When you purchase through links on our site, we may earn an affiliate commission. JEDEC, the organization responsible for defining the specifications of industry-standard memory types, is close to finalizing SPHBM4, a new memory standard designed to deliver full HBM4-class bandwidth with a 'narrow' 512-bit interface, higher capacities, and lower integration costs by leveraging compatibility with conventional organic substrates. If the technology takes off, it will address many gaps in the markets HBM could serve, but as we'll explain below, it isn't likely to be a GDDR memory killer. Although high-bandwidth memory (HBM) 1024-bit or 2048-bit interfaces enable unbeatable performance and energy efficiency, such interfaces take a lot of precious silicon real estate inside high-end processors, which limits the number of HBM stacks per chip and therefore memory capacity supported by AI accelerators, impacting both the performance of individual accelerators as well as the capabilities of large clusters that use them. JEDEC doesn't specify whether '4:1 serialization' means quadrupling the data transfer rate from 8 GT/s in HBM4, or introducing a new encoding scheme with higher clocks. Still, the goal is obvious: preserve aggregate HBM4 bandwidth with a 512-bit interface. Inside, SPHBM4 packages will use an industry-standard base die (probably made by a foundry using a logic fabrication process and therefore not cheaper as routing 'wide' DRAM ICs into a 'narrow' base die will probably get nasty here in terms of density and there will be clocking challenges due to slow wires from DRAMs and fast wires from the base die itself). It will also use standard HBM4 DRAM dies, which simplifies controller development (at least at the logical level) and ensures that capacity per stack remains on par with HBM4 and HBM4E, up to 64 GB per HBM4E stack. On paper, this means quadrupling SPHBM4 memory capacity compared to HBM4, but in practice, AI chip developers will likely balance memory capacity with higher compute capability and the versatility they can pack into their chips, as silicon real estate becomes more expensive with each new process technology. An avid reader will likely ask why not use SPHBM4 memory with gaming GPUs and graphics cards, which could enable higher bandwidth at a moderate cost increase compared to GDDR7 or a potential GDDR7X with PAM4 encoding. Get Tom's Hardware's best news and in-depth reviews, straight to your inbox. Although cheaper than HBM4 or HBM4E, SPHBM4 still requires stacked HBM DRAM dies that are physically larger and therefore more expensive than commodity DRAM ICs, an interface base die, TSV processing, known-good-die flows, and advanced in-package assembly. While a 512-bit memory bus remains a complex interface, JEDEC says SPHBM4 enables 2.5D integration on conventional organic substrates and does not require expensive interposers, significantly lowering integration costs and potentially expanding design flexibility. Meanwhile, with an industry-standard 512-bit interface, SPHBM4 can offer lower costs (thanks to the volume enabled by standardization) compared to C-HBM4E solutions that rely on UCIe or proprietary interfaces. Compared to silicon-based solutions, organic substrate routing enables longer electrical channel lengths between the SoC and the memory stacks, potentially easing layout constraints in large packages and accommodating more memory capacity near the package than is currently possible. Still, it is hard to imagine routing of a 3084-bit memory interface (alongside data and power wires) using conventional substrates, but we'll see about that. Tom's Hardware is part of Future US Inc, an international media group and leading digital publisher.

Historically, TSA hasn't been involved in hunting down people for ICE, but that's apparently changed during the second Trump era. TSA now sends ICE lists “multiple times a week” of people who will be traveling through U.S. airports, and those are compared to databases of travelers the Trump regime claims are subject to deportation, according to the Times. We say “claims” because ICE has deported people who are not subject to deportation, the most famous case being Kilmar Abrego Garcia who a judge ordered released on Friday. The Times notes that it's unclear how many people have been arrested at airports under the new information-sharing program between TSA and ICE. But it appears that Any Lucía López Belloza, a 19-year-old abducted at Boston Logan Airport on Nov. 20 and deported to Honduras just a couple of days later, was napped using this system. Lopez Belloza, who had been in the U.S. since she was just 7 years old, was ferried out of the country despite the fact that a judge ruled she should not be deported on Nov. 21, according to a different report from the Times. Last year's ceiling was 125,000 refugees under President Joe Biden. The new report from the Times notes that one reason ICE agents like arresting people at airports is that they've already been screened for weapons. Masked federal agents have invaded many cities with their terrorizing tactics and it's a common question among social media pundits who ask why there haven't been more shoot-outs in the streets. Given America's love for guns, it's fairly baffling, though Trump's federal agents have shot unarmed people while sweeping cities to harass anyone who isn't white. During an Oval Office appearance on Dec. 2, Trump said people from Somalia were “garbage.” I don't want them in our country. I'll be honest with you, okay?” Trump said. I don't want them in our country. On Tuesday, Trump called nations like Afghanistan, Haiti, and Somalia “shithole countries,” a term he's used before. And while Trump's defenders will often insist it's about crime, his more Nazi-adjacent supporters know the score. His efforts to expel immigrants is about race. It's extremely unusual for TSA to be handing ICE lists of passengers so that they can run a deportation program. And Reuters reporter Brad Heath, whose beat is criminal justice, helped put it in perspective: “To give you a sense of how big a departure this year, the government doesn't do this for wanted criminals,” Heath wrote. Subscribe and interact with our community, get up to date with our customised Newsletters and much more. It's a little bit unclear what the tracker will look like, but it'll apparently become available in the next few weeks. "The district court erred by placing too much weight on statements the President made on social media."

When you purchase through links on our site, we may earn an affiliate commission. Following reports that Nvidia has developed a data fleet management software that can track physical locations of its GPUs, Nvidia on Thursday detailed its GPU fleet monitoring software. The software indeed enables data center operators to monitor various aspects of an AI GPU fleet. Among other things, it allows for detecting the physical location of these processors, a possible deterrent against smuggling chips. However, there is a catch: the software is opt-in rather than mandatory, which may limit its effectiveness as a tool to thwart smugglers, whether nation-state or otherwise. The software collects extensive telemetry, which is then aggregated into a central dashboard hosted on Nvidia's NGC platform. This interface lets customers visualize GPU status across their entire fleet, either globally or by compute zones representing specific physical or cloud locations, which means the software can detect the physical location of Nvidia hardware. Nvidia stresses that the software is strictly observational: it provides insight into GPU behavior but cannot act as a backdoor or a kill switch. As a result, even if Nvidia discovers via the NGC platform that some of its GPUs have been smuggled to China, it cannot switch them off. Nvidia's new fleet-management software gives data center operators a detailed, real-time view of how their GPU infrastructure behaves under load. These indicators help expose load imbalance, bandwidth saturation, and link-level issues that can quietly degrade performance across large AI clusters. By catching hotspots and insufficient airflow early, operators can avoid performance drops that typically accompany high-density compute environments and, in many cases, prevent premature aging of AI accelerators. The system also verifies whether nodes share consistent software stacks and operational parameters, which is crucial for reproducible datasets and predictable training behavior. Any configuration divergence, such as mismatched drivers or settings, becomes visible in the platform. For example, DCGM is a local diagnostic and monitoring toolkit that exposes raw GPU health data, but requires operators to build their own dashboards and aggregation pipelines, which greatly shrinks its usability, but enables operators to build the tools they need themselves. There is also Base Command, a workflow and orchestration environment designed for AI development, job scheduling, dataset management, and collaboration, not for in-depth hardware monitoring. Get Tom's Hardware's best news and in-depth reviews, straight to your inbox. Meanwhile, all three tools represent a formidable set of knobs for data center operators. DCGM provides node-level probes, Base Command handles workloads, and the new service integrates them into a fleet-wide visibility platform that scales to geographically distributed GPU deployments. Tom's Hardware is part of Future US Inc, an international media group and leading digital publisher.

When a derecho slammed into the Duane Arnold nuclear plant in 2020, Diana Lokenvitz had time for exactly one glance out her window. Within seconds of the storm hitting the plant, 130-mile-per-hour winds had severed all six of its external power lines, triggering an automatic emergency shutdown. With the reactor core still dangerously hot, safety systems kicked in to help stabilize the reactor and vent excess heat, a process that lasted for hours. Twelve water-cooling towers once watched over the plant like two rows of soldiers, releasing steam from water used to cool the nuclear reactor. The derecho, a thunderstorm characterized by high wind gusts spanning several hundred miles, had swept across the Midwest, causing widespread power outages and catastrophic damage to buildings, trees and millions of acres of crops. After 45 years of operation, the Duane Arnold Energy Center was shut down. The plant was already scheduled for decommissioning within months, leaving little time or financial incentive to repair the storm-damaged facility. An initial analysis by the federal Nuclear Regulatory Commission's (NRC) Division of Risk Analysis estimated there had been a 1-in-1,000 chance of damage to the nuclear core at Duane Arnold during the derecho. There were no “significant precursors,” the highest risk level event, during that period. Duane Arnold has sat dormant since 2020, but with a power-purchasing agreement between the Florida-based owner, NextEra Energy, and Google, which is expanding its fleet of data centers in Iowa, the plant is now slated for a 2029 reopening. In the fall, NextEra unveiled a partnership with Google, which already operates several data center buildings in Council Bluffs and is currently building a data center campus outside Cedar Rapids, about 12 miles from Duane Arnold. The tech giant has agreed to help cover recommissioning costs and to purchase the bulk of Duane Arnold's energy output for 25 years. The cooling towers at Google's data center in Council Bluffs, Iowa. Operators say that improvements to the original plant design, including additional backup diesel generators and cooling towers with greater wind resistance, will enhance its resilience to future severe weather events. In 2024 alone, Iowa experienced a record 155 tornadoes, topping the previous record of 146 tornadoes set just three years earlier, in 2021. The National Oceanic and Atmospheric Administration (NOAA) tracked weather and climate disasters with losses exceeding $1 billion in the state from 1980 to 2024. In that time, the state saw slightly less than two storms a year where damage estimates were higher than $1 billion, after adjusting the numbers for inflation. However, NRC reports following the incident noted damage to both safety- and non-safety- related structures. Even with heightened risk of damage to the nuclear core, a great deal more would have had to go wrong for Duane Arnold to reach meltdown or radiation release, said Adam Stein, director of the Nuclear Energy Innovation program at the Breakthrough Institute, a nonpartisan and nonprofit global research center. “This was one of the most significant safety events in US nuclear history and yet it was not significantly dangerous to the public,” said Stein. “That speaks to the robustness of these plants.” NRC licensing requires reactor buildings to be built to withstand “tornado missiles,” or large objects colliding at high speeds, Stein added. “They are literally designed to withstand these kinds of events safely,” he said. Still, NextEra plans to increase weather-related safety measures at the reopened Duane Arnold plant. “We do look at those events and try to garner lessons learned and ask, what could make the plant even safer than it is?” said NextEra consultant Michael Davis at a public information meeting held in Cedar Rapids by the Iowa Utilities Commission on Nov. 13. The company is considering installing a third diesel generator to provide additional backup power, and will also design Duane Arnold's replacement water-cooling towers with a higher wind resistance threshold, said Davis. Google representatives did not respond to questions about whether the damage incurred at Duane Arnold during the 2020 derecho raised any concerns for nuclear safety during severe weather events. The NRC safety requirements mandate that applicants “consider the most severe meteorologic and seismic conditions known in the proposed area,” when selecting reactor sites, an NRC representative wrote in a statement to Inside Climate News. If the facility hadn't already been slated for decommissioning when the derecho struck, Duane Arnold would have been rebuilt and continued to produce power, she said. In your inbox: WIRED's most ambitious, future-defining stories Big Interview: Palantir's CEO Alex Karp goes to war Starlink devices are allegedly being used at scam compounds Livestream: What businesses need to know about agentic AI WIRED may earn a portion of sales from products that are purchased through our site as part of our Affiliate Partnerships with retailers. The material on this site may not be reproduced, distributed, transmitted, cached or otherwise used, except with the prior written permission of Condé Nast.

It's a new feature Google announced this week called Android Emergency Live Video, and it's rolling out to folks in the US, along with select regions in Germany and Mexico, for phones running Android 8 and newer with Google Play Services. This capability becomes available during an emergency call or text, as a dispatcher can send a request to your phone to share live video. This is the bougiest thing about me, but I love being able to control my smartwatch or fitness tracker without having to poke or tap at the screen. It helped Apple Watch wearers with accessibility issues use gesture-based controls. This year, Apple introduced Wrist Flick to dismiss calls or silence timers with WatchOS 26. Now Google is introducing gesture controls on the Pixel Watch 4. These features are named (with no apparent shame whatsoever) Double Pinch and Wrist Turn. The Pixel Watch will give you hints as to when Double Pinch might be helpful. Back when Wear OS was called Android Wear, Google had a few gestures that let you flick your wrist away or toward your body to scroll through notifications and tiles. It's one of the company's weakest offerings financially, as reported by Bloomberg. Management of the division recently rolled over to Apple Health head Sumbul Desai after explosive allegations that Jay Blahnik, the vice president of fitness technologies, had created a toxic work environment. Personally, I'm waiting for Fitness+ to get rolled into 2026's rumored new Health+ app. But it doesn't look like Apple is shutting it down just yet, since this week Apple announced that Fitness+ has expanded to a whopping 28 new markets, including Chile, Hong Kong, and India. Hundreds of Fitness+ sessions will now be available in languages such as Spanish, German, and Japanese, and there's a new K-pop music genre available for workouts. These new features tally with other developments we've seen across Apple's lineup, like live translation and heart-rate monitoring in this year's AirPods Pro 3. Fitness+ might not be more popular than Peloton just yet, but as an integrated part of a revitalized software and hardware lineup, it could get there. Steely Sound: The Braque speakers have a hefty price to match their metallic make-up. If you have been yearning for a pair of cubist speakers for your glamorous living room, Sweden-based Nocs Design has you covered with its new Braque speakers. The ultimate rectangles are formed from a 50-pound base of Swedish steel, which, to give you some perspective, means that each Braque plinth weighs about the same as a 9-year-old child or a bag of concrete mix. A wooden cabinet on top that lump of metal contains a UK-based Celestion coaxial driver and an Estonian wood cabinet. Don't plan to use them with your flatscreen (the speakers lack HDMI), but they do have RCA, XLR, optical, and coaxial inputs. Part of what makes it one of my favorites is how easy it is to upload photos and invite family members and friends to the frame to share pics (unlimited photo storage is no small feature, either). You'll add their phone number, then they'll get a text message inviting them to start texting back with photos, which will automatically be added to the digital frames. If you have more than one frame on your account, as I do, it'll show them a list of all the frames available so they can choose which one to text. The whole process can be done within a few minutes. Big Interview: Palantir's CEO Alex Karp goes to war Starlink devices are allegedly being used at scam compounds Livestream: What businesses need to know about agentic AI WIRED may earn a portion of sales from products that are purchased through our site as part of our Affiliate Partnerships with retailers. The material on this site may not be reproduced, distributed, transmitted, cached or otherwise used, except with the prior written permission of Condé Nast.

This article is part of Gizmodo Deals, produced separately from the editorial team. We may earn a commission when you buy through links on the site. Apple's official store keeps the MacBook Air with M4 chip locked at its full $999 price without budging an inch, following the company's legendary resistance to discounting its own products. Amazon just shattered that pricing wall by dropping this ultraportable 256GB laptop to $749 which marks an all-time low that actually beats the Black Friday deals we saw just weeks ago. This 13-inch model packs the new M4 chip, 16GB of unified memory, and 256GB of SSD storage and offers professional-grade performance in a package that weighs just 2.7 pounds. The M4 chip is Apple's newest silicon technology and speeds up the MacBook Air so that it can handle tasks that would have needed a Pro model just a generation ago. With 16GB of unified memory, you can run multiple applications at once without any problems and you can also switch between them right away. The M4 chip's efficiency means that the laptop can run for a long time, whether you're editing spreadsheets or streaming 4K video – and it is not the case with Intel-based systems, which quickly run out of power when they're under load. The high pixel density makes text rendering look very clear, which makes it easier on the eyes when you have to write for a long time or review code. The screen can get up to 500 nits of brightness which makes it easy to see even in sunny outdoor settings or brightly lit offices where glossy screens usually become unusable mirrors. Apple Intelligence is built into the whole system, so you can use natural language to write emails, summarize long documents, and organize your thoughts. This local processing also makes sure that responses are instant, without the delay that cloud-based AI assistants cause which makes the experience feel more natural and conversational. There are two Thunderbolt 4 ports that can transfer data, display output, and charge through a single cable. There is also a separate MagSafe charging port that frees up those Thunderbolt connections for other devices. Four speakers with Spatial Audio create an immersive soundstage that makes movies and music sound surprisingly full for such a thin device, and three microphones pick up clear sound during calls without picking up too much background noise. Getting this M4 MacBook Air 256GB for $749 instead of $999 is a great deal for a laptop that competes directly with laptops from other companies that cost $1,500 or more.