Hardware

Who needs power outlets when you can charge off of animal labor? The next time someone asks you if a hamster running on a wheel can produce a measurable amount of energy, you can point them to one inventive young YouTuber who has proven that the answer isn’t just yes - it’s also enough energy to harvest. Flamethrower, a Spanish YouTuber whose channel is full of experiments and DIY projects, most recently took to the video platform to post about his experiment in turning his brother’s hamster’s wheel into a machine he could use to charge his smartphone. “After my parents prohibited me from genetically or cybernetically modifying it,” Flamethrower explained in the video, “I deemed its existence … unacceptably useless. So what did I do? Exploit it for energy production, of course.” As anyone who’s owned a hamster knows, midnight runs on a squeaky wheel can be obnoxious to say the least. Sleepless nights, combined with memories of cartoons where hamster wheels powered the contraptions of evil geniuses, gave the young maker an idea: Strap a turbine to the wheel and make that hamster earn its keep. Of course, it’s not as simple as just attaching a 5V electric motor to use as a turbine and wiring it up to a USB-C plug. Hamsters can’t exactly be cajoled onto a wheel for regular charging shifts, and even if they could be, keeping them running at a charging pace would be pretty much impossible. “Say you have one of those common 5V DC motors,” Flamethrower explained. “Unless you spin that at over 10,000 RPM you wouldn't even reach the standard 15-watt charging speed” of most modern smartphones. “Forget about quick charging,” he added. “The motor would probably melt before that.” To get around those limitations, Flamethrower turned to the CJMCU-2557 low-power energy-harvesting chip. Energy harvesters like the 2557 (albeit larger ones) are designed to boost and regulate tiny amounts of input power, from sources such as solar cells or generators, into a usable voltage suitable for charging components like a capacitor or, in this case, a single lithium-ion battery cell. After doing some wiring work and putting the contraption together, Flamethrower left it to run overnight, and woke to a battery with enough charge in it to provide juice to his phone, though the amount wasn’t substantial. “I haven't measured exactly how much battery it can charge,” Flamethrower told us in an email. “It's not a lot, so it certainly won't fill the entirety of a phone's battery in a single night, although of course the energy generated each day can vary a lot based on the hamster's mood.” “I did a very rough calculation and the current generated might be like around an amp when the fella is running,” Flamethrower added. In other words, as frenetic as a hamster can get, it’s still not producing that much energy. Regardless, the young inventor told us that he is still using the device, and that it’s proved to be fantastic for providing energy to charge the family's smartwatches. As for whether he might consider attaching a bigger battery, that might be a bit overkill, he told us. “In any case, the CJMCU-2557 is only meant to charge 1 cell,” Flamethrower said. “There are, admittedly, slightly higher capacity cells, but I use it frequently enough to not have to care about reaching their limit.” A useful invention, then, if not a bit of an on-the-nose realization of a science-fiction dystopia trope. Hey, at least the hamster is happy and putting all that necessary calorie burning to work for a good cause - if we all used our pets' episodes of the zoomies to charge our devices, think of the load on the grid we could save. ®

Shortwave radio enthusiasts are sure to know the problem: You're trying to tune in to your favorite global broadcast only to find that the signal is fuzzy. Is it you? Your equipment? It might just be the conditions in the ionosphere, which you'd know if you built this DIY device.  As detailed in a YouTube video and project writeup on Hackster, North Macedonian maker Mirko Pavleski built himself a wee machine that measures ambient RF energy across the shortwave band, using it as a rough proxy for propagation conditions. For those not in the know about shortwave radio frequencies between ~1 and 30 MHz, their ability to travel around the world is thanks to the ionosphere. When the layer of ionized gas that makes up the aptly-named ionosphere is thick and dense, it refracts more shortwave signals back to Earth, meaning shortwave signals bounce off the atmosphere more and arrive at receivers stronger and clearer.  A weaker ionosphere, meanwhile, means more signal escapes into space and that broadcast is harder to make out.  Knowing the level of ambient RF energy across the shortwave band can be useful information when you're a shortwave listener, in other words, and that's exactly what Pavleski's device displays. The build is relatively simple, with basic components like an Arduino Nano R3, a small OLED, and, most crucially, a CA3089 chip used to translate raw RF signals into data the Arduino can display in an intelligible form.  The CA3089 is the trickiest part of the build, as you'll need to follow Pavleski's circuit diagram included on Hackster to wire up the passive components needed to make the whole thing work.  The data that the CA3089 outputs is also fairly raw, with its built-in received signal strength indicator providing a logarithmic voltage output corresponding to the strength of the received RF signals. "I map the entire input from 0 to 1023, which means that the values on the display from 0 to 1024 correspond to a voltage from 0 to 5V," Pavleski explained, while noting that the Arduino code he included in the project can be altered to show a different value if desired.  However, you choose to have it display - you could even cut out the Arduino and attach an analog multimeter instead - Pavleski notes that it takes a bit of time to get a sense of what a particular reading means.  "After a few days of comparing the value of the instrument with the real received signal on the radio, we will know for sure when the propagation of the RF signals is bad, good or excellent, and that at the same moment when we look at the instrument," the Macedonian maker explained. While shortwave radio operation is the obvious use of a tool like this, we can imagine other handy uses of it too: Maybe you want to record ionosphere activity for space-related purposes, are trying to correlate RF radiation to your local weather, or are a data hoarder looking for a new source. No matter the purpose, here's a project that'll suit your niche needs. ®

We've all been there: You're doing maintenance on a Weyland-Yutani hauler dragging mineral ore back toward Earth, and there’s no terminal handy to tap into the MU/TH/UR AI to check ship systems. Lucky for you, one enterprising maker has created just the machine for the job. Okay, maybe the megacorporations, starships, androids, and hostile xenomorphs of the Alien film franchise aren't real, but the aesthetic popularized by the 1979 film and its successors has captivated plenty of people, including Jeff Merrick, who has a passion for building his own "cyberdecks," custom-designed computers that often mimic cyberpunk and retro sci-fi aesthetics.  The 1970s retro future aesthetic is perfect for an Alien porta-terminal build, which is where the PS-85 comes in.  Built as a “rugged barebones slate-style portable computer inspired by the Alien universe,” Merrick’s Typeframe PS-85 packs in a small LCD display, a 40% mechanical keyboard with hot-swappable switches, and a Raspberry Pi Zero 2 W - a low-power board with 512 MB of RAM - meaning it can't do too much, but it'll sure look cool doing it. "With a Pi Zero you're likely going to want to run Raspberry Pi OS Lite and have a command-line only interface," Merrick wrote on his PS-85 build page, where you can find instructions, 3D print files, and everything you need to build your own.  For those wondering about getting some more useful keycaps for yours, Merrick told us that any MX-compatible keycaps will work for the design if you want to forego all the Alien aesthetic for something a bit more useful. Merrick doesn't just take his look from retro science fiction, either: He's also a lover of retro computing, as seen in the PX-88, which was the predecessor to his newer PS-85 design. Based on the 1985 Epson PX-4, Merrick's PX-88 was a gift for his wife, who wanted a bare-bones cyberdeck for writing.  "I built the PX-88 first and it took a few months of working on it in my free time," Merrick told The Register in an email. "There was quite a bit of trial and error with learning CAD software and just figuring out how everything could fit together." The PS-85 went considerably quicker, Merrick told us. Whichever one you're interested in building, you're going to have to do it on your own - Merrick has no plans to sell completed units.  "I open-sourced the full plans and files so folks could build their own, and a few people have," Merrick told us, before giving advice that everyone who's monetized a hobby would likely give him. "I plan to keep it solidly as a hobby rather than a business, it's more fun that way." ®

FEATURE In an unassuming three-story office building in Cupertino, California, engineers from Amazon Web Services are busy trying to make networking inconspicuous. They work in windowless hardware development labs at the center of the structure, surrounded by a ring of office cubicles that afford a view of scarce parking spaces and perimeter tree cover. Their latest project, which The Register and several other publications agreed not to discuss in advance of the pending official announcement, may get some attention. But their networking ambition differs from the promotional goals of the AWS communications team. A network should be like a light switch, said Matt Rehder, VP of global network engineering at AWS, during a tour of AWS's Torre Avenue lab in late April. It should be something that just works. "No one really cares about the network at the end of the day," he said. "It serves a function. You care about it when it's broken. But otherwise you want it to be out of your way. So that's been our mental model for the last 15 years – how do we get the network out of the way?" Networking was broken for AWS in 2010, at least from a business perspective. James Hamilton, SVP and distinguished engineer at Amazon, said as much in a presentation titled, "Datacenter Networks are in my Way." "This was in the very early days of the cloud," Rehder explained. "But even at that time, with the growth of bandwidth we were seeing, it was very clear that the way networks had been built wasn't going to scale into the future and that something fundamentally different had to happen." Hamilton objected to the vertically integrated networking stack that slowed innovation and kept margins high for network equipment makers. He likened it to the mainframe business model, and said he preferred the server business model, where there's competition and open source software. Networks for AWS, Rehder explained, consist of three primary types of hardware: network devices, including switches and routers, built on application-specific integrated circuits (ASICs) that forward data from one port to another; optical transceivers, which send and receive light signals via laser; and cabling, which may be fiber-optic glass or copper wire. When AWS was being built out a decade and a half ago, the cloud biz decided it needed to take control of its network technology. "It's so foundational to what we built," said Rehder. "And so we decided we needed to start developing our own hardware and developing our own software." The company started small, working with third parties to develop network devices, iterating on that until the footprint of its homegrown technology covered its datacenters, its core network, and its border network.  What's unique about AWS, Rehder said, is that other network providers typically use one type of switching ASIC for their aggregation network, another for their core network, and another for their border network, because each has different needs in terms of memory, performance, and throughput. "They would all use different silicon for different switches," he said. "We use one for everything." The reason, he explained, is simplicity. "If you have one thing and you overly invest in making it really good, you're putting all of your energy into that hardware and software making it super-reliable," Rehder said. "It also helps us scale the network because when we're managing our supply chain or figuring out how to scale, we're not trying to balance all these competing SKUs." That does create some challenges, he admitted. "There's a good reason people use different types of switching ASICs because there is different functionality," Rehder explained. "And that's where controlling our own software really comes to play. We've effectively been able to remove the need for that custom silicon by being smart with our software and just finding creative ways to keep what needs to happen on the device in the hardware as simple as possible while still delivering great performance and functionality for our customers." The switches have specialized ASICs to maximize the efficiency of packet routing. The ASIC can move many millions of packets per second from one port to another, without going through a CPU. So now AWS has its own hardware, running on its own software, a version of Linux called NetOS. The company's current homegrown switch is capable of transmitting 51.2 terabits per second of traffic, via 64 ports operating at 800 gigabits per second. Within the next 12 months, its next generation switch will provide 102.4 terabits per second via 64 ports running at 1.6 terabits per second. "Everything runs the same operating system as well, which is super powerful for us," said Rehder. "From a security perspective, it means the code's all ours. We can scan it, we can fix bugs … we can patch and update our devices very, very regularly." Owning everything has allowed AWS to do some difficult things. As an example, Rehder pointed to the high precision time network that AWS released a few years ago. That required unique hardware and unique software that integrates with the company's Nitro server chip. The technology, he said, allows applications such as high-frequency trading and distributed databases to operate across long distances. "The only way we could have achieved that is because we were able to bring our own hardware and our own software and then do something that was unique to solve some AWS customer problem," he said. "The bigger problem we're trying to solve is how do we keep all the server clocks in sync," said Satish Vangala, director of network product development at AWS. "We had to build a dedicated network to ensure that we have the timing synchronization at microsecond accuracy for all the servers in our datacenter." AWS's network consists of about two million devices and about 50-60 million optical links and transceivers. It includes about 20 million kilometers of terrestrial and subsea fiber at the moment, which Rehder says is enough to reach from the Earth to the Moon and back 25 times. And that's just cable between buildings. If you measure the cable within its datacenters, the amount of cable is maybe an order of magnitude higher. One of the ways AWS has been improving its network recently has been through the deployment of hollow core fiber, which has been around for a while but only recently became something that could be manufactured at scale.  With normal fiber optic cable, light-based networking signals travel through the glass fiber. Hollow core fiber consists of a glass tube surrounding air or vacuum, which offers less refractive interference and allows light to travel at a speed closer to its natural limit. The result is a 30 percent reduction in latency, which Rehder says is significant, particularly for datacenter placement. He explained that when an AWS region is built and has, for example, three availability zones, the datacenters have to be near each other but not too close.  For subsequent expansion, latency between structures constrains building placement – it has to be low enough that customer applications in different datacenters within the same region behave as if they were located in the same place.  So hollow core fiber expands the potential resources available to AWS datacenters – in terms of land and power – by allowing structures to be placed within a larger radius. "We do have hollow core deployed in a few places now," said Rehder. "It's more expensive than the traditional fiber. But if it enables us to improve latency or better serve customers, in the grand scheme of things, the cost of the fiber is small when you look at the entire cost of datacenters, servers and network devices and everything else." Network improvements like hollow core fiber are necessary because the demand for bandwidth keeps growing.  Rehder said that the need for bandwidth has been growing throughout his career but more so in the last four or five years as generative AI services have taken off. "The accelerated server types tend to have three to four times the bandwidth needs of the more traditional CPU-based server types," he explained. "We're still using the same hardware and we're using the same software, but we're packaging it together in a different way." To get more servers with more bandwidth under one network in a datacenter with less latency between servers, AWS uses fewer networking devices in the path between two servers.  "So the UltraCluster network lets you scale out to be much larger than the other traditional network that we use," Rehder said. "Effectively a different network topology. Instead of having seven network devices in the path between any two points, it has five network devices in the path between the two points." As AWS has built more capacity and expanded its network, the scale of its operations has demanded innovation. Rehder explained that issues like physical cabling infrastructure – the number of connectors required and how they can be optimized for ease, speed of deployment, and reliability come into play. "When you're into fiber optic, it's not like the Ethernet port you plug in," Rehder explained. "With fiber optic cabling, because you're sending a signal, even though you may plug the cable in right, if it's not seated perfectly or if it's dirty at all, that can obscure the signal and that can reduce the reliability of it. It's a major challenge when you're operating at high scale and the latest technology is on the edge of what you can cleanly send and receive and everyone's trying to deploy that." "That's a big area of focus for us – not only building all this capacity but making sure it can be built quickly and then, once it's built, it runs extremely reliably," he said. One of the ways AWS tries to ensure a smooth setup is with a device called a firefly, a connector that looks a bit like an alien from the arcade classic Space Invaders. Its function is to verify a fiber signal path so that's not a variable when a new endpoint gets added. "Each of these will have a send and a receive," said Rehder, "and this basically takes the send and receive and loops it, so that when we get the fiber into the datacenter, it'll be connected to a network switch at the other side and it can send a signal and if you see the signal come back to itself, you can make sure the fiber path is clean. So when the client comes in you can just plug it in and it's good to go." When the network works – more than 99 percent of the time, usually – you may not even notice the engineering. ® Updated at 2118 to correct the title for Matt Rehder.

GPS spoofing, which sends fake satellite-like signals, and GPS jamming, which drowns receivers in noise, are increasingly serious problems. Researchers at Oak Ridge National Laboratory in Tennessee have created what they say is the most effective system yet for detecting GPS interference, which could help blunt such attacks. ORNL said Wednesday that a group of boffins led by researcher Austin Albright has developed a new portable device that can detect both spoofing, which sends fake signals that mimic GPS satellite signals to provide bad location data, and jamming, which simply floods GPS receivers with noise. The device can operate from a vehicle to detect attacks on commercial trucks and warn drivers, the lab said, and tests with the US Department of Homeland Security suggest it's sensitive enough to outperform industry-developed systems that already exist.  That sensitivity would be notable enough, but ORNL said that the device is able to do something else that no known GPS interference detector can: It's able to detect spoofing even when fake and real signals are equally strong.  The ORNL device also operates entirely independently of GPS: It doesn't even have a GPS-specific receiver or knowledge of expected GPS signals, according to the lab. Instead, it consists of just a couple of well-known pieces of equipment, namely a software-defined radio and an embedded GPU, and what ORNL said is a new mathematical radio frequency analysis method to separate legit signals from malicious ones. The GPU's role is simply to perform the math in real time to detect spoofs or jams.  "Trucking needs a solution that works without special conditions or dependence on a trusted reference source," Albright said of the new device in ORNL's writeup. "Ours is the best in the world."  With the successful testing of the device completed, Albright and his team are now looking at ways to make the thing cheaper to produce, which we can imagine might include replacing the GPU with something less in-demand by the AI industry.  GPS spam: Not just a problem for planes We've reported plenty on GPS spoofing and jamming at The Register, but most of our writing on the topic has focused on aviation, with issues like GPS spoofing rampant at multiple airports in India, disrupting a flight carrying European Commission President Ursula von der Leyen, and generally rising to the level of being a serious flight safety concern for aviators around the world.  ORNL acknowledged the problem of GPS interference in aviation in its writeup, and while the device could potentially help detect attacks against aircraft, the lab’s immediate focus appears to be protecting truckers moving goods across the US. As an example, ORNL pointed to an incident last year in which two tractor-trailer loads of tequila from a brand co-founded by celebrity chef and Flavortown mayor Guy Fieri and former Van Halen singer Sammy Hagar were stolen. GPS spoofing was used during the crime to keep those waiting for the estimated 24,000 bottles from getting suspicious that the trucks weren't on course.  Some of the booze was eventually recovered in California (it was supposed to be delivered to Pennsylvania), but not before Fieri said the company had to lay people off due to the losses.  While stolen tequila is bad, the same attacks could also be used to waylay or misdirect shipments carrying everything from personal packages to nuclear materials and other essential goods. "Everyone uses cargo monitoring with GPS tracking, whether for your personal packages, your pizza, or nuclear materials," Albright said, adding that the device would act like any other sort of alarm to alert a driver that something's amiss.  "Like a carbon monoxide alarm alerts you to an invisible danger, spoofing detection is critical to alerting us to a new invisible danger," Albright said. Drivers with one of the ORNL devices, for example, could get an alert, "know something bad is happening and call someone," potentially protecting the driver, their shipment, and people who would be harmed by its loss.  We reached out to ORNL to learn more about the future of the project, but the lab wasn't able to meet our deadline. ®

If you follow PC hardware prices, you’ll know AI demand has pushed memory prices higher as manufacturers prioritize memory for datacenters. To deal with that, you can pay through the nose, buy less memory, or ... try to build your own DRAM. Dr. Semiconductor is a YouTuber who joined the platform in February and has only published two videos so far. One shows his process of turning a backyard shed into a cleanroom tidy enough to make his own semiconductors, while the other shows him producing working sample memory cells. The YouTuber opens his second video on the making of his sample chips with an acknowledgment that the whole experiment has been driven by RAM prices skyrocketing last year and into the beginning of 2026, leading to vendor quotes that constantly fluctuate and even delays in broadband expansions caused by a lack of memory chips.  Against that backdrop is Dr. Semiconductor, whose video walks through the process inside a tiny cleanroom equipped to make DIY RAM cells. Like any good RAM manufacturing process, Dr. Semiconductor started by designing a 5x4 array of capacitors and transistors - barely a fraction of the number of cells on a modern DRAM chip - then transferring the design onto silicon, coating it with photoresist for patterning and etching, and doping parts of the chip with phosphorus to increase conductivity. He then built up microscopic layers one by one without any of the industrial automation available to Samsung, SK hynix, or Micron. After making a few small sample chips, Dr. Semiconductor tested them using his parameter analyzers. Some additional DIY manipulation was necessary to probe the tiny chips, as the 5x4 array is small enough that the device had to be measured with micromanipulators and extremely fine probe tips. "Because the devices are at the nanoscale you can't just attach regular wires," the self-described doctor explained. "So for testing purposes I have a number of micromanipulators with some incredibly fine probe tips in order to feed current and voltage into the device."  The tests showed that the DIY RAM cells all functioned as expected, though higher voltage led to punch through due to the source and drain portions of each RAM cell being just a single micron apart. This "shows the trouble with scaling," according to Dr. Semiconductor, but he said it won't be a problem unless he tries to feed higher voltage into his memory.  As for the capacitors in the homemade chip, each was found to be able to hold 12.3 picofarads, which was right in line with what Dr. Semiconductor hoped to achieve, describing it as "pretty close to the perfect ideal theoretical of a little less than 15 pF I designed for."  Unfortunately, the capacitors in the DIY RAM cells in his test chip were only able to hold a charge for around 4 milliseconds - far less than commercial RAM, which he noted can retain charge for more than 64 milliseconds. Not perfect, but for a first try at creating homemade DRAM in a backyard microfab, it's quite the start.  "This is awesome - first time ever RAM has been made at home," Dr. Semiconductor said in the video. We can't confirm that claim, but we're not aware of anyone else making RAM in their backyard shed.  That said, the fact that it's an array of just 5x4 cells means this RAM isn't going to meet any important memory metrics anytime soon.  "While you can store data on it, you can't run DOOM in it quite yet - this is just a few cells to prove it will work," Dr. Semiconductor explained. He's not done here, though. Next up is chaining a bunch of his homemade cells together to see if they can actually serve as memory for a PC. When the results of that test might be published is unknown, but don't get too excited: Just because one enterprising individual managed to fabricate homemade RAM cells doesn't mean a cottage (or shed, to be fair) computing industry is likely to pop up overnight.  AI memory demand may become more efficient, and there may be early signs of pricing pressure easing, but the shortage is unlikely to end anytime soon. Buckle up, or consider building your own fab. ®

GCHQ's cyber arm has entered the hardware game with its first device designed to prevent cyberattacks on display devices. Called SilentGlass, the small gadget's intellectual property is courtesy of the UK's National Cyber Security Centre (NCSC), and the signals intelligence agency licensed it out to UK-based Goldilock Labs to make it commercially available to all businesses and consumers. SilentGlass is the NCSC's first branded device to hit the market. Announced publicly on Wednesday, the HDMI and DisplayPort-compatible device has already been deployed across "government estates," for several years and is capable of protecting "most high-threat environments." Naturally, The Register had a bunch of questions, but the NCSC refused to answer any. Through the powers that be, however, we are reliably informed that beyond the information included in the NCSC's blog, these devices are equipped with hardware that identifies malicious traffic in the data channel, blocking the transfer between computer and display. We're also told that the SilentGlass gizmos are threat-agnostic, meaning they are capable of detecting any kind of nastiness and preventing it from reaching and ultimately altering or manipulating a display. Anything potentially malicious that travels between HDMI or DisplayPort connections and a monitor is blocked. You might be thinking "it's not every day we hear about monitors being pwned via HDMI," and you'd be right. You wouldn't be alone either. Since the NCSC announced SilentGlass, infoseccers have taken to social media to question the need for this device. However, it is understood there are legitimate attack paths that are both applicable to modern environments and have been abused by known attackers. Very little exists in the research literature about these kinds of attacks. A team based out of Montevideo's Universidad de la República published findings in 2024 about the potential for highly technical individuals to intercept the electromagnetic radiation emitted from HDMI cables and use deep learning algorithms to reproduce text intended to be displayed on a monitor. The team called the finding Deep-TEMPEST, an evolution of the TEMPEST analog signal interception phenomenon of yesteryear. But, as with all side-channel attacks, the real-world application is significantly different from a remotely exploitable software bug, for example.  Most organizations probably don't need to worry about highly motivated foreign spies lurking around their cables looking for electromagnetic emissions. However, for those safeguarding highly sensitive data within the context of critical national infrastructure operators, it's potentially a slightly more credible threat. In any case, SilentGlass devices are available to anyone who wishes to purchase one, starting today. Attendees of Black Hat or 44con back in 2012 may also remember NCC Group's presentations about the potential for exploiting vulnerabilities in HDMI's EDID and CEC parsers, as well as CDC and NEC protocols. Again, these are fringe cases of which we hear very little from real-world scenarios, outside a conference keynote. Despite the lack of published cases of these attacks, the NCSC believes external computer monitors are "a hugely attractive target" for adversaries, particularly those with an espionage focus. It did not mention China specifically, although that is the country most often associated with cyberespionage, in the context of the UK's four main adversaries - China, Russia, Iran, and North Korea. The timing of the launch also coincides with the agency's CEO, Richard Horne, declaring China "a peer competitor in cyberspace," within the context of a steady rate of nationally significant cyberattacks directed at the UK by nation-states. The org also said such attacks can be effective if the people behind them are looking to cause disruption, or generate some financial gains, which essentially implicates each of the other three countries. Ollie Whitehouse, the NCSC's CTO, said: "Display screens and monitors are everywhere in modern business environments, and the SilentGlass device will help protect previously vulnerable IT infrastructure with unprecedented ease. "Its development and commercialisation shows the impact that the NCSC can have, alongside industry partners, with an affordable and effective product now globally available.  "By helping to launch a UK company onto the global market with this world-class innovation, we are breaking new ground and helping to strengthen national prosperity." NCSC gave Goldilock Labs, in partnership with Sony UK, the license to produce and sell SilentGlass, which comes as separate devices - one for HDMI and another for DisplayPort, each protecting one cable only. The NCSC wouldn't tell us the price, so we're waiting on Golidlock to tell us more information on that front. Stephen Kines, co-founder of Goldilock Labs, said the device meets a security problem that to date has been "widely overlooked," as many have not viewed HDMI and DisplayPort connections as a serious security boundary. "What was once confined to national security environments is now being applied with a low-cost, easy-to-deploy solution for CNI and businesses where the same risks exist," he said. "SilentGlass is the first step in a wider effort to enforce behaviour at hardware interfaces before it reaches complex software. It reflects a shift toward treating physical connectivity as a point of control rather than an assumed trust boundary." ®

Framework, maker of modular and repairable laptops, has spruced its line-up with a completely redesigned 13-inch model sporting the latest Intel CPUs, new components for its 16-inch system, and a dock that lets users add devices like a desktop graphics card. The California-based biz, which champions the right to repair, detailed its latest gear at an event in San Francisco, saying that each is a direct response to the requests of Framework's user community. Framework Laptop 13 Pro is a redesign powered by Intel's Core Ultra Series 3 Processors, which the firm claims to delivers more than 20 hours of battery life when video streaming. The buyer can choose which ports they want via plug-in Expansion Cards, with options including USB-C, USB-A, HDMI, DP, and Ethernet. Memory is also upgradable thanks to the use of an LPCAMM2 module that is screwed flat onto the motherboard. The display is a purpose-built power-optimized 13.5in screen with touch support and 2880 x 1920 resolution, while the backlit keyboard is available in a range of layout, language, and color options. This model is available to order with first shipments due in June. Prices start at $1,199 for the DIY Edition (i.e. assemble yourself) or $1,499 for pre-built configurations. Owners of the Framework Laptop 16 get new component options in the shape of a one-piece haptic touchpad and one-piece keyboard options, plus a new Bezel color and a Ryzen 5 processor configuration, the latter available for pre-order today in a pre-built configuration from $1,599 and a DIY Edition from $1,249. This laptop already offers buyers a number of input combinations, with the latest haptic touchpad module centering the touchpad in a single rigid aluminum palmrest. Likewise, the one-piece keyboard essentially offers a more seamless fit with the rest of the laptop. Also coming soon for the Laptop 16 is the OCuLink Dev Kit. This is described as a modular adapter and dock system that lets users connect external hardware such as a desktop graphics card. The OCuLink Dock supports standard off-the-shelf PCIe cards like 100 Gbps NICs, video capture cards, and more. As its name suggests, this uses OCuLink, the cable version of PCIe, to connect to the laptop, with up to 128 Gbps bidirectional throughput. Users will need an OCuLink Adapter Board in their Laptop 16 Expansion Bay Shell to bring the PCIe interface to a connector on the rear of the system. Framework says this has been designed as a kit. So you get the core electronics, structure, and reference 3D-printable designs, meaning users can choose what to build around it. Sample images supplied by the firm do not look particularly elegant. Finally, another product previewed but not yet available is a simple wireless keyboard with an integrated touchpad. This can connect to a system via a Wired (USB-C), Bluetooth, or USB-A Dongle, presumably meaning it can be used with any PC and not just a Framework model. However, Framework says the circuit board containing the core electronics is a module it will also make available separately in the Framework Marketplace so that enthusiasts can build their own wireless keyboard designs. ®

Windows has always had a built-in portal to the very recent past: Task Manager's CPU usage meter. "The CPU number in Task Manager is a moving little obituary for the immediate past," explained former Microsoft engineer Dave Plummer, "Not what happened at the moment that your eyeballs landed on the row." Plummer wrote the original version of Task Manager, back when it was a lean, mean, process-killing machine rather than the considerably chubbier and cuddlier tool of today. He has since led viewers through a tour of the source code, admitting along the way that he left his telephone number in the comments while chasing a strange bug in how CPU numbers were being reported. That bug is the subject of his latest explanation. So how did Task Manager report the CPU percentage? The answer is complicated. Windows had no magical CPU usage value waiting to be read. Instead, Plummer's Task Manager was timer-driven. Every time the timer fired, the code asked the kernel for cumulative execution times and compared them to the previous sample. "For an individual process, the math essentially is the cumulative current CPU time minus the previous cumulative CPU time. And that gives you how much CPU has been consumed by that process during the interval between the samples," he said. "The per-process percentage is just that process's delta divided by the total delta." "Now," the veteran engineer continued, "if that looks like the sort of thing you write when you've been locked in an office too long with a copy of Petzold and a lot of coffee, that's because it basically is." "Petzold" refers to the books by Charles Petzold, which were an indispensable companion to Windows programmers in the 1990s, well into the 2000s and beyond. Many long-serving engineers are likely still to have a well-thumbed copy or two on their shelves. Elegant as the solution was, it had wrinkles. Occasional quirks in the Windows kernel's reporting could make percentages fail to add up to 100, prompting a festooning of asserts in the code - and a request to contact Plummer directly if CPU usage ever exceeded that figure. And then there is the problem of modern hardware. Plummer said: "Back in the day, the scheduler's time accounting and the processor's actual throughput were much more tightly coupled because the CPU clocks were comparatively static. On modern CPUs, though, the hardware is constantly changing gears. "A mostly idle core may be downclocked, parked, or dropped into a state of sleep, sipping power through a cocktail straw, and then the instant that real work shows back up, it can jump up to a much higher frequency or even turbo past its nominal clock." As Task Manager's accounting was fundamentally time-based, the work accomplished in any given interval varied wildly depending on what frequency the silicon was running at. "The meter isn't wrong, but it's measuring sort of occupancy rather than productivity." It was built for a simpler era, before CPU frequency scaling and throttling became the norm. "When the numbers [today] feel a little slippery, it's not because the tool is broken so much as the hardware stops being simple enough for a single percentage to tell you the whole story." Plummer told The Register: "My main impetus was to account for every cycle, making sure each was properly attributed to the right 'cost center' and then determining how much actual work occurred in that time window. That seems pretty accurate, and I guess more important, 'felt' right to me in term of what the machine was doing." He was unable to comment on how the modern iteration of Task Manager performed the trick, telling us, "I know how I would have done it but hate to assume!" ®

California's proposed legislation to put the burden of blocking 3D-printed firearms onto printer manufacturers could effectively sideline open source tools and create new surveillance concerns, digital rights activists argue. Advocates at the Electronic Frontier Foundation (EFF) say that such legislation could empower manufacturers to introduce restrictive policies affecting consumer choice. It could lead to widespread surveillance of users' printing activity, which they fear could lead to copyright lawsuits, if that data were shared with other companies looking to protect against 3D-printed spare parts, for example. The bill in question is AB 2047, the scope of which, on paper, appears strict. The primary goal is clear and simple: to require 3D printer manufacturers to use a state-certified algorithm that checks digital design files for firearm components and blocks print jobs that would produce prohibited parts. Federal law does not impose a blanket ban on making firearms for personal use, though ghost guns are subject to various federal and state restrictions, and the practice remains controversial nationwide. Gun crime rates in the US far outweigh those in all other developed countries, so introducing legislation to curb the easy manufacture of untraceable firearms will be seen as a positive initiative to many, particularly in regions where guns are more strictly regulated. However, Cliff Braun and Rory Mir, who respectively work in policy and tech community engagement at the EFF, claim that the proposals in California are technically infeasible and in practice will lead to consumer surveillance. In a series of blog posts published this month, the pair argued that print-blocking technology - proposals for which have also surfaced in states including New York and Washington - cannot work for a range of technical reasons. They argued that because 3D printers and other types of computer numerical control (CNC) machines are fairly simple, with much of their brains coming from the computer-aided manufacturing (CAM) software – or slicer software – to which they are linked, the bill would establish legal and illegal software. Proprietary software will likely become the de facto option, leaving open source alternatives to rot. "Under these proposed laws, manufacturers of consumer 3D printers must ensure their printers only work with their software, and implement firearm detection algorithms on either the printer itself or in a slicer software," wrote Braun earlier this month. "These algorithms must detect firearm files using a maintained database of existing models. Vendors of printers must then verify that printers are on the allow-list maintained by the state before they can offer them for sale. "Owners of printers will be guilty of a crime if they circumvent these intrusive scanning procedures or load alternative software, which they might do because their printer manufacturer ends support." Braun also argued that it would be trivial for anyone who uses 3D printers to make small tweaks to either the visual models of firearms parts, or the machine instructions (G-code) generated from those models, to evade detection. Mir further argued that the bill offers no guardrails to keep this "constantly expanding blacklist" limited to firearm-related designs. In his view, there is a clear risk that this approach will creep into other forms of alleged unlawful activity, such as copyright infringement. "This could look like Nintendo blocking a Pikachu toy, John Deere blocking a replacement part, or even patent trolls forcing the hand of hardware companies," wrote Mir. "Repressive regimes, here or abroad, could likewise block the printing of 'extreme' and 'obscene' symbols, or tools of resistance like popular anti-ICE community whistles." Braun and Mir have a list of other arguments against the bill. They say the algorithms are more than likely to lead to false positives, which will prevent good-faith users from using their hardware. Many 3D printer owners also have no interest in printing firearm components. Most simply want the freedom to print trinkets and spare parts while others use them to print various items and sell them as an income stream. That said, Gun Owners of California also opposes the bill, arguing that it does not target criminals, only innocent consumers and businesses. "Californians deserve policies that focus on criminal misuse – not sweeping mandates that expand bureaucracy and restrict lawful activity," it wrote in a response to the bill's introduction in February. Addressing the community behind 3D printer manufacturer and slicer provider Prusa Research, community manager Tommy Muszynski said the company is keeping a close eye on developments. "At Prusa, safety is obviously the highest priority," he said in a comment on Reddit. "We want everyone to have a safe experience in this hobby, but at the same time, we have always been firm believers in the 'right to repair' and the right for you to use the machine you bought however you see fit. "We've built our community on open source principles and the idea that your printer is a tool for your own creativity, not a device that should be locked down or surveilled." ®

It's taken nearly a full version number to get the pieces in order, but the long-awaited end of 486 chip support in the Linux kernel appears to be nigh with Linux 7.1's release later this year.  Slated for the 7.1 merge window is a patch that veteran Linux kernel contributor Ingo Molnar queued up at the end of March, but which went widely unnoticed until over the weekend. If merged, the patch would begin phasing out support for 80486-generation chips by removing the M486, M486SX, and MELAN configuration options from Kconfig, effectively preventing new upstream kernels from being configured specifically for 486-class systems. The change has been a long time coming, and would begin the process of removing processor architecture support from the kernel for the first time since 2012, when support for 80386 processors was removed.  As he noted in 2022 when first contemplating removal of 80486 chips from the kernel, Linux maestro Linus Torvalds was similarly unsentimental when killing the 386.  "I *really* don't think i486 class hardware is relevant any more," Torvalds said in 2022, noting that while some people may still operate 486 systems they aren't relevant from a kernel development standpoint. "At some point, people have them as museum pieces. They might as well run museum kernels." In other words, if you want to run an old piece of hardware, you're just gonna have to rely on an old version of the Linux kernel going forward from 7.1, finally.  That's not to say the kernel maintainers haven't been working on the 486 support purge for some time – Molnar first proposed dropping 486 support in April 2025. Molnar noted a discussion he had with Torvalds in the patch notes, which were published in late April 2025, where he reiterated his feeling that it was time to ditch 486 support because it was wasting developers' time. Molnar similarly justified the elimination of 486 support in the notes.  "We have various complicated hardware emulation facilities on x86-32 to support ancient 32-bit CPUs that very very few people are using with modern kernels," Molnar explained. "This compatibility glue is sometimes even causing problems that people spend time to resolve, which time could be spent on other things."  To get around that wasted time, Molnar originally proposed eliminating 486 support by requiring the next kernel version to require chips to support Time Stamp Counter and the CMPXCHG8B instruction, which aren't present in 80486-family chips and some 586 derivatives.  It's not clear what happened in the just shy of a year since this proposal was made, but the kernel.org discussion chain that began with those recommendations has been ongoing for a year, with Molnar making multiple rounds of changes to his proposal since then. As of this latest merge request, it appears simply cutting off the configuration options for 486-family chips has been chosen as the way forward. With the final release of the Linux kernel 7.0 due sometime in the next few months, 7.1 can be expected sometime in the middle of this year, though whether this 486-killing patch proposal finally makes the cut remains to be seen.  Either way, Molnar noted in his request, there's no recent kernel package that supports 486 chips, so "actual users should not be impacted" either way. "Legacy users can keep using older kernels," Molnar added. ®

Apple's American Manufacturing Program (AMP) is expanding, with new suppliers signed on to produce iPhone components - though those parts will still be shipped overseas for final assembly. Tim Apple may continue avoiding tariffs but he probably won't win a lot of brownie points with President Trump. The iBiz confirmed on Thursday that Bosch, fabless chip firm Cirrus Logic, TDK, and component maker Qnity Electronics have all committed to make parts for Apple devices in US facilities. Apple intends to pay the new AMP members out of a $400 million pot through 2030. The AMP program is part of a broader pledge to spend $600 billion on US manufacturing by the end of the decade, in a bid to appease the Trump administration and avoid tariffs on iPhones and other devices manufactured overseas. Apple has so far announced plans to produce servers, chips, and Mac Minis in the US.  iPhones, which Trump signaled he wants made in the US, remain conspicuously absent from today's announcement, as was the case in prior US manufacturing plans published in recent months. Apple has long maintained that it is unlikely to manufacture iPhones on US soil as it is impractical due to costs, supply chain complexity, and workforce requirements training.  The new AMP members will make iPhone parts that are then shipped to factories in India and China for final assembly. Specifically, TDK will soon begin making sensors in the US for iPhone cameras "shipped all over the world," said Apple. TSMC will begin manufacturing integrated circuits used in Apple sensors for things like crash detection, activity tracking, and elevation at its Washington state facility on behalf of Bosch. And Cirrus Logic will work with initial AMP member GlobalFoundries to establish new semiconductor process tech for use in manufacturing integrated circuits for Face ID hardware.  Qnity Electronics, for its part, will supply "materials and technologies essential for semiconductor manufacturing and advanced electronics," Apple said.  Apple claimed its initial AMP partners, including Broadcom, Corning, Samsung, and others, have already made progress expanding their US manufacturing base for Apple products, though it didn't detail how.  None of this amounts to a US-made iPhone, which might explain why Apple boss Tim Cook, criticized of late for his attempt to cultivate a relationship with the White House, was absent from the list of Silicon Valley leaders picked for the President's Council of Advisors on Science and Technology.  Apple didn't respond to a request for comment. ®

The 5500FP is a ternary CPU implemented on an FPGA. It's not very fast, but it makes it easier to experiment with computers that don't use binary. Independent researcher Claudio Lorenzo La Rosa recently published 5500FP: A 24-Trit Balanced Ternary RISC Processor. The paper is quite technical, but it's only seven pages long. It describes how to implement a ternary procesor on a conventional binary-based FPGA: There's no inherent reason computers have to encode everything in binary. The natural world is more subtle than just "on" or "off". Another way of encoding logic that's quite well-suited to digital electronics is ternary logic. This encodes numbers in trits instead of bits. A trit can hold one of three values, rather than the two represented by "on" or "off". A common version is balanced ternary, in which one trit can hold one, zero, or minus 1. The word "bit" is short for binary digit and was coined by the late American mathematician John W Tukey, who also came up with the word "software" and as an encore devised the Fast Fourier Transform algorithm. By extension, if a Binary digIT is a bit, then a TRInary digiT is a "trit". That's trinary as in having three discrete states, not the BOFH version. There is a long historical precedent for ternary computer logic. The great Donald Knuth is a fan; in volume 2 of The Art of Computer Programming, he called it "perhaps the prettiest number system." The University of Iowa's Douglas Jones offers the Ternary Manifesto, which explains how number encoding works and much more. Back in 2017 a team came up with a more specialized ternary FPGA chip, and a couple of years later a South Korean team looked into wafer-scale fabrication of ternary components. Ternary computers have been built in the past. At Moscow State University in the late 1950s – around the time the late Sir Tony Hoare was there – a team led by Sergei Sobolev and Nikolay Brusentsov built a ternary machine, the Setun Computer. A lot has been written about Setun - much is in Russian - but the Road Not Taken – Setun, the Cold War, and the Lost Future of Non-Binary Computing is interesting. Brusentsov co-wrote a technical description, Ternary Computers: The Setun and the Setun 70 [PDF], which is in English. Setun inspired the later American TERNAC machine, which was also ternary – but it only existed as a simulation in Fortran on top of a Burroughs binary-based machine. TERNAC's developer, Gideon Frieder, published two oft-cited 1972 papers from the project: Ternary computers part I: motivation for ternary computers and Ternary computers part 2: emulation of a ternary computer. One of the reasons that Setun didn't lead to more ternary hardware is that the underlying implementation used binary logic, with two gates holding each trit. This is easier to implement, but it's inherently wasteful: one trit can hold approximately 1.58× more data than one bit, but using two bits wastes that advantage – you could be storing a third more data in binary (two bits can represent four states, while one trit only represents three states.) However, this is also how the 5500FP processor does it, both for the good reason of using off-the-shelf logic parts and also to make it easier to interface with basically any and all other existing computer hardware, which is all binary-based. La Rosa sees great potential in the idea, and has a whole website about it: Ternary Logic CPUs – Performance and efficiency in the third Millennium. ®

KETTLE It's The Most Wonderful Time of the Year - if you're an AI aficionado, that is, as chip giant Nvidia, now the most valuable company in the world, is kicking off its GPU Technology Conference (GTC) on Monday. For this week's episode of The Kettle, Brandon Vigliarolo talks to systems editor Tobias Mann and US editor Avram Piltch about what we can expect this Monday through Wednesday at GTC.  To be brief, gamers are probably going to feel left out since Nvidia seems to have decided renting cloud rigs to them is better than selling consumer hardware, small companies looking for AI chip compromises will be excited, and agentic AI is gonna be so hot that our Mann on the ground this week in San Jose isn't gonna need a jacket. We're making these predictions in advance of any official word from Nvidia, so we'll see if Tobias ends up being right or wrong - either way, keep an eye out for his reactions during and after GTC.  You can listen to The Kettle here, or find us on Apple Podcasts and Spotify. ®

When patient care is delayed in a hospital because something is broken, biomedical technicians would like you to understand that it's not usually their fault. Right-to-repair advocates from the Public Interest Research Group published the results of a survey of biomedical equipment technicians (BMETs) on Thursday that found widespread frustration at the tendency for equipment manufacturers to withhold repairability information and equipment, leading to delays in fixes.  Most of the BMETs surveyed (83 percent) said delays in receiving parts, service keys, manuals, and other necessary repair materials “somewhat frequently” or “most of the time” increased equipment downtime. The number is similarly high (70 percent) for those who said diagnostic tool restrictions "commonly" delayed prompt patient care.  According to the study, the most common repair restrictions that BMETs encounter are OEMs' refusal to provide passwords or service keys needed to read diagnostic information, and limiting access to and overcharging for training courses they require BMETs to pass before getting access to certain info and gear.  So, why not just go with an OEM contract for equipment maintenance? Most hospitals have OEM contracts, but not for everything, PIRG right-to-repair campaign senior director Nathan Proctor explained.  "Every hospital has a mix of all three -- they have OEM contracts, they have in-house people, and they have certain things they work with independent service organizations [ISOs] on," Proctor told The Register in an email. "The whole mix of health technology management is complex and involves a range of routine maintenance, serious device failures, and in-depth repairs, IT support, and everything in between." Relying on OEMs for timely repairs isn't a great idea.  "OEM staffing shortages have become a huge issue," Proctor told us. "OEMs have not been hitting the timeline stipulated in their contracts."  Medical equipment used in hospitals has been exempted from several recently passed right-to-repair laws, PIRG noted in the report. PIRG is hoping to change that. The agricultural equipment industry is finally being forced to come around through a combination of government lawsuits and state-level bills specific to them, with PIRG and other groups pushing those efforts. In this case, there's an even more pressing reason to make changes to healthcare equipment repairability, Proctor told us.  "Oftentimes there is tension between things that cut costs and things that improve quality in health care," Proctor opined. "Right to Repair is one of the rare things that both cuts costs and improves the quality of care. It's past time we enacted these reforms." ®

The US Army's attempt to turn Microsoft HoloLens headsets into battlefield kit may have failed, but the AR goggles aren't going into the garbage. Instead, they're being repurposed for remote cargo inspection support. When it comes to ensuring that pallets of military equipment are properly load-balanced for air transport, there's no one better suited to confirm requirements have been met than the US Air Force. Unfortunately, there's no way for delicate airmen to be everywhere Army grunts are stuck (this vulture is an Army veteran), and even if they're allowed to jump out of airplanes you may not want the average ground-pounder securing cargo that's supposed to actually complete its flight.  Faced with that conundrum, the Air Force and Army teamed up to use HoloLens headsets in one of the ways Redmond actually intended, by letting someone qualified see through the eyes of a peon.  Members of the 724th Air Mobility Squadron, based in Aviano, Italy, have been working with soldiers from the Army's 173rd Airborne Brigade, based in Vicenza, Italy, to give airmen a chance to help soldiers inspect equipment pallets prior to loading onto aircraft.  "We spent a year working with the manufacturer [Microsoft] and experimenting with different add-ons to figure out the right software and process we needed to get to where we are today," said 725th Air Mobility Squadron superintendent, Chief Master Sgt. Anthony Sewejkis. "Now it's plug and play. We can connect [from] anywhere just using the HoloLens, a Wi-Fi hotspot and a laptop." Airmen on laptops were able to see what soldiers were looking at, highlighting areas that needed attention using visual cues to direct soldiers to adjust rigging, reposition cargo, or whatever else needed doing. The augmented reality capabilities afforded to the troops by HoloLens headsets, 724's parent org the 521st Air Mobility Operations Wing noted, "increase the speed of maneuver to sustain joint force lethality across the competition continuum."  In other words, they think it's pretty cool.  Second time's a charm? The inspection project, which was just a proof of concept but one the Squadron intends to keep refining, was a far greater success than the Army's first foray into making use of the HoloLens. It all started back in 2018, when the Army awarded Microsoft a contract to build the Integrated Visual Augmentation System (IVAS) - custom HoloLens-derived headsets meant to give soldiers a battlefield heads-up display. Despite a deal described as worth up to nearly $22 billion over a decade, testing didn't go smoothly. A Pentagon watchdog report found the goggles caused “mission-affecting physical impairments,” including headaches, eyestrain, and nausea, among many of the soldiers who tested them. Attempts to continue the program, by dumping millions more into it, were met with resistance from Congress, which slashed funding for the program. Microsoft ultimately gave up, passing control of the initiative to Oculus inventor Palmer Luckey's Anduril. Anduril in turn brought Meta into the mix before the Army ultimately relaunched the program under a new name as the Soldier Borne Mission Command (SBMC) program, with the intention of developing an entirely new headset that hopefully isn't such a headache for soldiers.  It doesn't help for Microsoft's prospects as a supplier of military mixed reality headsets that the company canned development of the HoloLens in late 2024, with support for existing models slated to run through the end of 2027. It's not clear whether the headsets used in the cargo loading experiment were leftovers from IVAS, discounted leftovers acquired by the Air Force, or if they were purchased specifically for the project. We reached out to the Army and Air Force to learn more, but didn't hear back. ®

Intel continues to lose market share to rival AMD across server, desktop, and mobile processors, and this has been noticeable in PCs thanks to supply constraints on Chipzilla's processors. The latest figures from PC component watcher Mercury Research, covering Q4 of 2025, indicate that AMD now accounts for about 36 percent of the CPUs going into desktop PC systems, quite a jump from just under 27 percent for the same time last year. With mobile chips for laptops, the gain isn't quite so noticeable, with 26 percent coming from AMD in Q4, up from 23.8 percent for the same period a year ago. The flip side of that is that Intel is still filling 74 percent of mobile PCs. This is despite Intel suffering supply constraints on client processors due to a decision the company made earlier in the year to reallocate manufacturing capacity to output more server chips. This hurt its ability to build and sell processors for PCs. As a consequence, total x86 processor unit shipments declined during the last quarter of 2025, which was unusual as the fourth quarter would normally see the highest sequential growth in unit shipments for the year, Mercury says. But it wasn't all going AMD's way as shipments of the firm's gaming SoCs declined as expected. This is because we are said to be nearing the end of the current console cycle, so demand drops off in anticipation of new models. Server CPU shipments were up strongly during Q4, according to Mercury, with Intel's deliveries growing at nearly double the seasonal average, but AMD saw growth at more than triple the seasonal average. For AMD, its 5th Gen Epyc chips accounted for more than 50 percent of server revenues for the first time during the quarter, and Intel's Emerald Rapids 5th Gen family is also believed to have taken the lead from the Sapphire Rapids 4th Gen portfolio to become Intel's best-selling server parts. When it comes to share, Intel still takes the lion's portion at just over 71 percent of the server processor market. But AMD can take solace as its figure of 28.8 percent means it is closing in on a third market share, an increase of 3.1 percentage points from the 25.7 percent it stood at a year ago. So it doesn't look too bad for Intel, all things considered, which is odd as the company went to the trouble of sending us comments relating to these latest Mercury Research figures. For the record, Chipzilla wants you to know that it is "leaning hard into a simplified, accelerated roadmap – backed by strong server demand and major client portfolio updates – to stabilize and strengthen share as we move further into 2026." Mercury says that there is a larger than usual uncertainty in its estimates covering Arm-based CPU shipments into the PC market for this quarter. It believes that these systems have declined to 13.3 percent in Q4 2025 from 13.7 percent in the preceding quarter. However, the total share, including Arm servers against all x86 shipments, is estimated to stand at 12 percent, up from 11.7 percent earlier in the year. ®

Developers looking to gain a better understanding of machine learning inference on local hardware can fire up a new llama engine. Software developer Leonardo Russo has released llama3pure, which incorporates three standalone inference engines. There's a pure C implementation for desktops, a pure JavaScript implementation for Node.js, and a pure JavaScript version for web browsers that don't require WebAssembly. "All versions are compatible with the Llama and Gemma architectures," Russo explained to The Register in an email. "The goal is to provide a dependency-free, isolated alternative in both C and JavaScript capable of reading GGUF files and processing prompts." GGUF stands for GPT-Generated Unified Format; it is a common format for distributing machine learning models. Llama3pure is not intended as a replacement for llama.cpp, a widely used inference engine for running local models that's significantly faster at responding to prompts. Llama3pure is an educational tool. "I see llama3pure as a more flexible alternative to llama.cpp specifically when it comes to architectural transparency and broad hardware compatibility," Russo explained. "While llama.cpp is the standard for high-performance optimization, it involves a complex ecosystem of dependencies and build configurations, llama3pure takes a different approach." Russo believes developers can benefit from having an inference engine in a single, human-readable file that makes evident the logic of file-parsing and token generation. "The project's main purpose is to provide an inference engine contained within a single file of pure code," he said. "By removing external dependencies and layers of abstraction, it allows developers to grasp the entire execution flow – from GGUF parsing to the final token – without jumping between files or libraries. It's built for those who need to understand exactly what the hardware is doing." Russo also sees utility for situations where the developer is running legacy software or hardware, where client-side WebAssembly isn't an option, and where having an isolated tool without the potential for future dependency conflicts might be desirable. The C and Node.js engines, he said, have been tested with Llama models up to 8 billion parameters and with Gemma models up to 4 billion parameters. The main limiting factor is the physical RAM required to host model weights. The RAM required to run machine learning models on local hardware is roughly 1GB per billion parameters when the model is quantized at 8 bits. Double or halve the precision and you double or halve the memory required. Models are commonly quantized at 16 bits, so for a 1 billion-parameter model, 2GB would typically be required. According to Russo, the calculation for GGUF weights is different. "GGUF weights are loaded directly into RAM, which usually means the RAM usage matches the entire file size," he explained. "You can reduce the context window size by passing a specific parameter (context_size) – a feature supported by most inference engines, including the three I designed. While reducing the context window size this is a common 'trick' to save RAM when running models locally, it also means the AI won't 'remember' as much as it was originally designed to." He also said that llama3pure is presently focused on single-turn inference. He expects to implement chat history state management at a later date. For daily work, Russo says he uses Gemma 3 as a personal assistant, powered by his C-based inference engine, to ensure that sensitive data is handled privately and offline. "For a coding assistant, I recommend Gemma 3 27B," he said. "Regarding the latency concerns, while local models were historically slow, running optimized versions on modern hardware now provides an experience very close to cloud-based models like Claude and without the need to pay for such a service." While Russo expects common general use cases for AI assistance will continue to rely on cloud-hosted models, he foresees developers and businesses looking increasingly at local AI. While developer machines with 32GB or 48GB of RAM may lack the context window available with cloud-hosted models, they provide security and privacy without being dependent on service providers. Asked how he feels as a developer about the AI transition, Russo said he expects developers to eventually transition to AI supervisors. "Since AI models present answers with high confidence – even when incorrect – a human expert must remain in the loop to verify the output," he said. "Technical knowledge will not become obsolete; rather, it will become increasingly vital for auditing AI-generated work.  "While job titles may change, senior developers will always be necessary to maintain these systems, creating a workflow significantly faster than human-only development. For junior and mid-level developers, AI offers the opportunity to learn faster than previous generations. If managed correctly, AI can facilitate a significant leap in the industry's intellectual evolution." ®

It's no lightweight matter. DARPA is putting about $35 million in total funding on the table in the hope that it will spur researchers to work around fundamental physical constraints and build much larger-scale photonic circuits that do more of the computing with light, not electronics. A recent solicitation from the Department of Defense's research arm - the Photonic Integrated Circuit Architectures for Scalable System Objectives (PICASSO) - aims to scale photonics beyond today’s narrow demonstrations. Citing the need for more capable photonic systems, including those relevant to artificial intelligence workloads, DARPA is calling on researchers to submit proposals showing how circuit-level design can overcome the fundamental limitations that constrain current photonic computing approaches, using today’s photonic components rather than waiting for new ones to be invented. For those unfamiliar with photonic computing, it's the use of light instead of electrons to process and transmit data. If that sounds far-fetched, it's not, as DARPA points out - there are already photonic circuits around today, albeit in limited form. Using light to process data signals has advantages that are ideal for heavy workloads like AI due to greater bandwidth, less latency, and improved energy efficiency. Unfortunately, as DARPA points out, "systems incorporating photonic circuits struggle to show significant system-level performance advantages over electronic systems."  The current generation of photonic circuits is limited in depth, which restricts their ability to do much beyond single linear mathematical operations. Individual photonic circuits included in larger systems also have to convert optical signals to electronic ones in order to hand them off to other components, essentially eliminating the advantage of nanosecond latency due to the millisecond latency of electronic circuits, which DARPA notes is a 106 performance degradation.  So, what's holding photonic circuits back aside from, according to DARPA, an industry focus on component-level research? Why, just physics of course.  "The primary limitation to further scaling of circuit size and functionality is rooted in the fundamental properties of signaling with light," DARPA noted. Cool - easy fix, provided researchers can address the two fundamental technical challenges the research organization points out.  First there's signal degradation. Unlike metal-oxide-semiconductor circuits, which regenerate signals and filter out noise, photonic circuits have a fundamental issue of optical attenuation and noise that can't be amplified out, as any attempt to amplify optical signals also amplifies the noise.  Second, there's the issue of spurious wave interference, which leads to scattering, coupling, mode leakage, back reflections, and unwanted resonance. "Over many components, control of these errors becomes unpredictable, especially when combined with manufacturing variability and thermal and environmental instabilities," DARPA explained.  Overcoming those limitations, as mentioned above, has traditionally resulted in photonic circuits being interfaced with electronic ones, and DARPA doesn't want that.  "Heavy usage of electronics prevents system-level gains in latency, efficiency, and bandwidth offered natively by photonics," the agency said.  Here's where PICASSO comes in: What DARPA wants from its chosen participants is, like resolving the limitations imposed by the physics of light, simple: Just make better circuits so those two technical challenges aren't an issue anymore.  "Drawing inspiration from modern electronics, in which clever circuit design overcomes the limitations of individual transistors, the program will foster innovative circuit-level strategies to achieve unprecedented system performance and stability," DARPA said. "PICASSO confronts these challenges by embracing a new paradigm: creating tomorrow's photonic circuits using today's components." DARPA isn't just soliciting ideas, either. By the time PICASSO phase 1 wraps up, 18 months after it kicks off in July, DARPA wants a "demonstration of predictable performance of photonic circuits," and by the end of phase 2 (an additional 18 month period), it expects "generalized circuit functionality" to be demonstrated.   It also expects all that within a total program budget of about $35 million, spread across multiple awards. Proposals are due by March 6 - good luck. ®

Arduino boards power everything from robots to RGB lights, but they're a little on the small side. YouTuber UncleStem has his own solution: build a gigantic, yet fully functional one. In a video published last month on his fledgling YouTube channel, UncleStem built what he described as a seven-times-bigger-than-the-original Arduino Uno - though it seems larger than 7x - out of 3D printed parts, three layers of plywood to mimic the PCB and hide his wiring work, and, instead of an Arduino Uno, an Arduino Nano secreted inside to do the actual computing work.  "I wanted to hide the Arduino inside the board layers, and the Nano's profile is much slimmer than UNO's," UncleStem, who asked us to refer to him by his channel name, told us in an email. As for the motivation behind the project, UncleStem explains it not only as a fun hobbyist challenge (who wouldn’t want to have a fully functional giant Arduino to play with?), but also as a teaching tool, rather than a way to address one’s failing eyesight from too much squinting at microcontrollers. "If you've ever tried teaching Arduino to a class, you know the struggle. Everyone's hunched over these tiny boards, trying to see where the pins are," the video voiceover explains. "With this? No problem. Hold it up, point to the pins, and boom - everyone gets it."  Making it fully functional just adds to the cool factor, and when we say that the giant Arduino is fully functional, we really mean that. Sure, most of the exterior parts, like the ATmega chip, resistors, capacitors, and power adapter, are all just 3D printed for appearance, but if it's at all interactive or blinky, then it works. That means functional LEDs, a working reset button, and a full set of giant-yet-operational GPIO pins.  For the LEDs, that means a 3D printed clear enclosure for regular-scale LEDs, and, for the reset switch, that means a large enclosure for a smaller button. GPIO pins are a whole other matter.  To make those, UncleStem got hold of some large spring contacts, wired each of the 32 pins up and connected them to the Arduino Nano housed inside the three-layer wood PCB, and tested them out with a multimeter to be sure they actually worked.  UncleStem tested his fully-assembled giga Arduino using, obviously, some custom-made giant LEDs and resistors that fit onto the giant GPIO pins without needing to be shimmed in place and some code written on a laptop and fed to the Arduino Nano. Per the video, it all worked like a charm.  The YouTuber told us that it took him about a month of work, mostly in the evenings - and god knows how much 3D printer filament was fed through his Bambulab P1S - to build it, but he's not done yet.  When asked whether small components would work with the board, UncleStem told us that standard-sized components would work - but why would you do that? "Other components like buttons, sensors, LCD displays etc. work just fine with the big board," he told us, "but doesn't make much sense due to scale mismatch."  Along with the functional giant resistors and LEDs, UncleStem said he has plans to work on other giant components as well.  "Maybe building giant sensors, motors, buttons, LCD displays - who knows? The possibilities are pretty endless," he noted in the video.  For those who want to build their own giant Arduino, be it as an educational aid or just a fun toy to have around the shop, UncleStem has published all his 3D printed component plans on Google Drive. The rest of the bits, and the know-how to get them working, are up to you. ®