A day after AMD announced the Ryzen AI 400 (Gorgon Point) processor for laptops, PCWorld and a handful of other reporters sat down with Rahul Tikoo, senior vice president and general manager of the client business at AMD, to ask about AMD’s client processors: its mobile Ryzen processors, the Ryzen AI Max, desktop processors, and more.
Below are excerpts of the interview, edited for space and clarity.
Client played a very minimal role in Lisa Su’s keynote last night. What does that mean?
Tikoo: It was supposed to be about a 75-minute keynote, and client was about 15 of the 75 minutes, right? So if that gives you a clue, it’s roughly 25% to 30% of the time, and client business is roughly 30% of our revenue right now, right? I mean, so it’s an important part of our revenue profile, and it’s very, very important to us.
This is just my characterization, but it appears that the Ryzen AI 400 is a modest upgrade to the Ryzen AI 300, which was a very good chip. How do you see it?
I mean Qualcomm, kudos to them for continuing to fight the good fight. But you know, Arm is a big challenge in this marketplace, just because of the application compatibility, I feel really good about our [Gorgon Point] portfolio. Of course, we haven’t had a chance to get our hands on the competitive products yet, but everything that we heard yesterday did not surprise us, because, you know, we have our own market intelligence and what’s happening and what the competitive landscape look like, and so we didn’t see any surprises there. Based on that, what I would say is we have a pretty good head [of steam].
You had a certain number of design wins heading into the Ryzen AI 300, and a number of wins with the Ryzen AI 400. If you can’t give us specifics, which are larger?
It’s about the same. What we’re going to see is about between the Ryzen AI 300 and 400 product, and the Ryzen Halo product, we have roughly a little over 250 designs that will be in the market. That’s all three chips. All three chips, yeah, roughly a little over 250 designs, give or take, that will be in the marketplace by the middle of this year, right? Because we just have notebooks that are coming out this month. Desktops will come out in early Q2. [The additional] Strix Halo is also coming out this month. Pro is March. So let’s just call it, the first three-to-four months of the year are going to be busy for us launching the portfolio.
You mentioned the AI 400 desktop. It’s going to be a socketed AM5 part?
Yeah. It’s a socketed AM5 part. I think the interesting thing about the desktop Gorgon part is that it’s going to be the first Copilot+ part, so the first part with a 60 TOPS NPU. We’ve been working with Microsoft and our partners on optimizing for desktop, because you can imagine desktop has a different set of challenges, right?
I think we have a lot of opportunity in that space, and we weren’t there two years ago. We weren’t playing as heavily. We didn’t have enough of a portfolio last year, we had a really reasonable portfolio. This year, we’re going to have even better portfolio.
What we’re seeing is a lot of interest in mobile on desktop, even small desktops and even in large desktops, they’re actually putting mobile on because the socket infrastructure is cheaper on mobile.
Even traditional desktops?
Okay, yeah, even traditional desktops, we’re seeing mobile on desktop now. It’s more relevant in the smaller form factors, like, you know, you have the one liter boxes, the eight liter boxes, the small form factor. So that’s where it’s more relevant, right? But we’ve seen all kinds of desktops use mobile parts.
There was a time a few years back where mobile shifted into two categories, high performance and thin-and-light, right? And it’s sort of the same inflection that you see in desktops.
Let’s talk about what the prices of RAM and storage are doing, and the effects they’ll have. What are your customers telling you about how they’re going to configure their systems? Are they going to continue on pushing upwards to 2TB SSDs or 16GB of RAM?
It depends on the market segment. If you think about creators, they want all the capabilities they can get.
Let’s talk about a car company. They’re designing a car. They’re running wind tunnel simulations on a car. Are they going to sweat a 20% or 30% increase in price and say, well, you know, my seven-year research on the car is going to have to be slower? No, they’re going to invest.
Now, consumers, on the other hand, you and I, you know, when we sit at home and we’re using the laptop for basic internet, web browsing, or email, we’re going to have to make a choice, right? Do we really need the highest end components in the laptop, or not?
Now, we do know there’s a floor. A floor has been set where people like 1TB SSDs are the norm. Nobody buys anything smaller, you know? I mean, even phones, nobody tends to buy anything smaller than a certain capacity, right? So, I think consumers will have to make a choice based on that. But I do expect gamers will continue to invest. Creators will continue to invest.
There’s a rumor that AMD was going to launch a Ryzen X3DX2, which didn’t materialize. What’s going on there?
X3D dual-cache, right? Stay tuned. Stay tuned.
I just came back from Intel, where they planned to invest heavily into the handheld space, which you’ve dominated. They claim that you’re selling “ancient silicon.” What’s your strategy going forward in the handheld space?
We’re very committed to the handheld [space]. I mean, we created the space, so it’s a space that we’re very committed to.
Here’s the beauty, though, of AMD and why we have a much higher chance of success in that space: because of our console business, or how we develop semi-custom silicon for the console business. You can’t just use mobile silicon and put it in the handheld. You can, but the handheld or the consoles, they care about high graphics. They don’t care about as much compute, and they don’t care about the I/O.
So, if you’re putting a notebook chip like Panther Lake in there, and you’re not purpose building it, you have all this baggage that Panther Lake is going to carry around their chiplet architecture. You know, the interconnects of the chiplet architecture, the I/O that they have in there. I mean, it’s a Swiss army knife, and it’s good for certain things.
We can do that, too. In fact, we do that in the handheld space in some segments. But when you think about the core of the handheld space, they want purpose-designed, purpose-built chips that have great graphics technology, great software like FSR, integration with game developers on Xbox, PlayStation, etc. We can have high battery life, good fidelity of content, high frame rate, and we do that very well.
Intel believes their low-power E-cores give them an advantage, as they extend battery life. Does AMD have a response to that?
We haven’t seen any issues there. I’ll tell you this, Intel does play games sometimes, and it’s very interesting.
We had a customer. They said the same thing. They’re like, hey, I can get more battery life with Lunar Lake against the 300 series.
So, we’re like, okay, let’s do a quick experiment. And we did this in the lab. And actually, Qualcomm did a video on this too, because we didn’t want to go out and do a video and everything. Qualcomm did a video on this: Lunar Lake has great battery life when measured with MobileMark with the power connected. As soon as you go in DC Mode, battery life climbs and performance drops. The Core i7 performs like a Core i3.
So, the E-cores are very good for efficiency, very bad for performance. We balance the two, and we’re already making those choices for our customers and saying, hey, you don’t have to worry about it.
Can you talk about the desktop X3D processor and the direction that it’s going?
It’s a very critical part of our portfolio. I mean, the channel market overall. If you look at IDC, the DIY market is about 30, 35 million units. And give or take, we’re close to 60 points of share in that market, right? We’re pretty high. And then as you look at X3D, which is the top of that market, we have over 80 points of share in that market, and it’s driven by the fact that there’s really nothing else that comes even close in terms of performance.
And then with the new X3D part that we just announced, the new part to the stack that, with that boost clock you see on it, it now separates us even more, right? We used to be about 20% better now, or 27% better, when you look at average game performance, and so we’re very committed to that space. That customer base is very demanding, as you can imagine, right? And they’re very vocal.
Do you have anything to say about AMD’s ability to supply chips to its customers?
We’re using the biggest and the best supplier in the world, TSMC. And our Gorgon portfolio is based on four nanometer technology and is a fully ramped, highly yielding, very proven technology. So, we don’t have the same challenges our competition has where they’re bringing up a new technology. We feel very good about it. No challenges.
Threadripper, X3D, and the Ryzen AI Max: these are all innovative though niche products. Does AMD remain committed to all three?
We are very committed to those spaces. We’re very, very committed to those spaces.
How do you see the Ryzen AI Max going forward?
First of all, we will continue to invest in that space. That’s an important space for us. Stay tuned. There will be more announcements in that space over the course of this year.
Our focus has been in ramping developers and gamers around that product. You know, thin-and-light gaming is a space where that product has done well. Creative users is another space that product has done well, and now AI.
