The Six Five team reviews the launch of Marvell’s new Octeon 10.
Watch the clip here:
If you are interested in watching the full episode you can check it out here.
Disclaimer: The Six Five Webcast is for information and entertainment purposes only. Over the course of this podcast, 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 do not ask that you treat us as such.
Patrick Moorhead: So, Marvell launched a new DPU based on Octeon 10 and threw a little 5- nanometer action in there.
Daniel Newman: A little, Pat? I mean, come on. That is the news. The company basically invented the DPU. There might be some argument here. Wasn’t necessarily the fastest to get the marketing nomenclature correctly. But in terms of doing it, Matt Murphy, another great part of our Six Five Summit this year, CEO there has turned this company around. We talked about that here. In fact, watch the session we really talk about it there. But the DPU space and essentially for those not totally familiar, it’s about building silicon to offload certain work that the general purpose CPU’s have been doing that isn’t most efficient for them to do. So you’re offloading things like networking and storage.
And Marvell has been really leading in this space. The Octeon 10 is basically all about accelerating, helping process the whole spectrum. Security. Networking. Storage workloads. And really the company’s focus here is 5g cloud carrier and the data center, which are big booming businesses, and the ability to shift off these workloads more efficiently. So, Pat, as you said, 5 nanometer here, well listen. This is huge. The goal of this next … of the Octeon 10 is three times performance, 50% lower power consumption. And it’s taking advantage of the ARM Neoverse N2 5-nanometer process.
And Pat, I believe it’s the first DPU to do this. It’s got hardware acceleration. It’s got integrated terabits switch, vector packet processing accelerators. So this thing is essentially loaded to the gills with features and function. And as I see it, and I’ll say it, Pat, the thing that gets me excited is Marvell continues to push the envelope here. They’re pushing the envelope, they’re challenging. I’ll give Jensen Huang, another guest of our event, by the way, everybody came to the summit, Pat, a lot of credit for nailing the marchitecture of the DPU.
But I want to also give Marvell a lot of credit for actually building these products, getting them into market and being side-by-side with these big silicon makers, these cloud providers, 5g automotive, building what is necessary to make our CPUs more efficient and just understanding exactly … another company that by the way, understands exactly where they belong, is building out products and services to win there and that’s why I was so excited. I was so excited, in fact, Pat, that I actually gave Marvell a quote in their press release this week about the Octeon 10 launch because, and I don’t just give those away. I’ve got to be really excited and I really think the company is nailing it in this DPU space.
Patrick Moorhead: All good analysis, Daniel. And I guess the only thing I’ll add is that just kind of to reinforce the architecture of the data center is dis-aggregating. There was a day that server storage, networking and security was all kind of packed into one place. And in order to get the scale-up capabilities and to be able to more easily jockey between different types of applications requires dis-aggregation. And we see in storage. Intelligent SAN and masses that have acceleration like AQUA on AWS. The other part is networking and you’re 100% accurate that Marvell created the DPU as an appliance. I think AWS perfected it, but if you’re not AWS and haven’t spent 10 years and acquired a company like Annapurna to do increasing offloads, who are you going to turn to? You’re going to turn to Marvell in the cloud.
If the enterprise, not to be confused with Nvidia who’s really going after the enterprise DPU market, but they continue to move forward. One of the things like you said that didn’t get enough credit was this is using TSMC’s 5-nanometer process. And also ARM’s latest server chip, which you could run general purpose on that from a SPECint standpoint. Right now, at least on paper haven’t seen this yet, is higher SPECint than any chip from Intel or AMD. Now, this is not supposed to do general purpose computing. I understand that because it’s a DPU. But a way to put a massive compute engine on the Edge and I’m looking forward to, hey, beyond maybe security DNS or things like that. What else could they do here? Well, they did add other own unique machine learning engine, which is pretty cool. But anyways, great stuff, great chip analysis here. God do we love chips.
Daniel Newman: Well, you can’t have SAS without chips.
Patrick Moorhead: You can’t chips and SAS, the two bookends that make the world go around.
Daniel Newman: You really can’t have anything without chips in the tech space. Everything runs on a chip. And you know, I think there’s a lot of people who love tech that do not appreciate the fact, whether it’s your gaming console, your mobile device, your laptop, the server that runs your businesses applications. Look, Pat, we don’t geek out on chips because it’s just a popular topic. We geek out on chips because as chips become more capable and as infrastructure becomes more elastic with what semiconductors are doing, this stuff we get to do is cooler. Speaking of this stuff we get to do that’s cooler, you love to look at cameras. So, I’m not the host, but I just took over that turn. Normally I just let you go and I just couldn’t help myself.
Patrick Moorhead: It’s okay.