Ask Paul: August 29 (Premium)

Happy Friday! A very busy month of travel for me begins on Monday, so let’s kick off my last weekend as a free man with some great reader questions. Clearly, we have worries.

? The future of Xbox is portable, or hybrid, or something

hastin asks:

Any thoughts on the “Xbox Full Screen Experience” they have been demoing as part of the ROG Xbox Ally preview? Does this feel like the “future of Xbox” to you?

Yes. Yes, it does.

Personally, I think this could solve some of the gripes that gamers have about Windows 11, especially if you can do the full-screen experience on any device. Might be great for standalone PCs as well, especially if hooked up to a display.

It astonishes me how negatively people–for lack of a better term, maybe “fans”–react to things these days. But when I look at what’s happening with Xbox, and this has been true for several years, I see this set of huge problems that could have easily destroyed this business. And then I see a team at Xbox doing everything it can to not just rescue the business but turn the platform into something truly special. This isn’t just “making lemonade.” It’s more like they were handed lemons and turned it into gold.

Backward compatibility, cross-play, cross-progression, Play Anywhere, on and on it goes, and what that all adds up to is the most gamer-centric gaming platform there is. And now we’re seeing what appears to be a transition from a console family that’s failed to a PC-based gaming platform that can work across a range of devices, from desktop PCs to laptops to gaming handhelds and, I think, to what will essentially be PC-based consoles. And instead of doom and glooming this, seeing it as Xbox somehow losing everything that it was supposed to be, I see this all as making Xbox better. As a business, but also for me as a customer and fan.

But I don’t have to watch this video to know that this UI makes sense. Microsoft has been working toward this moment for years via an endless series of improvements to the Xbox app, which becomes the shell on these gaming handhelds, and the Game Bar overlays. And we learned more recently about its efforts to simplify and optimize Windows so that it makes sense as a console/appliance-like system that, by the way, still runs Windows apps. And this is important. A PC-based Xbox means we will be playing Xbox console games on PCs, and playing PC games on Xbox consoles in the future. This is exciting. It broadens what’s possible dramatically.

Today, you can join the Xbox Insider Program, get the latest Xbox app with these updates and the third-party store integration, put it in Compact mode, and use it with a controller and see what this is like. And it’s a lot like using an Xbox console … but with the power of full Windows and whatever a PC can do too. This might be viewed like Apple stripping down Mac OS X to make sense in an iPhone. It might also be like Apple finally turning the iPad into a true PC without ruining the iPad-ness that’s always made that device special. It’s like both, I think.

Granted, Microsoft has failed at this type of thing in the past. Windows Media Center tried to cover the stock Windows user interface with a full-screen UI and remote control-based navigation. And sometimes you’d get the random Windows dialog box or whatever other complexities. But that was 20+ years ago. And the Apple examples noted above kind of point the way to what’s possible in this era. I bet they get it there.

I know there’s a lot of hand-wringing out there now about Linux performing better than Windows on these handheld devices. But I also don’t think that matters much in that it’s a relatively small market anyway. And perhaps this will be close enough in v1. We’ll see. The bigger deal to me is the underlying platform shift to a Windows base and the resulting compatibilities that gives us. And that will impact us much more on PCs of all kinds and, someday, I think, on future Xbox consoles too. That’s incredibly exciting to me.

?‍? Copilot+ PCs matter

helix2301 asks:

I heard Copilot Plus has been kind of a flop. People’s video cards and internet-connected devices are much faster than ai on chip. I was wondering your thoughts.

Lenovo, HP, and Dell are the three biggest PC makers in the world, and all three have made the point that what they call “AI PCs,” which are now mostly Copilot+ PCs, are the most lucrative PCs they sell, with the best growth, and the most impact on the future of their respective businesses. So I wouldn’t call it a “flop.” Copilot+ PC is a terrible brand, but it’s also the future of the PC, meaning that its constituent parts serve as a baseline of sorts for what a modern PC looks like now and in the future.

We need to put this in perspective.

Which isn’t all that difficult: When you consider how long it took for a chip like the TPM to go from an idea to an optional part to a normal feature to a requirement (in this case, to make the Copilot+ PC spec), you’re looking at 20 years. When Microsoft announced the technology behind the TPM, it was poorly received, and mostly seen as a way to prevent Linux from being used on new PCs. Today, it’s part of what makes a PC a PC, but it’s also the basis for the security models in all modern digital platforms, mobile, desktop, whatever.

Basically, Copilot+ PC is just something that’s going to happen. People and businesses upgrade on whatever schedules, and as they do, Copilot+ PCs will become a bigger and bigger part of the overall user base. And then they will simply be the user base. I hope the brand goes away, it’s terrible. But claiming that this thing has flopped is like claiming that Apple’s C1 modem chip has failed because it’s only on a single low-end iPhone model today. Looking forward in time, all Apple devices will use these chips.

If you really mean Snapdragon X-based Copilot+ PCs, I’m struggling to understand how a brand new architecture like this could be more successful given the calcified nature of the PC market and the lingering impact of decades of Intel paying off PC makers to use its chips and ignore its competitors. I wrote a bit about this in my HP OmniBook 5 16-Inch first impressions post, but Qualcomm has scaled its Snapdragon X chips to meet a range of needs and price points, and even the lowest-end version in that laptop is eye-opening, with excellent battery life, reliability, performance, and efficiency. Qualcomm is poised to announce its second-generation Snapdragon X chips for PCs at the end of September. So we’ll see where they take it, though rumors suggest bumping up the GPU is one of the big design goals.

As for NPUs, they’re widely understood to be more efficient than GPUs for what I’ll call hardware accelerated local AI tasks. But less well understood is that video cards (GPUs) are not “faster” than NPUs at these tasks. If there is a task that can run on an NPU, it will run faster there than it will on any GPU you can put in a PC. GPUs are just more common, more versatile, and they’ve been around long enough that they’re well understood by OSes and apps, many of which have been optimized for these types of chips.

You should (re)watch Stevie Bathiche’s 2024 Build discussion about NPUs for a better explanation, but I wrote about this in my Orchestration (Premium) post. He runs the same workload against the CPU, GPU, and NPU, and the NPU is 32 times faster than the CPU and about 16 times faster than the (Nvidia) GPU. This thing isn’t just more efficient, it’s also a lot less expensive to operate. These differences are so great, he refers to the NPU as “an AI step function change.” Meaning, it’s not an evolutionary change, it’s revolutionary.

The “problem” with NPUs is the same as the problem that GPUs had 20 years ago. They weren’t common. When Windows Vista shipped, most integrated graphics chips in the installed base couldn’t run Aero Glass. A few years later, this was no longer a concern, largely because enough new PCs with updated Intel chipsets had sold into the market. But even then, what did the typical user “get” with a better GPU? Unless they played games or really cared about Aero Glass, it didn’t seem like anything changed. But once these capabilities are just available broadly, things do change. Basically, transitions don’t happen overnight.

To bring this full circle, the real issue today with NPUs and AI is a set of contradicting strategic goals. Microsoft is trying to get more and more AI processing to occur locally to save it money in datacenters and to help prop up its PC maker partners through a wave of hardware upgrades. But this hybrid AI future, like that of Copilot+ PC, isn’t just assumed, it’s assured. It will happen. It just requires time and a lot to evolve across the stack. Every day, almost, we get little hints of this happening, baby steps towards that future. The first Microsoft AI models are the most obvious very recent example. It’s a “chicken and the egg”-type problem. But it’s happening.

? Oh, snap

Tied to the above, jchampeau asks:

What is Microsoft saying, if anything, about the seemingly slow adoption of Snapdragon processors in consumer and business PCs? And separate from Microsoft, what are your thoughts? Dell, HP, and Lenovo each just have a very small number of SKUs with Snapdragon processors. HP doesn’t have a way to filter the PCs on their website by Qualcomm or Snapdragon processor (just an “other” option with 80 items in it), and when you use the default “Featured” sort on Best Buy’s site, the first Snapdragon-powered laptop is the 26th one listed. For such a meaningful change, few seem to be taking advantage of it.

As noted above, I don’t see this in the same negative light as some do, and I don’t quite understand what anyone thought was going to happen. AMD has made better CPUs (and much better APUs/GPUs) than Intel for years (decades?) and it’s only now starting to make inroads, and only because Intel is struggling and can’t pay off PC makers anymore. It’s not like Qualcomm was going to swoop in and dominate this market. Per above, these things take time.

Qualcomm was able to get all the top-tier PC makers to adopt Snapdragon X across the board, and that’s impressive. With AMD finally getting some traction, and with a year of compatibility wins behind us and negating all those complaints, the platform is in a great place. It’s about to get a major generational upgrade, and we’re going to see competition in this space soon as well. I would pay most attention to MediaTek in that space.

? And speaking of flops

helix2301 asks:

What is the deal with Surface and the NFL? There has been no news or anything. I am wondering if Microsoft making money on that that might be only thing keeping surface going?

The NFL and Microsoft expanded and extended their Surface partnership for multiple years this past week with the stated aim of delivering more “game day AI operations.” As part of the renewal, the NFL is buying more than 2,500 Surface Copilot+ PCs, giving coaches and players access to a new feature built with GitHub Copilot for “filtering plays based on criteria such as down and distance, scoring plays, and penalties to quickly analyze formations, decipher coverages and make more data-driven and strategic decisions.” A Microsoft 365 Copilot-powered dashboard will give in-booth analysts “actionable insights faster to influence game strategies — such as personnel groupings and snap counts.” And Microsoft and the NFL are working on ways to use AI agents to assist game operations managers, create Azure AI video tools, and analyze draft prospects.

I had almost forgotten how long this deal has been in effect: The NFL started using Surface in 2013 on-screen, and though some on-air personalities infamously referred to them as “iPads” in the early days, we’re kind of past that now. I guess Surface will remain a fixture in the NFL for years to come.

? Get out of the sidecar, Ralph

dremy1011 asks:

Hi Paul, what are your thoughts on Googles plan to restrict side loading of Android apps?

This is clearly tied to antitrust action in the U.S. and elsewhere, and against not just it but also Apple. It’s a fundamental change to Android, which has to date differentiated itself from iOS/iPadOS in various ways, one of which is a level of pseudo-openness that seems more tied to Google’s roots in hacker idealism than to anything pragmatic. This openness was key to getting hardware and software partners on board. And though the complaints we see now are from tech enthusiasts, I don’t see this impacting the mainstream user base at all.

It’s kind of the perfect tech enthusiast hot button topic, really. Something that literally no one uses, statistically, but that everyone seems to feel very strongly about. It sort of reminds me of Linux being open source and what an advantage that is because anyone can see the source code. But no one looks at the source code, not really, and we get weird Linux vulnerabilities that might have been caught earlier or avoided entirely if that theoretical advantage had amounted to anything.

According to Google, the theoretical advantage of side-loading on Android is undercut by the very real instances of malware that are delivered into this otherwise protected platform using that method: Side-loads are responsible for 50x as much malware as apps delivered through the Google Play Store. But I’m curious Google didn’t give us the other relevant number there, which is how many sideload installs vs. Google Play Store installs. Because you know the former is almost exponentially smaller than the latter. And all the malware is from unverified developers.

But that’s the point. It’s not just sideloading. It’s unverified developers creating apps that get sideloaded onto Android. And that’s a different thing. Google isn’t getting rid of sideloading, it’s creating accountability. Commercial developers like Epic Games that want to bypass the Google Play Store legally with their own stores, app ecosystems, updating mechanisms, and so on already have the means to do that, and before the recent changes, they had to use sideloading to accomplish any form of app distribution outside the Store. The system Google just announced is for student and hobbyist developers.

None of this impacts how apps are distributed through Google Play or third-party stores. It doesn’t go into effect until September 2026, giving interested third parties time to test the new developer console in preview. And that initial release will only be in Brazil, Singapore, Indonesia, and Thailand, with a global launch coming even later. Given how well some of Google’s initiatives have gone in the past, it’s possible this will change dramatically based on feedback. Or even disappear. We’ll see.

Google has a useful presentation explaining this change to developers. It impacts almost no one, either on the developer side or for customers. It doesn’t seem onerous in any way. I get that change is unwelcome, and that in this space, everything Google does is (and should be) viewed with suspicion. But one has to turn to the theoretical to generate any real angst. I mean, what if Google changes the terms yet again a few years down the road?

Yep. Sure. What if I get hit by a bus? It could happen. But I’m not going to lose sleep over it. Or this. Though I do agree it’s interesting, and I’ll likely write it up separately today or tomorrow.