I’d describe myself as a skeptic of the generative AI revolution—I think the technology as it currently exists is situationally impressive and useful for specific kinds of tasks, but broadly oversold. I’m not sure it will vanish from relevance to quite the extent that other tech fads like the metaverse or NFTs did, but my suspicion is that companies like Nvidia and OpenAI are riding a bubble that will pop or deflate over time as more companies and individuals run up against the technology’s limitations, and as it fails to advance as quickly or as impressively as its most ardent boosters are predicting.
Maybe you agree with me and maybe you don’t! I’m not necessarily trying to convince you one way or the other. But I am here to say that even if you agree with me, we can all celebrate the one unambiguously positive thing that the generative AI hype cycle has done for computers this year: the RAM floor for many PCs and all Macs is now finally 16GB instead of 8GB.
Companies like Apple and Microsoft have, for years, created attractive, high-powered hardware with 8GB of memory in it, most egregiously in $1,000-and-up putative “pro” computers like last year’s $1,599 M3 MacBook Pro or the Surface Pro 9.
This meant that, for the kinds of power users and professionals drawn to these machines, that their starting prices were effectively mirages; “pay for 16GB if you can” has been my blanket advice to MacBook buyers for years now, since there’s basically no workload (including Just Browsing The Web) that won’t benefit at least a little. It also leaves more headroom for future software bloat and future hobby discovery. Did you buy an 8GB Mac, and then decide you wanted to try software development, photo or video editing, CAD design, or Logic Pro? Good luck!
It’s been a big year for Windows running on Arm chips, something that Microsoft and Arm chipmakers have been trying to get off the ground for well over a decade. Qualcomm’s Snapdragon X Elite and X Plus are at the heart of dozens of Copilot+ Windows PCs, which promise unique AI features and good battery life without as many of the app and hardware compatibility problems that have plagued Windows-on-Arm in the past.
But Qualcomm has unceremoniously canceled the dev kit and is sending out refunds to those who ordered them. That’s according to a note received by developer and YouTuber Jeff Geerling, who had already received the Snapdragon Dev Kit and given it a middling review a couple of weeks ago.
“The launch of 30+ Snapdragon X-series powered PCs is a testament to our ability to deliver leading technology and the PC industry’s desire to move to our next-generation technology,” reads Qualcomm’s statement. “However, the Developer Kit product comprehensively has not met our usual standards of excellence and so we are reaching out to let you know that unfortunately we have made the decision to pause this product and the support of it, indefinitely.”
Qualcomm’s statement also says that “any material, if received” will not have to be returned—those lucky enough to have gotten one of the Dev Kits up until now may be able to keep it and get their money back, though the PC is no longer officially being supported by Qualcomm.
Whether you care about Microsoft’s Copilot AI assistant or not, many new PCs introduced this year have included a dedicated Copilot key on the keyboard; this is true whether the PC meets the requirements for Microsoft’s Copilot+ PC program or not. Microsoft’s commitment to putting AI features in all its products runs so deep that the company changed the Windows keyboard for the first time in three decades.
But what happens if you don’t use Copilot regularly, or you’ve disabled or uninstalled it entirely, or if you simply don’t need to have it available at the press of a button? Microsoft is making allowances for you in a new Windows Insider Preview build in the Dev channel, which will allow the Copilot key to be reprogrammed so that it can launch more than just Copilot.
There are restrictions. To appear in the menu of options in the Settings app, Microsoft says an app must be “MSIX packaged and signed, thus indicating the app meets security and privacy requirements to keep customers safe.” Generally an app installed via the Microsoft Store or apps built into Windows will meet those requirements, though apps installed from other sources may not. But you can’t make the Copilot key launch any old executable or batch file, and you can’t customize it to do anything other than launch apps (at least, not without using third-party tools for reconfiguring your keyboard).
Intel has formally announced its first batch of next-generation Core Ultra processors, codenamed “Lunar Lake.” The CPUs will be available in PCs beginning on September 24.
Formally dubbed “Intel Core Ultra (Series 2),” these CPUs follow up the Meteor Lake Core Ultra CPUs that Intel has been shipping all year. They promise modest CPU performance increases alongside big power efficiency and battery life improvements, much faster graphics performance, and a new neural processing engine (NPU) that will meet Microsoft’s requirements for Copilot+ PCs that use local rather than cloud processing for generative AI and machine-learning features.
Intel Core Ultra 200V
The most significant numbers in today’s update are actually about battery life: Intel compared a Lunar Lake system and a Snapdragon X Elite system from the “same OEM” using the “same chassis” and the same-sized 55 WHr battery. In the Procyon Office Productivity test, the Intel system lasted longer, though the Qualcomm system lasted longer on a Microsoft Teams call.
If Intel’s Lunar Lake laptops can match or even get close to Qualcomm’s battery life, it will be a big deal for Intel; as the company repeatedly stresses in its slide deck, x86 PCs don’t have the lingering app, game, and driver compatibility problems that Arm-powered Windows systems still do. If Intel can improve its battery life more quickly than Microsoft, and if Arm chipmakers and app developers can improve software compatibility, some of the current best arguments in favor of buying an Arm PC will go away.
Intel detailed many other Lunar Lake changes earlier this summer when it announced high-level performance numbers for the CPU, GPU, and NPU.
Like Meteor Lake, the Lunar Lake processors are a collection of silicon chiplets (also called “tiles”) fused into one large chip using Intel’s Foveros packaging technology. The big difference is that there are fewer functional tiles—two, instead of four, not counting the blank “filler tile” or the base tile that ties them all together—and that both of those tiles are now being manufactured at Intel competitor TSMC, rather than using a mix of TSMC and Intel manufacturing processes as Meteor Lake did.
Intel also said it would be shipping Core Ultra CPUs with the system RAM integrated into the CPU package, which Apple also does for its M-series Mac processors; Intel says this will save quite a bit of power relative to external RAM soldered to the laptop’s motherboard.
Keep that change in mind when looking at the list of initial Core Ultra 200V-series processors Intel is announcing today. There are technically nine separate CPU models here, but because memory is integrated into the CPU package, Intel is counting the 16GB and 32GB versions of the same processor as two separate model numbers. The exception is the Core Ultra 9 288V, which is only available with 32GB of memory.
For a long time, Microsoft’s Surface hardware was difficult-to-impossible to open and repair, and devices as recent as 2019’s Surface Pro 7 still managed a repairability score of just 1 out of 10 on iFixit’s scale. 2017’s original Surface Laptop needed to be physically sliced apart to access its internals, making it essentially impossible to try to fix the machine without destroying it.
Now, iFixit has torn apart the most recent Snapdragon X-powered Surface Pro and Surface Laptop devices and has mostly high praise for both devices in its preliminary teardown video. Both devices earn an 8 out of 10 on iFixit’s repairability scale, thanks to Microsoft’s first-party service manuals, the relative ease with which both devices can be opened, and clearly labeled internal components.
Beneath the Surface
To open the Surface Laptop, iFixit says you only need to undo four screws, hidden beneath the laptop’s rubber feet; at that point, the bottom of the machine is only attached by magnets, rather than breakable retention clips. Opening the bottom of the laptop provides easy access to the battery and an M.2 2232 SSD. Labels inside the device indicate which screws need to be removed to replace which parts, and what kind of screwdriver you’ll need to do the job; scannable barcodes also make it easier to find repair manuals and parts on Microsoft’s site. Most other parts are easy to remove and replace once the bottom of the laptop is off.
The Surface Pro’s best repairability feature remains its easily accessible M.2 2232 SSD, present under a pop-off cover on the back of the tablet. From there, things get more difficult—accessing the battery and other components requires removing the screen, which is still held in place with adhesive rather than screws or magnets. This adhesive needs to be removed—iFixit cut it away with a thin plastic tool, and closing the tablet back up securely would likely require new adhesive to be applied. Once inside, the parts and screws are still labeled clearly, but you do need to remove the entire heatsink before you can replace the battery.
iFixit uses slightly different criteria for evaluating the repairability of laptops and tablets since tablets are more tightly integrated devices. So despite the identical repairability scores, the Surface Pro remains slightly more difficult to open and fix than the laptop; iFixit is just comparing it to devices like the iPad Air and Pro rather than other PC laptops, and the Surface Pro still looks better than other tablets by comparison despite the use of adhesive.
The teardown video didn’t detail exactly why iFixit knocked points off of each device’s repairability score, though iFixit took note of the soldered-down non-upgradeable RAM and Wi-Fi/Bluetooth modules. Both devices also use way more screws and clips than something like the Framework Laptop, which could also be a factor.
We’ve been using the new Snapdragon-powered Surface devices for a few days now, and we’ll have more thoughts to share about the hardware and its performance in the coming days.
For the vast majority of compatible PCs, Microsoft’s Windows 11 24H2 update still isn’t officially available as anything other than a preview (a revised version of the update is available to Windows Insiders again after briefly being pulled early last week). But Microsoft and most of the other big PC companies are releasing their first wave of Copilot+ PCs with Snapdragon X-series chips in them today, and those PCs are all shipping with the 24H2 update already installed.
For now, this means a bifurcated Windows 11 install base: one (the vast majority) that’s still mostly on version 23H2 and one (a tiny, Arm-powered minority) that’s running 24H2.
Although Microsoft hasn’t been specific about its release plans for Windows 11 24H2 to the wider user base, most PCs should still start getting the update later this fall. The Copilot+ parts won’t run on those current PCs, but they’ll still get new features and benefit from Microsoft’s work on the operating system’s underpinnings.
The wider 24H2 update rollout will also likely enable the Copilot+ PC features on Intel and AMD PCs that meet the hardware requirements. That hardware will supposedly be available starting in July—at least, if AMD can hit its planned ship date for Ryzen AI chips—but neither Intel nor AMD seems to know exactly when the Copilot+ features will be enabled in software. Right now, the x86 version of Windows doesn’t even have hidden Copilot+ features that can be enabled with the right settings; they only seem to be included at all in the Arm version of the update.
Unfortunately for Microsoft, the Copilot+ PC program (and, to a lesser extent, the 24H2 update) has become mostly synonymous with the Recall screen recording feature. Microsoft revealed this feature to the public without first sending it through its normal Windows Insider testing program. As soon as security researchers and testers were able to dig into it, they immediately found security holes and privacy risks that could expose a user’s entire Recall database plus detailed screenshots of all their activity to anyone with access to the PC.
Microsoft initially announced that it would release a preview of Recall as scheduled on June 18 with additional security and privacy measures in place. Microsoft would also make the feature off-by-default instead of on-by-default. Shortly after that, the company delayed Recall altogether and committed to testing it publicly in Windows Insider builds like any other Windows feature. Microsoft says that Recall will return, at least to Copilot+ PCs, at some point “in the coming weeks.”
Aside from the Copilot+ generative AI features, which require extra RAM and storage and a PC with a sufficiently fast neural processing unit (NPU), the main Windows 11 system requirements aren’t changing for the 24H2 update. However, there are older unsupported PCs that could run previous Windows 11 versions that will no longer be able to boot 24H2 since it requires a slightly newer CPU to boot.
Microsoft will be delaying its controversial Recall feature again, according to an updated blog post by Windows and Devices VP Pavan Davuluri. And when the feature does return “in the coming weeks,” Davuluri writes, it will be as a preview available to PCs in the Windows Insider Program, the same public testing and validation pipeline that all other Windows features usually go through before being released to the general populace.
Recall is a new Windows 11 AI feature that will be available on PCs that meet the company’s requirements for its “Copilot+ PC” program. Copilot+ PCs need at least 16GB of RAM, 256GB of storage, and a neural processing unit (NPU) capable of at least 40 trillion operations per second (TOPS). The first (and for a few months, only) PCs that will meet this requirement are all using Qualcomm’s Snapdragon X Plus and X Elite Arm chips, with compatible Intel and AMD processors following later this year. Copilot+ PCs ship with other generative AI features, too, but Recall’s widely publicized security problems have sucked most of the oxygen out of the room so far.
The Windows Insider preview of Recall will still require a PC that meets the Copilot+ requirements, though third-party scripts may be able to turn on Recall for PCs without the necessary hardware. We’ll know more when Recall makes its reappearance.
Why Recall was recalled
Recall works by periodically capturing screenshots of your PC and saving them to disk, and scanning those screenshots with OCR to make a big searchable text database that can help you find anything you had previously viewed on your PC.
The main problem, as we confirmed with our own testing, was that all of this was saved to disk with no additional encryption or other protection and was easily viewable and copyable by pretty much any user (or attacker) with access to the PC. Recall was also going to be enabled by default on Copilot+ PCs despite being a “preview,” meaning that users who didn’t touch the default settings were going to have all of this data recorded by default.
This was the version of Recall that was initially meant to ship out to reviewers this week on the first wave of Copilot+ PCs from Microsoft and other PC companies. After security researcher Kevin Beaumont publicized these security holes in that version of Recall, the company promised to add additional encryption and authentication protections and to disable Recall by default. These tweaks would have gone out as an update to the first shipments of Copilot+ PCs on June 18 (reviewers also wouldn’t get systems before June 18, a sign of how much Microsoft was rushing behind the scenes to implement these changes). Now Recall is being pushed back again.
A report from Windows Central claims that Recall was developed “in secret” and that it wasn’t even distributed widely within Microsoft before it was announced, which could explain why these security issues weren’t flagged and fixed before the feature showed up in a publicly available version of Windows.
Microsoft’s Recall delay follows Microsoft President Brad Smith’s testimony to Congress during a House Committee on Homeland Security hearing about the company’s “cascade of security failures” in recent months. Among other things, Smith said that Microsoft would commit to prioritizing security issues over new AI-powered features as part of the company’s recently announced Secure Future Initiative (SFI). Microsoft has also hired additional security personnel and tied executive pay to meeting security goals.
“If you’re faced with the tradeoff between security and another priority, your answer is clear: Do security,” wrote Microsoft CEO Satya Nadella in an internal memo about the SFI announcement. “In some cases, this will mean prioritizing security above other things we do, such as releasing new features or providing ongoing support for legacy systems.”
Recall has managed to tie together all the big Windows and Microsoft stories from the last year or two: the company’s all-consuming push to quickly release generative AI features, its security failures and subsequent promises to do better, and the general degradation of the Windows 11 user interface with unwanted apps, ads, reminders, account sign-in requirements, and other cruft.
Microsoft’s Windows 11 Copilot+ PCs come with quite a few new AI and machine learning-driven features, but the tentpole is Recall. Described by Microsoft as a comprehensive record of everything you do on your PC, the feature is pitched as a way to help users remember where they’ve been and to provide Windows extra contextual information that can help it better understand requests from and meet the needs of individual users.
This, as many users in infosec communities on social media immediately pointed out, sounds like a potential security nightmare. That’s doubly true because Microsoft says that by default, Recall’s screenshots take no pains to redact sensitive information, from usernames and passwords to health care information to NSFW site visits. By default, on a PC with 256GB of storage, Recall can store a couple dozen gigabytes of data across three months of PC usage, a huge amount of personal data.
The line between “potential security nightmare” and “actual security nightmare” is at least partly about the implementation, and Microsoft has been saying things that are at least superficially reassuring. Copilot+ PCs are required to have a fast neural processing unit (NPU) so that processing can be performed locally rather than sending data to the cloud; local snapshots are protected at rest by Windows’ disk encryption technologies, which are generally on by default if you’ve signed into a Microsoft account; neither Microsoft nor other users on the PC are supposed to be able to access any particular user’s Recall snapshots; and users can choose to exclude apps or (in most browsers) individual websites to exclude from Recall’s snapshots.
This all sounds good in theory, but some users are beginning to use Recall now that the Windows 11 24H2 update is available in preview form, and the actual implementation has serious problems.
“Fundamentally breaks the promise of security in Windows”
Security researcher Kevin Beaumont, first in a thread on Mastodon and later in a more detailed blog post, has written about some of the potential implementation issues after enabling Recall on an unsupported system (which is currently the only way to try Recall since Copilot+ PCs that officially support the feature won’t ship until later this month). We’ve also given this early version of Recall a try on a Windows Dev Kit 2023, which we’ve used for all our recent Windows-on-Arm testing, and we’ve independently verified Beaumont’s claims about how easy it is to find and view raw Recall data once you have access to a user’s PC.
To test Recall yourself, developer and Windows enthusiast Albacore has published a tool called AmperageKit that will enable it on Arm-based Windows PCs running Windows 11 24H2 build 26100.712 (the build currently available in the Windows Insider Release Preview channel). Other Windows 11 24H2 versions are missing the underlying code necessary to enable Recall.
The short version is this: In its current form, Recall takes screenshots and uses OCR to grab the information on your screen; it then writes the contents of windows plus records of different user interactions in a locally stored SQLite database to track your activity. Data is stored on a per-app basis, presumably to make it easier for Microsoft’s app-exclusion feature to work. Beaumont says “several days” of data amounted to a database around 90KB in size. In our usage, screenshots taken by Recall on a PC with a 2560×1440 screen come in at 500KB or 600KB apiece (Recall saves screenshots at your PC’s native resolution, minus the taskbar area).
Recall works locally thanks to Azure AI code that runs on your device, and it works without Internet connectivity and without a Microsoft account. Data is encrypted at rest, sort of, at least insofar as your entire drive is generally encrypted when your PC is either signed into a Microsoft account or has Bitlocker turned on. But in its current form, Beaumont says Recall has “gaps you can drive a plane through” that make it trivially easy to grab and scan through a user’s Recall database if you either (1) have local access to the machine and can log into any account (not just the account of the user whose database you’re trying to see), or (2) are using a PC infected with some kind of info-stealer virus that can quickly transfer the SQLite database to another system.
But even given all of that, Intel still provides the vast majority of PC CPUs—nearly four-fifths of all computer CPUs sold are Intel’s, according to recent analyst estimates from Canalys. The company still casts a long shadow, and what it does still helps set the pace for the rest of the industry.
Enter its next-generation CPU architecture, codenamed Lunar Lake. We’ve known about Lunar Lake for a while—Intel reminded everyone it was coming when Qualcomm upstaged it during Microsoft’s Copilot+ PC reveal—but this month at Computex the company is going into more detail ahead of availability sometime in Q3 of 2024.
Lunar Lake will be Intel’s first processor with a neural processing unit (NPU) that meets Microsoft’s Copilot+ PC requirements. But looking beyond the endless flow of AI news, it also includes upgraded architectures for its P-cores and E-cores, a next-generation GPU architecture, and some packaging changes that simultaneously build on and revert many of the dramatic changes Intel made for Meteor Lake.
Intel didn’t have more information to share on Arrow Lake, the architecture that will bring Meteor Lake’s big changes to socketed desktop motherboards for the first time. But Intel says that Arrow Lake is still on track for release in Q4 of 2024, and it could be announced at Intel’s annual Innovation event in late September.
Building on Meteor Lake
Lunar Lake shares a few things in common with Meteor Lake, including a chiplet-based design that combines multiple silicon dies into one big one with Intel’s Foveros packaging technology. But in some ways Lunar Lake is simpler and less weird than Meteor Lake, with fewer chiplets and a more conventional design.
Meteor Lake’s components were spread across four tiles: a compute tile that was mainly for the CPU cores, a TSMC-manufactured graphics tile for the GPU rendering hardware, an IO tile to handle things like PCI Express and Thunderbolt connectivity, and a grab-bag “SoC” tile with a couple of additional CPU cores, the media encoding and decoding engine, display connectivity, and the NPU.
Lunar Lake only has two functional tiles, plus a small “filler tile” that seems to exist solely so that the Lunar Lake silicon die can be a perfect rectangle once it’s all packaged together. The compute tile combines all of the processor’s P-cores and E-cores, the GPU, the NPU, the display outputs, and the media encoding and decoding engine. And the platform controller tile handles wired and wireless connectivity, including PCIe and USB, Thunderbolt 4, and Wi-Fi 7 and Bluetooth 5.4.
This is essentially the same split that Intel has used for laptop chips for years and years: one chipset die and one die for the CPU, GPU, and everything else. It’s just that now, those two chips are part of the same silicon die, rather than separate dies on the same processor package. In retrospect it seems like some of Meteor Lake’s most noticeable design departures—the division of GPU-related functions among different tiles, the presence of additional CPU cores inside of the SoC tile—were things Intel had to do to work around the fact that another company was actually manufacturing most of the GPU. Given the opportunity, Intel has returned to a more recognizable assemblage of components.
Another big packaging change is that Intel is integrating RAM into the CPU package for Lunar Lake, rather than having it installed separately on the motherboard. Intel says this uses 40 percent less power, since it shortens the distance data needs to travel. It also saves motherboard space, which can either be used for other components, to make systems smaller, or to make more room for battery. Apple also uses on-package memory for its M-series chips.
Intel says that Lunar Lake chips can include up to 32GB of LPDDR5x memory. The downside is that this on-package memory precludes the usage of separate Compression-Attached Memory Modules, which combine many of the benefits of traditional upgradable DIMM modules and soldered-down laptop memory.