Ask Paul: August 16 (Premium)

Happy Friday! This has emerged as a blockbuster installment of this series, so I recommend settling in for a long read and an early start to the weekend.

Missing the point

jeroendegrebber asks:

When it comes to laptops, I care about battery life, but not about AI and having an NPU. So ARM CPU: yay! , NPU: meh (nay). Do you think there will be premium laptops without an NPU? Would there be a point? Or am I missing something entirely on this?

I write this jokingly, but … Yes, you are missing the point entirely. 🙂 But please don’t take that the wrong way. Everyone is missing the point, and it’s because Microsoft and its hardware partners are marketing this incorrectly, and selfishly, by focusing exclusively on on-device AI workloads that it hopes will lighten the bandwidth hitting its incredibly expensive cloud-based AI servers. This is a company that just spent $19 billion in a single quarter building out that infrastructure, and over $10 billion in each of the previous three quarters, and it promises similar spending–sorry, investments–in each of the next four quarters too. So, yes, Microsoft can afford this, but it’s also unsustainable. At some point, AI needs to pay for itself. In the meantime, we get Copilot+ PCs.

So. To level set this, think about it this way. Asking whether there will be processors without NPUs in the future is like asking whether you can buy a processor without an integrated floating-point math coprocessor (as we could briefly when there were 386SX/486SX chips in addition to 386DX/486DX chips) or without multimedia (SSE) extensions or whatever. And the answer is no, because costs go down, the technology gets better, and these things are just part of the base package. The thing is, in 1993 or whatever, you as a consumer might have thought, hey, I don’t need floating-point speed, I use Word, not Excel, or whatever. And that’s sort of the argument today: These NPU use cases are either highly targeted/vertical or just not of use to you at this point in time.

But that’s not what the NPU is for. I mean, it is. But we’re getting caught up in too many of these uninteresting and very specific features in certain apps and don’t see the big picture. Part of this story about processor advances over the years also includes integrating ever more powerful GPUs into the silicon, too. And as I wrote recently in A Few Thoughts on Portable PC Gaming (Premium), we’re on the cusp of an exciting new era in which even basic PC processors are good enough to play AAA games at decent resolutions and graphics quality. What’s interesting is that this already happened on mobile (more on that below, in a different answer, and then coming soon in a separate article), and that as these things move forward in tandem, the PC is catching up in some areas (integrated GPU, NPU), just as mobile had done over the previous decade.

Remember when Microsoft first started talking about hardware accelerated graphics in Longhorn and how exciting it was that we could get these effects without hitting the CPU? That was what I call a gimme: It looked better and it performed better. But that was followed by the trough of despair when we later realized that this would also require new hardware–new PCs with better integrated graphics, at the least–and everyone complained because Windows Vista looked terrible on most existing computers. But that changed over time, and while you basically needed a dedicated graphics card at first, that requirement disappeared. Evolution happened.

That’s where the NPU is today. You need special hardware–an NPU with 40+ TOPS–and everyone is complaining because there are dedicated graphics cards that can do 10x that. But that’s misunderstanding the problem the NPU is solving. Which, again, is kind of on Microsoft’s shoulders, because it’s marketing Copilot+ PCs to help revive the PC industry. We see a number, we see lame features we don’t want, and we think the NPU is useless.

It’s not. The NPU is incredibly important. It does for certain workloads (or tasks or whatever) what the GPU did for hardware-accelerated graphics, by offloading those tasks onto silicon that is much, much more efficient and optimized for these uses specifically. This frees the CPU and the GPU to do what they’re good at, improving performance and battery life, which these days is key. It’s the perfect storm. A gimme. It’s just that we don’t see the use cases that well. Or care. Or whatever.

I wrote about this in detail in Orchestration (Premium), so I guess I’ll just recommend reading (or re-reading) that for the big picture. But the short version is that the NPU is “a step function change” for this kind of processing, a change so significant that it will enable new types of computers that don’t resemble today’s PCs. (Stevie Bathiche compared it to the shift from vacuum tubes to microprocessors). What that means is that NPUs are so powerful and so efficient that the GPU comparison–“my GPU could do that 10x as fast”–is incorrect and misses the point. Even the basic NPUs we have today are on the order of 100 to 200x more efficient at these tasks as CPUs and GPUs: Stevie used an example in which a sub-4 watt NPU outperformed 20 Intel Core i7 processors (at 440 watts) or 320 watts of Nvidia GPU. And it does that without impacting battery life almost at all. That’s a super-gimme!

So, yeah. Cocreator in Paint is kind of lame. And you have to really look to find specific features in specific apps that use an NPU and may or may not be useful to you. But that’s how these things work. When you had to buy a specific graphics card in 1996 or whatever to play the first Quake with hardware-accelerated graphics, most people probably just shrugged and thought, I’ll never need that. But those graphics are infinitesimally primitive compared to the integrated graphics in even the most basic processor today. And its successors power experiences everyone uses, every day, on multiple devices of all kinds.

We’re just at the start of the NPU era. So, the question isn’t whether you will be able to buy a PC with a processor that doesn’t have an NPU, it’s why on earth would Intel or AMD even make such a thing right now? The question is whether those things disappear this year or next year at the latest. Even the Raspberry Pi 5 has an NPU, via an add-on kit. The next one will have that integrated into the silicon, I bet. And everything will just get better. Whether we as users understand that that’s at least partially due to the NPU is hard to say. It will just be there, getting better every year, doing more every year, making our devices more efficient and powerful and with better battery life.

Regarding your laptop scenario, I will just put it this way. Having an NPU in the silicon won’t hurt you in any way. (The Copilot key? God help us all.) But it will help you, even if you aren’t using it, or don’t think you’ll ever use it. There are tasks today, like Windows Studio Effects, that run over many minutes or even hours, and they could work using a CPU or GPU, but that would kill the battery life; when you run this against the NPU, there’s no hit to battery life. In fact, it improves battery life because the CPU and GPU can be idle (or focused on other tasks). The set of tasks that will use the NPU like that will only go up during the time in which you own the laptop. So even if you never, ever use the camera for meetings or whatever, it will still be better.

This is the argument I’ve made for Snapdragon X, essentially: Come for the AI (the marketing) but stay for the efficiency, battery life, and silent, effortless performance (the real-world advantages). This is what computing is like when the system is modern and optimized for mobile. Intel and AMD will likely get there too. It’s going to be great. And you can thank Arm and the NPU for this shift, even if you don’t care about them or use (or think you use) them. It’s just going to happen.

Video games play into this too, pardon the pun. More on that below…

Arm ISO Snafu

thewarragulman asks:

With the new breed of Snapdragon X devices possibly gaining more traction in the near future, do you think that Microsoft will finally make installation media for Windows 11 ARM64 available to download like they do with the existing x86-64 versions of Windows?

I don’t see how they can avoid doing so. That said, I wrote about this topic back in July, and while I remain hopeful, nothing has changed since then. The issues here are many. Just from a basic recovery standpoint, those who buy an Arm-based PC need to have that capability, and while some PC makers (like Microsoft, go figure) offer downloadable recovery images for specific computers, some don’t. That’s not good enough. For Arm to take off on Windows, it needs all the same options we see on x64. So … that’s inevitable. It just hasn’t happened yet.

Even on an ARM64 PC I’d still like to do my own installation of Windows rather than use OEM recovery media.

The article I link to above describes your options today, but none are ideal. There are Windows Insider Preview builds you can download directly from Microsoft, and there’s that possibility that one might find its way into stable and you could get off that train. (That said, there’s no Release Preview ISO there today.) And there’s a site called UUP Dump that is anything but official. Looking at that now, I see just one stable Windows 11 on Arm build, and it’s 23H2, not 24H2. So … great.

Beyond that, Windows now has additional recovery options that could help with that end of this problem–including a new Windows Update-based recovery feature that’s faster than Reset this PC I’m writing about now for the book–but nothing for those who want a clean install image. But the issues there are many, too: If you have a Copilot+ PC, it’s not just the OS image, it’s the 40+ small language models (SLMs) that Microsoft preinstalls in this system too. And it’s not clear how and whether we can get those separately. Or how big that download would even be.

So many questions. But this is Microsoft, so we’re on our own. Sorry.

Your kids are always your kids

spacecamel asks:

Need some “Dear Abby” advice. Since you are going through this now, what is the right time to “push” your children off of phone and streaming plans? I have one about to start his first job and one about two years from graduating so my wife and I are talking about when we can start saving more money to possibly, hopefully, retire someday. I want to gradually get them into it and not overwhelm them but I do want it to end. This seems like a complicated subject so I wanted your thoughts on how to approach it.

I’m curious where others land on this as well. Parenting is a huge topic and everyone has different ideas about the right ways to do things. Plus, the world is changing: Kids do things later in life now–dating, sex, marriage, etc.–than previous generations, and it’s still not clear where this ends or what it looks like on the other end. I was 23 when I was married, and my wife was 22. Our son is 26 and our daughter will turn 23 this year. So they’re both older than we were when we got married, and just steady dating seems like a far-off thing for them. I guess we’ll see. I don’t get it.

With that in mind, I have often joked that “kids are the best.” What I mean by that is that our kids will be at the center of the very best and very worst moments in our lives. It’s the one thing my mother, who is a horrible parent (and grandparent, sadly) got right: She told me that my kids would always be my kids and that I would never stop worrying about them, no matter how old they were. And yeah. God, I feel that. So much. It’s a terrible weight. I wouldn’t trade my kids for anything. But they’re the best and the worst. I guess that’s life.

With that in mind, we’re all a little different, but most of us, I bet, would do anything for their kids. And I pretty much mean anything. My wife and I watched a new documentary about the murder of Laci Peterson on Netflix this week for no reason either of us could come up with–we both know the story very well, and it’s not like there’s new information or whatever–and I was struck by the fact that the murderer’s sister is still so convinced that her brother, who obviously murdered his wife, that she studied law so she could become a lawyer and, I don’t know, get him out of jail or something. That’s not just demented and sad, it’s pathological. It’s what happens in families. (Others in his family know he’s guilty, etc.)

My kids haven’t murdered anyone that I know of. They’re just murdering my finances in every way imaginable, with the cars we’ve purchased, the car insurance we pay for, the schools we’ve paid for (and not just four years each, my daughter will end up with at least six years), phones, online services of every kind, gas money, payouts for unexpected bills (my daughter’s car just needed $500+ of routine maintenance somehow, and she works as a babysitter while going to school), and more that I’m either forgetting or blocking out because seriously what the hell where does it end. It doesn’t end. That’s the problem.

I guess what I’m trying to say–or, maybe, what I was working out in my brain as I wrote that–is that I’m not the best person to ask. 🙂 I try, I do. But when the kids call, or text, and need something, saying no is difficult. Also, my wife is usually the point person on this stuff. She can be less accommodating than me in some respects, but I also try to stay out of this, and was shocked–almost literally–last year when I discovered how much we spend to prop up just one of the kid’s lives.

So I asked her. Not because I’m completely out to lunch, but just to be sure I’m not missing something. And we don’t have a plan, per se. Just two kids in different phases of life. We’re kind of playing this by ear.

Our son graduated from college and has a job. It’s not a high-paying job, and he’s looking for a new job, but that’s difficult, and his job does at least have good health insurance. So he pays for that out of his salary, pays his co-pays and doctor bills, pays for contact lenses and all that. But we still pay for his car insurance, phone service, and some streaming, each of which is in bundles that include our stuff. (He’s on my wife’s Verizon account, for example, as is his sister and Steph’s parents, who do pay her for that.) In this case, he’s in transition, we no longer have that big school bill, but he’s not totally independent.

Our daughter is fully dependent on us, so we pay for everything. And that was our deal: We want her to focus on school, and we’re willing to foot that if she does well. She has that babysitting job, but that’s just spending money. So her life is pretty much on us. And she has a few years to go.

I’m not sure where I expected this to be, or if I even had a plan or idea for the future. But I was fully independent of my parents much sooner than this, as was my wife and pretty much everyone my age I know. We’re not helicopter parents or whatever. But we are enabling this to some degree, I guess.

Do I have advice?

And who would take it, given the narrative above?

The one thing we did instill in the kids early on is the need to save for retirement from the very beginning. We funded small accounts for both of them to start, and our son now contributes as much as he can right out of his paycheck so he won’t even know it was ever there. Our daughter can’t do that yet, so we contribute some small amount each year instead. This was the one thing Steph and I did right, financially, and it’s important to us that they do the same. They seem to get that.

I will occasionally poll the kids to see which services they use, which they could live without, etc. My son surprised me last year when he told me that he and some roommates pay for a few things on their own, and since he’s further along, he’ll be fully independent sooner. But Kelly is younger and still in school. It’s all good. But once she gets out into the world, I guess we’ll slowly push her off the dependence, lower our costs, and this will start winding down. (I have often discussed, but never implemented, this idea of one service like Netflix per month, rotated each month. I still think about doing that.)

I’m not sure if there are hard ages for anything. People are wherever they are in life, everyone’s different, and, as noted, people do things later in life now. At least neither of them live with us still, or are forced to. We have some friends with kids in their 30s (!!!) still living at home. I will think about them a bit today so I feel better about myself.

This is a tough one.

The sound of silence

madthinus asks:

Is it me or has the Xbox software team just gone silent? You don’t hear of any new API’s or optimizations being worked on or worst, being delivered to developers. On the Sony side, they have added a bit over this generation, smoothing out the edges of the system.

This has been a very, very strange year for Xbox. Everyone knows the story: Total victory with Activision Blizzard in 2023 has somehow led to an “annus horribilis” for the platform in 2024. There’s no reason to recount any of that here. But I feel like Microsoft’s broader silence–never explaining what’s happening with Activision games on Xbox Game Pass, quietly raising prices without announcing new perks, lackluster me-too hardware refreshes in a year in which Sony will launch a PlayStation 5 Pro, and more–are all tied together.

What we see on the software front from Xbox is minor and can be summed up as monthly dashboard updates of little significance. But we know that Microsoft is plotting a next-generation console–it has said as much, describing it as the biggest leap ever–and we know from leaks that it was seriously considering going with Arm. And per the discussion above about Windows on Arm, that means that a powerful NPU could be at the heart of this next platform. And that perhaps Microsoft is waiting on that, biding its time with the current (sadly flailing) generation because it has to.

Brad and I have speculated on First Ring Daily that the right technologies are coming together at the right time to make a next-gen, Arm-based Xbox console truly different and very exciting. We would like to see a Nintendo Switch-like system that could be portable and living room-based. An efficient, Arm-based system that might dispense with the high-end specs (as Nintendo as done) because that’s not necessary anymore: Thanks to things like AutoSR (super resolution/scaling), lower resolution games could run faster and look great. And thanks to Prism, the x64 emulator in Windows 11 on Arm 24H2+, this thing could run today’s games effectively, something that would be even easier in the closed Xbox ecosystem. This thing may basically be a stripped down Windows 11 on Arm PC using the Xbox app in Compact Mode as its primary UI.

This is almost 100 percent speculation. But it does make sense, and it aligns with the few tidbits of fact we do have. This doesn’t excuse the silence this year per se. But we all know about the Osborne Effect, and it’s obvious why babbling about a next-gen console a year or more before it could possibly ship it would undermine Xbox today even more badly than has already happened. So I can complain about the silence, but I also don’t have any ideas about the best way to disclose things is. I don’t even know what they have to disclose.

But I’ll just put it out there. I think the next Xbox is a Windows 11 on Arm PC, running on some future Snapdragon X, stripped down, portable when needed, and that the underlying software platform is just Windows gaming. It will run previous gen titles–Xbox and Windows PC–in emulation and use NPU-based AutoSR to make them look great. It all comes together. And … we’ll see. It’s just a guess and a bit of wishful thinking and projection. But there must be a reason we didn’t get an Xbox Series X Pro or whatever. I think they want to cut ties with this generation and do something truly exciting and new.

He says, knowing how naive that all sounds. And how much the next six months or more will be miserable for Xbox fans.

Surfacing a win

dremy1011 asks:

You’ve tested a couple of Snapdragon PC’s at this point. If you were able to go back, would you still have purchased the Surface Laptop 7?

As I’m sure you know, I review a dozen-ish PCs every year. And this year, this summer in particular, has been a busy time for PC reviews. I’m not sure why, exactly, but it’s been crazy. In addition to the PCs I’ve written about, I have two here I haven’t yet written about, and one has been sitting there in its box for over a week. I can’t keep up.

But I can tell you this with great certainty. The Surface Laptop 15-inch I purchased with my own money and at great expense is still my favorite of the lot. This type of thing is difficult to explain in some ways. It’s that matrix of choices that go into a big decision, where we each give whatever attributes more weight than others. It’s that emotion vs. logic thing. It’s that I wanted a MacBook Air but without macOS and this is absolutely the closest to it. It’s all over the map. But from where I’m sitting right now, which is the living room and not my office, I can literally see 7 laptops (OK, 9, there are two on the floor too), the iPad Air, and two smartphones. And of those laptops, the one I reach for, the one I want to use the most, is the Surface Laptop.

Normally, I would have moved onto to the “next” laptop. It’s really two or three in regular rotation at any time, and this year, as noted, has been particularly hectic. But last night, I finally published my HP EliteBook 1040 G11 review, which took longer than expected. And I just wanted to use the PC I want to use the most. And so I’m using the Surface Laptop this morning to write this. And I’ll move onto the other laptops I should be using after lunch. There’s so much to do.

One way I think of this is that Surface Laptop isn’t perfect, doesn’t have a few features–like presence sensing and a fingerprint reader–that I would like to have, especially in such an expensive PC. And yet I still gravitate to it. This is like a lot of things in life. When people ask me what my favorite city is, I immediately respond “Paris” without thinking. Paris isn’t perfect. The weather is terrible. It’s expensive, noisy, and full of–sorry–French people. And moving there would be prohibitively difficult given all the bureaucracy and language issues. And … yeah. So we don’t do that. But if we were just rich somehow, I would own a place in Paris and I would go there all the time. We’re not. And I don’t.

Surface Laptop isn’t problematic like that, of course. But that makes the choice easier. Objectively, it’s missing a few things I want. And it’s heavy. But I love it. And explaining why is a bit like explaining why one might like chocolate, or find a particular person attractive, or whatever. You just do.

And I do. I love Surface Laptop.

The PowerToys paradox

gg1 asks:

Any thoughts on reviewing new features released for PowerToys?

Yeah. PowerToys is … problematic.

I love the concept. I love that Microsoft brought it back. And I love that it’s been updating the bundled tools–toys is such a silly name–on such a regular basis, updating it with new tools regularly.

What I don’t like, or get, about PowerToys is why it requires a single monolithic main app. I almost hate it. I wish this was just an installer/updater, and that the constituent parts–the individual utilities–could be installed separately and only when wanted. This thing has turned into a monster, there’s almost too much there. (And seriously, we’re still on some weird sub-1.0 version? Stop. Just stop.) I vaguely wish there was a formal process for individual tools “graduating” into Windows directly, and then retiring from PowerToys. Etc.

After looking at the announcement for app workspaces, which looks useful in theory, plus previous features they’ve released, PowerToys seems to be adding real value to Windows (for a change), but it’s hard to keep up.

Yes. I agree. It’s something I need to pay more attention to. I did an episode about this on Hands-On Windows a while back, and Chris Hoffman cited PowerToys as his single favorite/best tip for Windows users earlier this year.

I will try to figure something out here. This should be in the book as well.

Game Pass confusion

helix2301 asks:

Companies like Stardock put their games on game pass for a while then take them off and put other games on. First I think this is great getting to play the game before you buy it. My question is what is the process? How does Microsoft pick what games go on and does game company make money for being on gamepass?

Game Pass has been top of mind this year for all the wrong reasons: As alluded to above, Microsoft acquired Activision Blizzard at great expense in late 2023 and then spent this year studiously avoiding putting any of its titles on Xbox Game Pass before quietly raising prices and turning what was once great into something not so great.

Microsoft isn’t particularly transparent about how Game Pass “works”–meaning how it selects games to be on the service, how much it pays developers for that, etc.–but here’s what I do know with certainty. When you see a game on Game Pass, Microsoft initiated that, not the game maker. Microsoft approaches developers and offers them some form of payment for putting a title on Game Pass, and Microsoft determines when (and whether) it leaves the service. I suppose the math there is easy enough: Subscribers pay for the service, Microsoft pays developers for the games. Whether it’s profitable, or makes sense, or whatever, we don’t have any insight there.

But there’s one important aspect to this story that’s relevant to what’s happening with Game Pass today. That is, the Game Pass of today is quite different from the Game Pass Microsoft originally offered.

This service was originally designed to do three things. Give Xbox fans a value-add service that would expose them to more games, similarly to how Netflix exposes subscribers to tons of content (though Game Pass requires downloading). Give developers a way to get their catalog titles out into the world again at a time they wouldn’t be making money anyway, giving them a new revenue stream and, hopefully, new fans. And give Microsoft a deeper presence in the crucial subscription services world; before Game Pass, all it had was Xbox Live Gold.

Over time, Game Pass evolved. There are now four tiers, one that replaces Xbox Live Gold, one for console, one for PCs, and one that’s the full meal deal with everything (plus game streaming, yet another avenue for reaching gamers). That’s all obvious enough, but the bigger change that was Microsoft promised to put all its in-house studio titles, even its biggest blockbusters (Halo, Gears, etc.) on the services on day one, meaning day and date with the versions they sell outright. This was a huge change, and it transformed Game Pass from something nice, and useful, into a no-brainer for many fans. All those games, one reasonable monthly price.

I assume Microsoft has negotiated with developers to get bigger, better, more recent games on Game Pass, and even day one AAA titles. But I don’t know anything about that, the math that would involve, and the very real possibility that it would just never make sense for either side. But that’s sort of the point of the virtuous cycle they are trying to create there: If Game Pass is so successful that Microsoft could afford to pay developers enough for day one AAA titles that it would obviate their fears of losing game purchase revenues, then it would feed a cycle of goodness in which more developers would want to jump on that train. We know this didn’t happen because Microsoft reneged on its day one promises on its most popular Game Pass tier and raised prices where it did not. But that’s the dream.

Anyway. It all starts and ends with Microsoft. But the process is opaque.

My other question is why do you think Microsoft has not brought free games from Blizzard to Games Pass they own them anyway. I mean Starcraft 2 is free to play why not bring it to Game Pass people might not know it’s free. World of War Craft could beef up their PC game pass offering.

I can only speculate because of the silence.

Activision Blizzard was a big, complex company with multiple studios, each of which supported whatever range of platforms across PC, Mac, Nintendo, PlayStation, Xbox, Android, and iPhone/iPad. Each of those studios made whatever deals, signed whatever license agreements, and Microsoft adopted all that, it can’t just unwind anything. Meaning, there are likely legal agreements that make this complex.

Above, I referenced the math of Game Pass. And Microsoft always discusses the ongoing costs of the Activision acquisition in its quarterly reports. Activision is a nice revenue bump, of course, but it’s also a huge cost center, and those costs far exceed the $68 billion it paid for the company. Some of the resulting transition costs will ease and go away. But many will never end: Just running that business, all those studios with all those employees, has a built-in run-rate. It’s expensive.

Obviously, Microsoft worked to determine the cost of putting whatever titles on Game Pass, also the cost/benefit math, if you will, of how this would (or would not) make sense financially. I have to assume the numbers were worse than they had predicted before the acquisition. If you think back to Microsoft’s Nokia acquisition a decade earlier, we later learned that Nokia had stopped developing new phones to save money after it agreed to be acquired, and that Microsoft was thus several months behind when it went through. In this case, it may have learned after the fact that there were costs–contracts, whatever–it didn’t know about. And that may have negatively impacted the Game Pass rollouts.

I don’t know. But whatever it was, I didn’t expect so little in this much time. In fact, I expected most of Activision’s catalog titles to be on Game Pass by mid-2024 at the latest. So there’s something going on there.

We know they are worried about Call of Duty but do they make that much off the Blizzard games again some of them are free to play already add them to game pass. People might like them and buy the upgrade pack. As pc gamer I think Microsoft missing out on the opportunity to upsell and grow the user base. Again many are FREE already it’s just adding it to the store.

Yeah. That’s the calculus right there. Call of Duty is easy, in a way, because you can see how much it makes when it’s there at retail, and what a hole it would be to lose that. You can figure out how many Game Pass subs they’d need to sell to counter that, but you also run into that issue where gamers would sign up for a few months, play the game, and then unsubscribe. Nothing is certain. You can predict, make projections, and just guess. And then what happens, happens. It’s not always what you expected.

My issue in all this is the same as ever. The silence. Dear God, Microsoft. Just tell us something. Give us some answer, some excuse, some justification. Anything.

It’s a tough year to be an Xbox fan. I don’t think anyone saw that coming. Not even Microsoft, apparently.