Anyone that knows me knows that if you put me in a room with someone to talk about Windows on Arm, or computing on Arm processors in general, I can talk until the sun goes down. When Arm reached out and asked if I wanted to speak to Paul Williamson, its SVP and General Manager of Client Line of Business, I knew it would be the highlight of my Mobile World Congress.
There was only one Windows-powered Arm device announced at the show, and that was the Lenovo ThinkPad X13s. There was a Chrome OS tablet as well though. Still, the first thing that I wanted to talk about was Apple. While I’ve had several interviews with Qualcomm on the subject, discussing it with someone from Arm is a bit different, because MacBooks are using Apple Silicon, which is based on Arm. Apple is just designing its own chips that use the Arm instruction set, rather than licensing Arm’s designs.
XDA-Developers VIDEO OF THE DAYHowever, this isn’t new to the computing market. Apple has been building custom Arm processors for over a decade, and those have been outperforming the chips that go in Android phones for a while too. So the biggest question on my mind was, why are Apple’s Arm processors better than the ones that are actually designed by Arm?
24-inch Apple iMac using M1 processor
Rich: The first thing I want to ask about is why Windows on Arm PCs can’t match Apple’s performance. Qualcomm’s chips use Arm’s designs, while Apple uses its own. Why are Apple’s Arm processors better than Arm’s?
Paul: If you’re looking for ultimate performance for laptop class, in the past, you would take a specific implementation of the Arm architecture, and it will target a certain performance point. You’ll see us do implementations at different sizes. So you’ll see our ultra cores, our X-series, the big cores (7 series), and small cores (5 series). Put those into a cluster and optimize for performance.
When you’re targeting laptops, you take a different approach to mobile because the tradeoffs are slightly different. You’ll have a different power envelope, so you can afford to have more bigger cores. If the price of the silicon is higher – you can put down bigger caches and more layers to increase the frequency of what you’re doing, because the power, the thermals, and the price allow you to do that. So, I think part of the headache for performance in the laptop space is that we are yet to see the max on what is possible with that.
When we announced our X2 series last year, we made a big point of saying this can be configured to be top-end performance if you put down the right frequency, if you put down the right cache sizes, if you put low enough latency path to memory, you can deliver really, really premium performance.
So, you can get there with Arm’s implementation IP. The question is, when you’re into a market, making that investment in premium silicon when you can’t guarantee the volume is a big challenge for some of our silicon partners.
Rich: Well there’s only one silicon partner right now.
Paul: It depends how you look at it, because you’ve got Qualcomm in the Windows space but you’ve also got MediaTek with their Kompanio chips for Chromebooks.
Rich: MediaTek also even said that they’re planning to enter the Windows space, whenever that exclusivity deal between Qualcomm and Microsoft ends.
One concern that I have about Windows on Arm in general is that the PC market is slow. You might recall that Qualcomm announced the Snapdragon 8cx Gen 3 and the Snapdragon 8 Gen 1 at the same time. But while the Snapdragon 8cx Gen 3 has Cortex-X1 cores, the Snapdragon 8 has the newer Cortex-X2 cores.
That’s an issue that’s directly related to how the PC market operates. OEMs operate slowly, and they like to have a new CPU in hand a good year to 18 months out. Unless Qualcomm wants to announce a chip and say devices will be out in a year – something that it actively tried to change when it announced the Snapdragon 8cx Gen 2 with almost no changes – it has to use the previous generation of Arm’s cores.
Snapdragon 8cx Reference Design
Rich: Also, since you mentioned the X2 design – obviously Qualcomm’s Snapdragon 8cx Gen 3 uses X1 – the way I understand it is that this is a problem less with Qualcomm’s chip and more with the PC market. The PC market moves so slowly that these OEMs want this chip in their hands 12 to 18 months out from when the product is released. So, Qualcomm’s release is always going to be a year behind what Arm is putting out, because they don’t want to announce their chip in December and say the products will be out in 16 months.
Paul: I think some of that is absolutely the case. There is definitely a longer development cycle in the PC industry. We’ve seen that the development cycle compresses in the smartphone industry. OEMs working with silicon partners have squeezed that timeframe.
Rich: Do you think they’ll squeeze it in the PC industry?
Paul: I think it’s entirely possible. It takes that ambition.
Rich: It’s a fundamental problem with the way it’s set up. When you’re Apple, everything is in-house and you can do it. When you’re Intel, you’re developing the chip in-house, so you own that roadmap. When you’re Qualcomm, you have to wait for Arm, they put out the design, then you put out the chip after the fact.
Paul: If you look at the mobile device timeframe, there’s no reason that you couldn’t match that with Arm’s IP timeframe. We work really early with the silicon partners.
Rich: But now you’re asking the PC market to shift.
Paul: To meet that ambition, it would be Microsoft on the software platform and it would be the silicon partners to condense their timeframe for development, and that is not what they’ve done historically.
Rich: This is an industry that is built around Intel. So, it’s like trying to disrupt from multiple angles, which is tough. You’re asking consumers to buy into it, you’re asking Microsoft to develop for it, and you’re asking OEMs to build faster. So it can’t be easy.
Do you ever try to think of ways that you can work better with them and make it happen?
Paul: Totally. We are working closely with our silicon partners, with the OEMs directly, and with Microsoft to ensure that the software ecosystem is there for a smoother transition. One of the other things that becomes beneficial is things like where we’re talking about Total Compute. Because Arm can offer consistency in the underlying approach toward building an SoC, it means a developer is building a game on one platform, you can just take it from mobile to tablet, and on one of our booth demos at MWC, we’re showing it on the front screen of a fridge. The compute capabilities are consistent to the developer and you can take it anywhere you like. Similarly, if we can do that so that the PC SoC has a similar level of portability and similarity, you’ll get a much more rapid cycle where you don’t have to validate a completely new platform every generation. So, we think the scope to bring the mobile SoC approach to the world of the PC. But many of the things you say are challenges.
Rich: They’re roadblocks. When the 8cx Gen 3 was announced, and I’m a big Windows on Arm fan, so I’m not knocking it, but some people were surprised that the chip was X1 based when X2 is the thing. X2 is in the Snapdragon 8. As a writer, I have to explain why it’s still X1.
Paul: I genuinely think that this will get better. With the right level of investment from the silicon partner side, they can be sampling silicon at the same time as their mobile silicon. So, this can definitely improve.
Lenovo ThinkPad X13s, using Snapdragon 8cx Gen 3
Rich: I know Qualcomm wants it to improve too. But now, they’re working on custom silicon too; they’re trying to go the Apple route.
Paul: Right, so they’re hiring their own team and they’re looking at building their own cores.
Rich: Where does that leave the Arm implementation IP?
Paul: We’re fully committed to an Arm implementation roadmap that will happen year after year. We’ve got hundreds of partners that use that technology, and we offer a business model that means people can build their own implementations.
Rich: I’m excited to see more competitiveness in the Arm chip space for PCs.
Paul: I’m loving some of the other trends we’ve seen in the industry. If you look at Google coming in to build their own silicon, it shows that people are taking the Arm implementation CPUs and bringing in custom accelerators like Tensor and making genuinely innovative steps. If we can bring that kind of mobile innovation to the PC platform I think it’s going to be really exciting.
Rich: What do you guys think about different tiers of Arm chips? In the PC space, we’ve had this focus on thin and light. The focus is in comparing with the Intel U-series. Qualcomm has compared the Snapdragon 8cx to a U-series Core i5. What do you think about gaming? Creator laptops? Stuff that competes with Intel’s H-series with dedicated graphics or even an M1 Max?
Paul: I think that’s an exciting prospect. There are different approaches and I think we will see discrete graphics with Arm. That will be exciting. It’s going to be interesting to see how the market evolves and how the disruption of Arm might influence that. There’s definitely an increased focus on that MacBook Air form factor and price point.
Rich: I asked Qualcomm three years ago if they were interested in gaming or creator laptops, and they told me it wasn’t happening. Then Apple switched to Arm and legitimized it. It’s good for everyone in the space. Do you think we’ll see designs like that?
Paul: I think it’s possible to see Arm packages with discrete graphics, and I’m really excited about Linux on Arm. I know the focus is Windows, but for developers that are developing for the cloud on Arm, they’re looking for Linux on Arm-based machines to do that development on. They want big rigs where they can do Linux on Arm development on. That’s an area where we’ll see some innovation as well.
Rich: I’d like to see some Linux on Arm laptops ship, because I don’t think they do. There’s some confusion, because you can’t download an ISO for Windows 11 on Arm. I know Microsoft will change this at some point.
What do you think are the key value propositions of using an Arm processor in a PC? I know what Qualcomm tells people and I want to know if it’s the same. Because it’s different with Apple.
Paul: There are a couple of really interesting factors that are playing out. One is people taking a mobile SoC approach to the design, so we’re seeing a lot more dedicated accelerators for certain use cases, for improved audio, better face recognition, more AI-type accelerators built-in, which really makes a big difference. Obviously, there’s thermal efficiency and the power envelope that you can achieve. That Lenovo device with 28 hours battery life is showing you can do that with Windows.
Rich: Well, I’ll believe the battery life when I see it. I feel like the Arm promises of battery life on laptops has not lived up to the promise.
Paul: I’m very keen to try the X13s to check it out.
Rich: Apple’s Arm laptops have great battery life, so I blame Microsoft at this point. I wonder how hard they’re working to optimize Windows for Arm. I know they’re focusing more on Arm now because Apple legitimized it but the reason they got involved in Arm was to light a fire under Intel.
Paul: I think the best thing for us to do is see what Microsoft says next. Build should be coming up soon. That’ll be the time for us to see what they say next. The devices are coming together and now it’s time to see the software support.
There’s a lot going on here, and this all turned out to be super interesting. It seems like Arm is poised to take over the entire world of computing. Whether it’s Windows or Linux PCs getting Arm Cortex-based processors or Apple getting custom Arm chips, the platform seems to be everywhere.
But Intel is still an incumbent in the Windows space. In fact, Intel actually says that it plans to retake the lead in performance-per-watt by 2025. To do that, it would have to try and make its chips more like Arm’s, something that it’s already doing with its new hybrid architecture.