SiliconANGLE theCUBESiliconANGLE theCUBE
  • info
  • Transcript
Innovating for sustainable IT
Clip Duration 01:39 / September 27, 2023
John Hampton, AMD & John Frey, HPE | HPE Discover 2023
Video Duration: 16:43
search

Hi everybody. Welcome back to Las Vegas. We're here in the Venetian Conference Center, and this is theCUBE's continuous coverage of HPE Discover 2023. This is day three. This is our 13th year covering HP/HPE Discover. John Fogerty last night was rocking the house. He was really, really good. Guys, Leonard, what's he in, his late seventies? He's 78 years old. The guy was amazing. Bad Moon rising. He brought out all his old songs. He had his kids playing behind him. The saxophone player was unbelievable. The crowd was rocking. Unfortunately for Leonard, no trombone, but hey, he was really, really good. Put me in coach. Just fantastic. I'm here with Rob Strechay. John Hampton is also here. He as the senior director for Worldwide Customer Field Support at AMD. And John Frey is back, Chief Technologist at HPE. Gents, good to see you. Yeah, great to be back. Thanks for having me, gentlemen. Yeah, you bet. So John, let's start with you. John Hampton from AMD. Give us the update on your business. I mean, AMD has been rocking. Lisa has the company on the right track. We see the markets, the stock market loves AMD and it's been fantastic to watch you guys come back. Give us the big, the update on the business. Yeah, I've actually been at AMD now about 20 years, and, really, right now it's been one of the most rewarding times to see AMD and the climb and, you know, the stock price and just being able to bring this new technology and innovation to the market. And, really, that's what we're about at AMD, innovation is in our blood. And that's what's really driven this success. Our mission at AMD is to use our technology, use our innovation in partnership with HPE to solve the world's most important challenges. One of those being here today is sustainable IT, sustainability. Yeah, and the epic, you know, architecture's been epic, essentially. And John, you guys know it well. It's one of your key vectors of your architecture. But we were talking earlier this week about sustainability. It's like all of a sudden this has become such a front and center issue. Why is that? Well, there's a couple reasons. One is institutional investors are starting to ask CEOs and boards about the topic. So of course they want to make sure they have great answers for them. Increasingly, customers are asking companies they do business with about their ESG and sustainability programs. And for a non-manufacturing company, the technology infrastructure is often the largest source of carbon emissions for the company. Three, I think that's a cost opportunity. We often think of sustainability as costing more, actually, sustainable IT saves lots of money. And one other driver is geopolitical situations like the war in the Ukraine, all of a sudden EU data center operators saw their power bills double or triple in the span of a month. So they really had to react to that. And I guess the last thing I would mention is we've got some emerging regulatory drivers in the EU and here in the United States that are going to bring a lot more focus on carbon emissions and technology can be such an enabler to reduce those. For all those reasons, I think it's a hot topic. You know, when SVB, you know, had the bank run, a lot of critics came out, no surprise, a lot of finger pointing and they were criticizing, you know, their focus on ESG. And that's why they... But the point is miss, this is off the base because ESG has a business case associated with sustainability is good business. Can you explain why? That's absolutely the case. I like the way you said that. There was a recent report that came out that said, "Businesses that are putting sustainable IT metrics in place today and focused on sustainability transformation are two and a half times more likely to be successful businesses in the future, high performing businesses." So we absolutely agree with that. John said it very well. There's a very clear business impact and we're seeing a lot more of a connection now. Between those decisions in IT. Of course, we're here to improve our environment, take care of the next generation, improve sustainability. But now more than ever before, we're really able to connect this to the bottom line. And in fact, one thing we have talked about with some of these organizations, particularly in Europe, that have seen rising power and electricity costs, we talked to them about the cost of doing nothing.

Their concerns are, I have to address these rising costs, so I can't invest in new technology. But we can sit down with them and that's why we're here to come and consult and say, by investing in new technology, the performance and performance per watt increases are so dramatic. We can actually increase your power, or excuse me, increase your performance and dramatically reduce your power and improve your sustainability. And it's about efficiency, right? That's right. You're sitting there and you need to be more efficient. Sustainability is not just about green power and everything else, right? It's about how do you take things out of that to really become more efficient in how you do things? How are you guys working together to really get at that? Yeah, so there's a lot of great ways, and in fact, if we go back a couple of years when HPE set our 30x product performance improvement goal, AMD instantly reached out and said, we can help you along that trajectory. So how do we do that? So we've been collaborating for years on that. When we think about system architectures, what do customers want? Higher core counts, smaller packages, but they want to be able to cool them effectively as well. So how do you balance all of those objectives? AMD has been a great partner in those. Customers saying, look, we want to have denser architecture, maybe even manage those workloads on a smaller number of pieces of equipment. So how do we do that? We come in with next generation architecture, higher core counts, and we can actually cut the size of the hardware infrastructure required for the same workloads, often by 50%.

So we collaborate there. We also collaborate further. Supply chain assurance. HP and AMD have worked together on that topic for years because we don't solve that as one company. We solve it through collaboration. I sit on AMD's external stakeholder advisory committee, so we share roadmaps with one another and help each other get better. Probably a week doesn't go by that we're not talking back and forth to drive programs forward. So collaboration is everything. I love this topic. I was 15 years too early on this. In 2007 we had a partnership with PG&E and we worked with them because they would give rebates to companies that would install energy efficient technologies. Virtualization was one. 3PAR was another one where we could show them that virtualized storage. And we would go in with the engineers, we would do tests. We had an energy lab like the ESG lab, that the company used to work with, different ESG. The research company. But it was a heavy lift to get people to even pay attention 'cause the CIOs would be like, "Eh, I don't pay the power bill." But now it's completely changed and the technology behind it. John, I know you have something to show us here. I'd love to take us through this sort of... Yeah, so I wanted to show our fourth generation EPYC chip. And what you're seeing here is under the hood, we've taken the lid off of this and want to draw your attention to all these different, what we call chiplets on the CPU. It's really another level of innovation in how we design chips. These chiplets allow us to improve efficiency into the future. It's going to dramatically help us improve sustainability. And the reason for that is, historically it was just one big monolithic die. Now AMD has all of this exciting technology, CPUs, GPUs, FPGAs, and AI engine for example. You can kind of swap in a flexible, efficient way exactly what our customers are looking for. So what took place on a motherboard in the past, a lot of that's now the innovation, right at a chip level all connected. So wanted to just showcase that and talk about where we're going because this is going to allow us to take another step function improvement in sustainability in the future. So explain the chiplets in a little more detail. So they're like mini workers that can do offloads for networking and storage and things like that. That's right. You may say, "Hey, I want to focus on high performance computing and I want maximum cores." Each one of these chiplets will have eight high performing cores. But you may say, "John, I'd love to actually bring in some AI as well." We can take an AI engine chiplet and drop that in a very flexible way. And again, it's all integrated and so that reduces latency and improves performance. And so the efficiency is just going to go through the roof. Do you have a sense as to, like, what the opportunity is? I mean, we've just basically taken the same architecture for years and, oh, we're running SAP and Microsoft, you know, whatever, you know, SQL database. But the inefficiency sort of over the years grew. Is it 30% opportunity to do that kind of specialized offloading? Is it less, more, any sense of that? It's funny you mentioned 30%. We actually have an AMD corporate goal that we're going to improve energy efficiency 30 times by the year 2025. So you were right on point there. I will add to that, together with HPE, we're quite proud of where we stand currently in the Green500. It's a great proof point. 70% of the top 10 is based on AMD CPU, GPUs with HPE. And just add one other thing. It's really incredible about what we're going to do, again, future looking here. El Capitan has been announced now, it will become the number one supercomputer in the world. And for this discussion, the number one most efficient supercomputer in the world, using our next generation MI300A. That's where you take a CPU and a GPU, integrate them together in one chip. We call it an APU with high bandwidth memory, shared memory. So that's going to allow us to jump the efficiency, I'd say another 5x. What's the A stand for? The APU. A single chip design. But what's the A standpoint? >> Accelerated. Accelerated processing. >> That's right, yeah. Yeah, and I think that's totally key to how people are looking at this because when you start to look at AI and training models, right now, you know, one of the even the more efficient models out of Meta would take 41 years worth of power to go and train that model for the average US house. So when you start to look at that amount of, I get that has to be where you guys are aiming is, hey, we know these workloads are happening. We know that they're coming along. People are going to do them in different ways. They have to not only train it, but they got to keep it going. Is that really the focus of what you're both... Yeah, it absolutely is. >> Absolutely. And one of the reasons is because so many customers are asking for example, about AI, but they don't understand the environmental implications of using AI solutions. So part of what we help them do, one is run those training models and run the solutions as efficiently as possible, but two, understand the environmental footprint of training those models. So can we optimize that and running the AI solutions so we can do that. So one of the announcements that was made this week as an example, was HPE GreenLake for large language models. We're making supercomputing technology available to a variety of users that might not ever buy a supercomputer themselves, but want to run AI models to take advantage of that. It's in a partnership in a data center in Canada, and on a high percentage, almost a hundred percent renewable energy, which allows us to offset some of that carbon footprint of bringing those technologies to bear.

We already mentioned some of the supercomputers, El Capitan, LUMI is another great example of that, of a partnership that sits in Finland. But what it enabled was the EU destination Earth to actually build a digital twin of the globe to understand how ecosystems interact with one another. So we do better climate modeling and have better solutions. So yes, those technologies are going into play. We've got to do them efficiently and effectively. That's why we look at the Green500 as well as the top 500. But we want to solve technological and societal problems with them. That's technology for sustainability. And that's really where we're going in the future as well. A couple of times this week, and it might've been on your earlier interview, John, I know Antonio brought it up. He said that we're really proud because our 2040 goals are science-based. They're not as a result of buying carbon offsets. Can you explain that a little bit more, what that's all about? Yeah, so there's a third party group called the Science Based Targets initiative. It's scientists that work with businesses from all industries. And when you want to set a really rigid and gold standard target for climate, or carbon reduction, they actually look at your goal. So in our case, it was a net zero goal by 2040. So they looked at that and said, one, is it in line with climate science to the one and a half degree scenario, the more rigid scenario. Number two, are you looking across all of your emissions, not just your own operations. 'Cause for us, that's only 3% or 4% of our total. So do you look across all emission sources, your supply chain, when your customers use your products? And finally what they say is the goal has got to involve significant reductions before you even think about neutralization.

In our case, 90% reductions. And what's happening is around the world, there's a variety of carbon claims being made. There's a lot of scrutiny from regulators and journalists. And as soon as we say we have a science-based net zero target, all of the scrutiny stops. And people completely understand how that's the gold standard. Yeah, and I think that the science-based is the key in the fact that you have to work together with your supply chain. So I would assume that you are really doing the same thing together with HPE as well. Oh yeah. That's no question. In fact, the theme for AMD right now that you'll see in our booth here is together we advance. And I love that because that really is what it's all about. It's the collaboration. In fact, with HPE, we just sat down with a very large enterprise customer here at the event. And we took science and data and modeled for them how they could take their current data center, basically, cut the number of servers in half. Dramatically increase the performance, reduce the power, and then of course beyond the power, they were able to reduce their hardware costs, their licensing costs. So the bottom line, TCO, again, the business impact is very well connected. But again, the beauty of all of this was the sustainable IT impact that we were able to make by doing that. So together we advance really does, I think, capture exactly what we're trying to do. Well, and thinking ahead of the impact of AI and what that's going to do in terms of driving demand for compute. So could you talk a little bit about your roadmap there? Yeah, and I'd love to. Dave: And what we can expect going forward? Yeah, in fact, we just made an announcement this past week here in June 23 on a new GPU that will really change the face of the industry. We recognize too, we need to bring more competition to this industry. And I hear it every day. >> No, I bet. So the AMD Instinct MI300X, especially, is going to allow us to scale into the largest language models out there. And just the capabilities are truly astounding what this chip will be able to do. So we're quite excited. We have a lot of fun conversations happening with HPE on this MI300 family. It's going to help us with supercomputing, high performance computing. But to your point, it's really going to change the game in AI. So I want to encourage the viewers to reach out to AMD, find out more about what they want to do with AI. We'd love to have the conversations that explain how we can really help them solve a lot of those challenges. It's a huge point. If we're going to democratize AI, we have to have more than one chip supplier. John: That's right. So keep up the good work. We really appreciate It. You got it. Dave: Alright John and John, thanks so much for coming into theCUBE. It's good to see you guys. Good to see you. >> Thanks so much, enjoy. All right, in a minute we're going to come back and talk about HPE GreenLake. It's the unifying force of the company and the company's cloud strategy for its customers. Dave Vellante and Rob Strechay from HPE Discover 2023. You're watching theCUBE, be right back.