companies, partnerships have to be very strategic and as they decide who do they want to build with, who do they want to invest with, who are we going to move forward and create the most advanced computing capabilities, they’ve chosen to build a deep partnership with Lenovo? Talk a little bit about what’s forming that partnership and how that partnership is developing.
Giovanni Di Filippo: Yes. So LRZ for us has since many, many years a very important partner, particularly of course with German market, as Dieter said they’re also offering their services across the markets. And from this perspective, we are investing on our side as Diet and his team is investing, so we are looking into roadmaps and for us it’s important, it’s not just a one-off, it’s really investment on both sides looking, understanding what are they looking for in regard of capabilities, also in regard of next generation of CPUs and GPUs of course. And so I’m really proud, to be honest, to sit here now together with Dieter and talk about LRZ and our partnership.
Patrick Moorhead: No, that’s excellent. You said goes back many, many years.
Giovanni Di Filippo: Yes.
Patrick Moorhead: You had talked about the foundation in 1962, when did Lenovo step in to the picture?
Giovanni Di Filippo: Oh, that’s a good question. Dieter, I have to say… Do you know exactly? Yes.
Dieter Kranzlmueller: I can tell you. What we did in 2012 is we got the first hot motor cooled system, which was positioned-
Patrick Moorhead: Wonderful.
Dieter Kranzlmueller: …four on the top 500 and that was IBM at that time, it then came the transition, so it was the same team working for us. So I would say we started in 2012, more or less-
Patrick Moorhead: That’s great.
Dieter Kranzlmueller: …working together.
Patrick Moorhead: So over a decade long-
Giovanni Di Filippo: Yes.
Patrick Moorhead: …relationship. That’s great.
Daniel Newman: So Dieter, I’d love to get your sort of more macro take, it’s great to hear about the partnership and the work that you’re doing with Lenovo, obviously day in and day out now you’re seeing the interest, the demand, the national economic value that’s being put on AI. How are you observing and how are you feeling that this more broad market momentum for AI is changing the work that you’re doing?
Dieter Kranzlmueller: Absolutely. So the thing we have is we’ve just been installing our latest system, SuperMUC-NG phase two, and that’s a system that was really originally indicated to be an accelerator for HPC. But with the technology we have, it’s also the perfect machine for doing AI, for doing large language models and to really attribute to what we need to do in this day with the advent of all these AI things. And the good thing again is this partnership also allows us really to bring the latest technology to AI. And we SuperMUC-NG phase two, what we’ve been doing is we have the first system that is really a hundred percent hot water cooled. So it’s not only the core, the CPU core, the GPUs, which are brand new Intel GPUs in that case, but it’s also that the entire system, even the power supplies are hot water cooled. And the good thing is I’m really happy that we’re working on this because we are also using this on our AI clusters, so the AI clusters have the highest density of what you can get of AI within one rack. And I think it’s 144 GPUs in one rack because of the technology advancement that we have been doing together.
Giovanni Di Filippo: And LRZ in particular, now they put a lot of value in regards of energy efficiency, and with warm water cooled systems, you can save up to 40% and think about on such an installed base when you can save 40% of energy costs, that’s a significant savings, which then of course they’re going to invest in different infrastructure, in different architectures. Yes.
Dieter Kranzlmueller: Well, in fact, all the money that we are saving is put into buying a bigger machine, so science profits from energy efficiency.
Patrick Moorhead: It sounds like a virtuous cycle here.
Giovanni Di Filippo: Yes, exactly.
Patrick Moorhead: Right? I have a technical question, but it’s a macro question. I talked a little bit in the run-up about FLOPS in AI and technology is very rarely, “We’re going to stop doing one thing and just do another.” Right? We saw those analytics to ML to DL, now generative AI, we’re still all doing all four of those and we’re still doing a lot of FLOPS based computing, but we’re also doing AI based and generative AI. I’m curious as to, is it possible to put a percentage on the types of workloads like AI versus SLOPs? By the way, I’m expecting AI numbers overall to be low because it’s relatively new and also when we polled people last year, it was low then, and these things don’t flip overnight.
Dieter Kranzlmueller: Well, I would-
Patrick Moorhead: Or are they?
Dieter Kranzlmueller: …I would say we have seen dramatic increases in the last year, so the AI performance is great, what we have now. Of course we expect much more growth even, but that’s what we already see. And bringing these things together really means for us that we need to balance all these topics that you mentioned, and that’s a tricky question. If you ask me how much percentage is HPC, I would say, “I don’t care.” I want the science to be done, and we need a system that is also able to reflect all these requirements and support all these different things, which is why I always say I don’t care about the top 500, yeah? Because what is the next scientific breakthrough? It’s solving the problem that our society has today with the tools we have and putting that together in a system that helps us inside.
Patrick Moorhead: Right.
Daniel Newman: Well, Dieter and Giovanni, it’s an exciting time. I want to congratulate you on all the progress in the partnership. This will be a big transitional year and next year we can be sure, Pat, as we show up here that there’s going to be more FLOPS, there’s going to be more tops, there’s going to be more applications that are going to be driving the future. And I kind of love to also see this convergence that’s going on because you got… If you think about accelerated computing and supercomputing and then we’re talking about in a lot of our conversations in labs around quantum, you’re seeing these forces coming together and I think there’s some really exciting things that are going to happen and a lot of them will be announced and heard here at Supercomputing. So Giovanni, Dieter, thanks so much for joining us here.
Giovanni Di Filippo: Thank very much.
Daniel Newman: …on the show.
Giovanni Di Filippo: Thank you. Thank you.
Daniel Newman: All right, thank you everyone. All right everyone, thanks for joining us here at Supercomputing 2023 it’s great to be here in Denver at the convention center looking at all this technology, but for now, for this show, for Patrick and our guests, it’s time to say goodbye. See you all later.