Nvidia unveiled its next wave of AI processors at GTC on Tuesday, announcing Blackwell Ultra chips that will ship in the second half of 2025, followed by the Vera Rubin architecture in 2026. CEO Jensen Huang also revealed that its 2028 chips will be named after physicist Richard Feynman.

The Blackwell Ultra maintains the same 20 petaflops of AI performance as standard Blackwell chips but increases memory from 192GB to 288GB of HBM3e. Nvidia claims these chips can process 1,000 tokens per second -- ten times faster than its 2022 hardware -- enabling AI reasoning tasks like running DeepSeek-R1 models with 10-second response times versus 1.5 minutes on H100 chips.

Vera Rubin will deliver a substantial leap to 50 petaflops in 2026, featuring Nvidia's first custom Arm-based CPU design called Olympus. Nvidia is also changing how it counts GPUs -- Rubin itself contains two dies working as one chip. The annual release cadence represents a strategic shift for Nvidia, which previously introduced new architectures every two years before the AI boom transformed its business.

Huawei will no longer be able to produce or sell Windows-based PCs as Microsoft's supply license to the Chinese tech company expires this month, according to Chinese tech site MyDrivers. The restriction comes as Huawei remains on the U.S. Department of Commerce's Entity List, requiring American companies to obtain special export licenses to conduct business with the firm.

Richard Yu, executive director of Huawei's consumer business unit, said the company is preparing to pivot to alternative operating systems. Huawei had previously announced plans to abandon Windows for future PC generations. The Chinese tech giant will introduce a new "AI PC" laptop in April running its own Kunpeng CPU and HarmonyOS, alongside a MateBook D16 Linux Edition, its first Linux-based laptop.

An anonymous reader quotes a report from 404 Media: Something very strange is happening inside Super Nintendo (SNES) consoles as they age: a component you've probably never heard of is running ever so slightly faster as we get further and further away from the time the consoles first hit the market in the early '90s. The discovery started a mild panic in the speedrunning community in late February since one theoretical consequence of a faster-running console is that it could impact how fast games are running and therefore how long they take to complete. This could potentially wreak havoc on decades of speedrunning leaderboards and make tracking the fastest times in the speedrunning scene much more difficult, but that outcome now seems very unlikely. However, the obscure discovery does highlight the fact that old consoles' performance is not frozen at the time of their release date, and that they are made of sensitive components that can age and degrade, or even 'upgrade', over time. The idea that SNESs are running faster in a way that could impact speedrunning started with a Bluesky post from Alan Cecil, known online as dwangoAC and the administrator of TASBot (short for tool-assisted speedrun robot), a robot that's programmed to play games faster and better than a human ever could.

[...] So what's going on here? The SNES has an audio processing unit (APU) called the SPC700, a coprocessor made by Sony for Nintendo. Documentation given to game developers at the time the SNES was released says that the SPC700 should have a digital signal processing (DSP) rate of 32,000hz, which is set by a ceramic resonator that runs 24.576Mhz on that coprocessor. We're getting pretty technical here as you can see, but basically the composition of this ceramic component and how it resonates when connected to an electronic circuit generates the frequency for the audio processing unit, or how much data it processes in a second. It's well documented that these types of ceramic resonators are sensitive and can run at higher frequencies when subject to heat and other external conditions. For example, the chart [here], taken from an application manual for Murata ceramic resonators, shows changes in the resonators' oscillation under different physical conditions.

As Cecil told me, as early as 2007 people making SNES emulators noticed that, despite documentation by Nintendo that the SPC700 should run at 32,000Hz, some SNESs ran faster. Emulators generally now emulate at the slightly higher frequency of 32,040Hz in order to emulate games more faithfully. Digging through forum posts in the SNES homebrew and emulation communities, Cecil started to put a pattern together: the SPC700 ran faster whenever it was measured further away from the SNES's release. Data Cecil collected since his Bluesky post, which now includes more than 140 responses, also shows that the SPC700 is running faster. There is still a lot of variation, in theory depending on how much an SNES was used, but overall the trend is clear: SNESs are running faster as they age, and the fastest SPC700 ran at 32,182Hz. More research shared by another user in the TASBot Discord has even more detailed technical analysis which appears to support those findings. "We don't yet know how much of an impact it will have on a long speedrun," Cecil told 404 Media. "We only know it has at least some impact on how quickly data can be transferred between the CPU and the APU."

Cecil said minor differences in SNES hardware may not affect human speedrunners but could impact TASBot's frame-precise runs, where inputs need to be precise down to the frame, or "deterministic."

Longtime Slashdot reader UnknowingFool writes: Gamer's Nexus performed tests on the effect of removing legacy 32-bit PhysX on the newest generation of Nvidia cards with older games, and the results are not good. With PhysX on, the latest generation Nvidia was slightly beaten by a GTX 580 (released 2010) on some games and handily beaten by a GTX 980 (2014) on some games.

With the launch of the 5000 series, NVidia dropped 32-bit CUDA support going forward. Part of that change was dropping support for 32-bit PhysX. As a result, older titles that used it would perform poorly with 5000 series cards as it would default to CPU for calculations. Even the latest CPUs do not perform as well as 15-year-old GPUs when it comes to PhysX.

The best performance on the 5080 was to turn PhysX off however that would remove many effects like smoke, breaking glass, and rubble from scenes. The second-best option was to pair a 5000 series with an older card like a 980 to handle the PhysX computations.

MojoKid writes: AMD just launched its latest flagship desktop processors, the Ryzen 9 9950X3D. Ryzen 9 9950X3D is a 16-core/32-thread, dual-CCD part with a base clock of 4.3GHz and a max boost clock of 5.7GHz. There's also 96MB of second-gen 3D V-Cache on board. Standard Ryzen 9000 series processors feature 32MB of L3 cache per compute die, but with the Ryzen 9 9950X3D, one compute die is outfitted with an additional 96MB of 3D V-Cache, bringing the total L3 up to 128MB (144MB total cache). The CCD outfitted with 3D V-Cache operates at more conservative voltages and frequencies, but the bare compute die is unencumbered.

The Ryzen 9 9950X3D turns out to be a high-performance, no-compromise desktop processor. Its complement of 3D V-Cache provides tangible benefits in gaming, and AMD's continued work on the platform's firmware and driver software ensures that even with the Ryzen 9 9950X3D's asymmetrical CCD configuration, performance is strong across the board. At $699, it's not cheap but its a great CPU for gaming and content creation, and one of the most powerful standard desktop CPUs money can buy currently.

Apple has refreshed its MacBook Air lineup with the M4 processor, adding a new sky blue color option and reducing prices across the board. The 13-inch model now starts at $999, while the 15-inch begins at $1,199. Both models are available to order immediately and will ship on March 12.

The updated MacBook Airs feature the same thin design as previous generations but now include the 12-megapixel Center Stage webcam found in current MacBook Pro models. Both variants come with the M4 chip, aligning them with Apple's recent Mac Mini, iMac, and MacBook Pro refreshes.

Base configurations include an M4 with a 10-core CPU and 8-core GPU, 16GB of unified memory, and 256GB of storage. Customers can upgrade to a 10-core GPU (matching the base 14-inch MacBook Pro), 32GB of RAM, and up to 2TB of storage. A significant technical improvement is the support for two external 6K displays while keeping the laptop's lid open, addressing a limitation of previous Air models.

Apple today announced a significant update to its iPad Air lineup, integrating the M3 chip previously reserved for higher-end devices. The new tablets, available in both 11-inch ($599) and 13-inch ($799) configurations, deliver substantial performance gains: nearly 2x faster than M1-equipped models and 3.5x faster than A14 Bionic versions.

The M3 brings Apple's advanced graphics architecture to the Air for the first time, featuring dynamic caching, hardware-accelerated mesh shading, and ray tracing. The chip includes an 8-core CPU delivering 35% faster multithreaded performance over M1, paired with a 9-core GPU offering 40% faster graphics. The Neural Engine processes AI workloads 60% faster than M1, the company said. Apple also introduced a redesigned Magic Keyboard ($269/$319) with function row and larger trackpad.

Lenovo demonstrated "a laptop with a foldable screen and one that can get extra battery life from solar power," reports CNBC, emphasizing that "These laptops are just concepts, meaning they are not commercially available."

But "Lenovo, the world's biggest PC maker, has a history of showing off imaginative concepts with some becoming reality, so it's worth keeping an eye on what the Chinese technology giant is up to..." The latest concepts were unveiled at the Mobile World Congress trade show in Barcelona... When fully unfolded, the screen is an 18-inch display [on the Lenovo ThinkBook 'flip' concept]... The screen can then be folded in half horizontally to create two screens — one on the front and one on the back. The entire display can be folded down flat so the laptop turns into a tablet-like device.
Lenovo also showed off a Yoga Solar PC concept, reports Gizmodo, calling it "relatively thin and light" despite a solar panel in its lid with "a supposed 24% solar conversion rate": Lenovo claims they achieved this by maneuvering the gridlines you usually find on a solar panel behind the solar cells, offering more real estate for energy absorption... Lenovo's software showed the power accumulation at around 7 V when facing away from the sunlight and 12 V when facing toward it. It could get more when getting direct sunlight. Despite the presence of the solar panel, the laptop still weighs a little more than 2.6 pounds, which isn't out of the realm of what to expect from most modern laptops.

We should note that the panel isn't generating the required power to run the PC continuously. Lenovo claimed that 20 minutes of direct sunlight will transform into about one hour of video playback battery life. Depending on the CPU and battery, that could be 1/20 of the laptop's battery life.
CNBC had slightly different statistics for the laptop's battery life. "Lenovo said that the solar panels can absorb even ambient light in a person's surroundings to give a user an extra hour of laptop use at the end of an eight-hour work day..."

Microsoft has updated its CPU compatibility list for Windows 11 24H2, excluding pre-11th-generation Intel processors for OEMs building new PCs. The Register reports: Windows 11 24H2 has been available to customers for months, yet Microsoft felt compelled in its February update to confirm that builders, specifically, must use Intel's 11th-generation or later silicon when building brand new PCs to run its most recent OS iteration. "These processors meet the design principles around security, reliability, and the minimum system requirements for Windows 11," Microsoft says.

Intel's 11th-generation chips arrived in 2020 and were discontinued last year. It would be surprising, if not unheard of, for OEMs to build machines with unsupported chips. Intel has already transitioned many pre-11th generation chips to "a legacy software support model," so Microsoft's decision to omit the chips from the OEM list is understandable. However, this could be seen as a creeping problem. Chips made earlier than that were present very recently, in the list of supported Intel processors for Windows 11 22H2 and 23H2.

This new OEM list may add to worries of some users looking at the general hardware compatibility specs for Windows 11 and wondering if the latest information means that even the slightly newer hardware in their org's fleet will soon no longer meet the requirements of Microsoft's flagship operating system. It's a good question, and the answer -- currently -- appears to be that those "old" CPUs are still suitable. Microsoft has a list of hardware compatibility requirements that customers can check, and they have not changed much since the outcry when they were first published.

Arm will reportedly start making its own chips this year after signing Meta as a customer, according to the Financial Times (paywalled). TechCrunch reports: The chip is expected to be a CPU for servers in large data centers and can be customized for various customers. Arm will outsource its production. The first in-house Arm chip will be unveiled as early as this summer, the Financial Times also reported.

This is a notable change in strategy for the semiconductor company, which usually licenses its chip blueprints to companies like Apple and Nvidia. Making its own chips will turn some of its existing customers into competitors.

LibreOffice, the open-source office suite that began as StarOffice in 1985, has marked its 40th anniversary with new features that it says could transform how users interact with the software. At the FOSDEM 2025 conference, developers unveiled LibreOffice 25.2, which introduces browser-based functionality and real-time collaboration capabilities through a technology called conflict-free replicated data types.

A key development is ZetaOffice, a version built for the WebAssembly runtime that enables the full office suite to run inside web browsers across operating systems and CPU architectures. The project, which entered public beta last November, allows websites to embed LibreOffice applications with complete user interfaces for editing documents, spreadsheets and presentations.

While the browser-based version currently requires about a gigabyte of code and additional memory to run, developers at Allotropia are working to modularize the codebase for faster loading times. The software, released under the MIT license, can be controlled via JavaScript and operates without requiring an internet connection, unlike Google Docs or LibreOffice's existing Collabora Online version.

Today, the KDE Project announced the release of KDE Plasma 6.3, featuring improved fractional scaling, enhanced Night Light color accuracy, better CPU usage monitoring, and various UI and security refinements.

Some of the key features of Plasma 6.3 include:
- Improved fractional scaling with KWin to lead to an all-around better desktop experience with fractional scaling as well as when making use of KWin's zoom effect.
- Screen colors are more accurate with the KDE Night Light feature.
- CPU usage monitoring within the KDE System Monitor is now more accurate and consuming fewer CPU resources.
- KDE will now present a notification when the kernel terminated an app because the system ran out of memory.
- Various improvements to the Discover app, including a security enhancement around sandboxed apps.
- The drawing tablet area of KDE System Settings has been overhauled with new features and refinements.
- Many other enhancements and fixes throughout KDE Plasma 6.3.

You can read the announcement here.

For the first time since 2004, PassMark's global CPU benchmark data shows a decline in average processor performance, with laptop CPUs dropping 3.4% and desktop CPUs falling 0.5% year-over-year. Tom's Hardware reports: We see the biggest drop in laptop CPU performance results. PassMark recorded an average result of 14,632 across 101,316 samples last year. But, in 2025, the average score sat at an average of 14,130 points between 25,541 samples, decreasing the average score by 3.4%. The average desktop PC result in 2024 netted 26,436 points for 186,053 samples. But for 2025, the average score currently sits at 26,311 points for over 47,810 samples -- a 0.5% drop from last year. While that drop is small, we should only see a continued progression of faster performance.

[...] Passmark itself mused on X (formerly Twitter) that it could be that people are switching to more affordable machines that deliver lower power and performance. Or maybe Windows 11 is depressing performance scores versus Windows 10, especially as people transition to it with the upcoming demise of the latter. We've certainly seen plenty of examples of reduced performance in gaming with some of the newer versions of Windows 11, particularly as Intel and AMD struggled to upstream needed updates into the OS. [...] PassMark also muses that bloatware could contribute to the sudden decline in performance, but that seems like a longshot.

Slashdot reader headlessbrick writes: Google security researchers have discovered a way to bypass AMD's security, enabling them to load unofficial microcode into its processors and modify the silicon's behaviour at will. To demonstrate this, they created a microcode patch that forces the chips to always return 4 when asked for a random number.

Beyond simply allowing Google and others to customize AMD chips for both beneficial and potentially malicious purposes, this capability also undermines AMD's secure encrypted virtualization and root-of-trust security mechanisms.
Obligatory XKCD.

Arm has dropped its attempt to terminate Qualcomm's Architecture License Agreement (ALA), allowing Qualcomm to continue developing and producing Arm-compatible chips for PCs, smartphones, and servers. "The Brit biz had sought to end that license in a lawsuit it brought against Qualcomm in 2022," notes The Register. "That suit is rooted in Qualcomm's 2021 acquisition of a startup called Nuvia, which was co-founded by the brains behind Apple's custom processors and had signed an architecture license agreement (ALA) with Arm that allowed it to design its own Arm-compatible CPU cores." From the report: On Wednesday, Qualcomm's latest quarterly financial report [PDF] revealed Arm had indicated on January 8, 2025 it was no longer seeking to kill off Qualcomm's ALA. During Qualcomm's Q1 2025 earnings conference call with Wall Street, CEO Cristiano Amon confirmed Arm "has no current plan to terminate the Qualcomm Architecture License Agreement. We're excited to continue to develop performance leading, world-class products that benefit consumers worldwide that include our incredible Oryon custom CPUs." [...]

On the other side of the fence, Arm noted in a regulatory filing [PDF] that post-trial motions had been filed on both sides to clarify the legal situation following the jury's verdicts, and a new trial may be sought. On its own latest quarterly earnings call, which like Qualcomm's took place on Wednesday, Arm's CFO Jason Child was asked about the impact of the case. He said Arm's revenue forecasts assumed the biz was "not going to prevail in that lawsuit," and that it expected to continue receiving payments from Qualcomm, which licenses various technologies from Arm and doesn't just hold an ALA.

"The primary reason for the lawsuit very much was around defending our IP and that's important," Child said. "But from a financial perspective, we had assumed that we'll continue to be receiving royalties at basically the same rates that they've been paying for in the past and will continue to pay." Qualcomm continues to pursue another case against Arm, alleging the UK outfit didn't honor some of its contractual obligations. Arm reckons that matter will reach the courts in the first half of 2026.

The DeepComputing RISC-V Mainboard that Framework announced last year for its 13-inch laptops is now available for $199. Liliputing reports: If you already have a Framework Laptop 13 with an Intel or AMD motherboard, the new board is a drop-in replacement. But if you don't have a Framework Laptop you can also use the mainboard as a standalone computer: Framework sells a $39 Cooler Master case that effectively turns its mainboards into mini desktop computers. The RISC-V Mainboard comes from a partnership between Framework and DeepComputing, the Chinese company behind the DC-ROMA laptops, which were some of the first notebook computers to ship with RISC-V processors.

The board features a StarFive JH7110 processor, which is a 1.5 GHz quad-core chip featuring SiFive U74 RISC-V CPU cores and Imagination BXE-4-32 graphics, 8GB of onboard RAM, and a a 64GB SD card for storage (there's also support for an optional eMMC module, but you'll need to bring your own). Since the board is designed to fit in existing laptop frames, it's the same size and shape as AMD or Intel models and has four USB ports in the same locations. But these ports are a little less versatile than the ones you might find on other Framework Laptop 13 Mainboards [...]. There's also a 3.5mm audio jack. You can check out the new board via the Framework Marketplace.

Further reading: Late last year, Framework CEO Nirav Patel delivered one of the best live demos we've ever seen at a tech conference -- modifying a Framework Laptop from x86 to RISC-V live on stage.

Zyxel customers are facing reboot loops, high CPU usage, and login issues after an update on Friday went awry. The only fix requires physical access and a Console/RS232 cable, as no remote recovery options are available. The Register reports: "We've found an issue affecting a few devices that may cause reboot loops, ZySH daemon failures, or login access problems," Zyxel's advisory reads. "The system LED may also flash. Please note this is not related to a CVE or security issue." "The issue stems from a failure in the Application Signature Update, not a firmware upgrade. To address this, we've disabled the application signature on our servers, preventing further impact on firewalls that haven't loaded the new signature versions."

The firewalls affected include USG Flex boxes and ATP Series devices running ZLD firmware versions -- installations that have active security licenses and dedicated signature updates enabled in on-premises/standalone mode. Those running on the Nebula platform, on USG Flex H (uOS), and those without valid security licenses are not affected.

Intel is advocating for modular PC designs to improve repairability, reduce e-waste, and align with the right-to-repair movement. A trio of executives makes their case for such designs in a recent blog post. The Register reports: Intel's approach to the problem is to draft three proposals targeting different market segments, saying that a one-size-fits-all approach would not be able to address the nuanced demands of these varied segments. Those three segments comprise "Premium Modular PC" (actually a laptop design); "Entry/Mainstream Modular PC" (another laptop); and "Desktop Modular PC."

The first envisages a three-board system, comprising a core motherboard plus universal left and right I/O boards, the latter engineered to be common across fan-less Thin & Light designs with a 10W power envelope, and premium fanned designs for up to 20W or 30W. The Entry/Mainstream Modular PC is similar, with a core motherboard and left and right I/O boards, although in this segment, Intel says these can be redesigned to allow multiple SKUs of the design. The circuit boards are also cost-optimized here to cater to the mainstream segment, it says.

The Desktop Modular PC design appears from Intel's diagram to use a midplane that has the Platform Controller Hub (PCH) silicon, with other modules connecting to this. These include CPU, memory, and GPU modules, removable using slide rails, along with hot-swappable storage, all designed to fit inside a 5 liter desktop chassis. Intel also said it is introducing subsystem-level replaceable modules. In practice, this means something like a Type-C connector on a flexible printed circuit (FPC) or an M.2 circuit board. The idea is that the module can easily be swapped out if the port or connector is damaged.

"I know, I know: 'Year of the Linux desktop ... yadda, yadda'," writes Steven Vaughan-Nichols, a ZDNet senior contributing editor. "You've heard it all before. But now there's a Linux-powered PC that many people will want..."

He's talking about Nvidia's newly-announced Project Digits, describing it as "a desktop with AI supercomputer power that runs DGX OS, a customized Ubuntu Linux 22.04 distro." Powered by MediaTek and Nvidia's Grace Blackwell Superchip, Project DIGITS is a $3,000 personal AI that combines Nvidia's Blackwell GPU with a 20-core Grace CPU built on the Arm architecture... At CES, Nvidia CEO Jensen Huang confirmed plans to make this technology available to everyone, not just AI developers. "We're going to make this a mainstream product," Huang said. His statement suggests that Nvidia and MediaTek are positioning themselves to challenge established players — including Intel and AMD — in the desktop CPU market. This move to the desktop and perhaps even laptops has been coming for a while. As early as 2023, Nvidia was hinting that a consumer desktop chip would be in its future... [W]hy not use native Linux as the primary operating system on this new chip family?

Linux, after all, already runs on the Grace Blackwell Superchip. Windows doesn't. It's that simple. Nowadays, Linux runs well with Nvidia chips. Recent benchmarks show that open-source Linux graphic drivers work with Nvidia GPUs as well as its proprietary drivers. Even Linus Torvalds thinks Nvidia has gotten its open-source and Linux act together. In August 2023, Torvalds said, "Nvidia got much more involved in the kernel. Nvidia went from being on my list of companies who are not good to my list of companies who are doing really good work." Canonical, Ubuntu Linux's parent company, has long worked closely with Nvidia. Ubuntu already provides Blackwell drivers.
The article strays into speculation, when it adds "maybe you wouldn't pay three grand for a Project DIGITS PC. But what about a $1,000 Blackwell PC from Acer, Asus, or Lenovo? All three of these companies are already selling MediaTek-powered Chromebooks...."

"The first consumer products featuring this technology are expected to hit the market later this year. I'm looking forward to running Linux on it. Come on in! The operating system's fine."

An anonymous reader quotes a report from TechCrunch: On Saturday, Triplegangers CEO Oleksandr Tomchuk was alerted that his company's e-commerce site was down. It looked to be some kind of distributed denial-of-service attack. He soon discovered the culprit was a bot from OpenAI that was relentlessly attempting to scrape his entire, enormous site. "We have over 65,000 products, each product has a page," Tomchuk told TechCrunch. "Each page has at least three photos." OpenAI was sending "tens of thousands" of server requests trying to download all of it, hundreds of thousands of photos, along with their detailed descriptions. "OpenAI used 600 IPs to scrape data, and we are still analyzing logs from last week, perhaps it's way more," he said of the IP addresses the bot used to attempt to consume his site. "Their crawlers were crushing our site," he said "It was basically a DDoS attack."

Triplegangers' website is its business. The seven-employee company has spent over a decade assembling what it calls the largest database of "human digital doubles" on the web, meaning 3D image files scanned from actual human models. It sells the 3D object files, as well as photos -- everything from hands to hair, skin, and full bodies -- to 3D artists, video game makers, anyone who needs to digitally recreate authentic human characteristics. [...] To add insult to injury, not only was Triplegangers knocked offline by OpenAI's bot during U.S. business hours, but Tomchuk expects a jacked-up AWS bill thanks to all of the CPU and downloading activity from the bot. Triplegangers initially lacked a properly configured robots.txt file, which allowed the bot to freely scrape its site since the system interprets the absence of such a file as permission. It's not an opt-in system.

Once the file was updated with specific tags to block OpenAI's bot, along with additional defenses like Cloudflare, the scraping stopped. However, robots.txt is not foolproof since compliance by AI companies is voluntary, leaving the burden on website owners to monitor and block unauthorized access proactively. "[Tomchuk] wants other small online business to know that the only way to discover if an AI bot is taking a website's copyrighted belongings is to actively look," reports TechCrunch.