❌

Reading view

Dev hacks Xiaomi's Smart Humidifier to free it from the cloud, now works with Home Assistant locally β€” custom firmware allows the product to evade planned obsolescence

If the one thing about smart home appliances that deterred you from ever investing in this connected future was how your data was always being routed through servers, you're in for a treat. A skilled developer has hacked his new Xiaomi Humidifier with ESPHome, making the device compatible with Home Assistant.

Industry preps new 'cheap' HBM4 memory spec with narrow interface, but it isn't a GDDR killer β€” JEDEC's new SPHBM4 spec weds HBM4 performance and lower costs to enable higher capacity

JEDEC is nearing completion of SPHBM4, a standard that enables full HBM4 bandwidth over a 512-bit interface using a 4:1 serialization, reusing standard HBM DRAM dies and a base die. The tech promises to enable a 2.5D integration on organic substrates to support up to 64 GB per stack and more stacks than HBM4 and HBM4E.

Nvidia details new software that enables location tracking for AI GPUs β€” opt-in remote data center GPU fleet management includes power usage and thermal monitoring

Nvidia's GPU fleet management software can track spikes in power usage, monitor utilization, detect hotspots, spot anomalies, identify software errors, and detect the physical location of processors. However, the software is completely optional for its clients.

Creality Falcon2 Pro 60 Watt Laser Engraver review: Deep cuts and fine engravings

The Creality Falcon2 Pro 60W is a fully enclosed diode laser engraver with multiple safety features, intuitive first-party software, and a pull-out tray for easy clean-up that can be purchased on sale for $1,899 for the Single Unit package. With the included 1.6W laser module, the Falcon2 Pro is capable of cutting through thick material as easily as it can produce detailed engravings.

Intel Core Ultra 7 270K Plus and Core Ultra 9 290K Plus spotted at Indian retailer β€” listings appear to corroborate prior leaks but don't reveal pricing or new info for upcoming Arrow Lake refresh

We've gotten our first retail appearance of the Arrow Lake refresh family, listed at an Indian retailer but with barely any information. The specs mentioned seem regurgitated from existing media coverage and leaks, and there's no price, but at least the vendor is promising a 3-year warranty.

Huawei's latest mobile is armed with China's most advanced process node to date despite using blacklisted chipmaker β€” Huawei Kirin 9030 mobile SoC made on SMIC N+3 process, but can't compete with 5nm node

TechInsights says Huawei's Kirin 9030 is built on SMIC’s N+3 process: an incremental, DUV-based extension of its 7nm-class technology that pushes density without EUV but falls well short of true 5nm nodes amid rising yield challenges.

Photonic latch memory could enable optical processor caches that run up to 60 GHz, twenty times faster than standard caches β€” optical SRAM stores and outputs data entirely as light, but density challenges remain

Researchers from the University of Southern California Information Sciences Institute and the University of Wisconsin-Madison have demonstrated what they describe as the first regenerative photonic memory latch.

China mulls $70 billion domestic chip fabrication injection, would be largest of any government semiconductor investment β€” Huawei and Cambricon among candidates in push to compete with Nvidia, other U.S. firms

China is considering investing up to an additional $70 billion in its domestic chip manufacturers in an effort to better compete with US firms like Nvidia. Although exact figures and investment strategies have yet to be decided, this move would be in line with China's "whole nation" approach to tackling its chip shortages.

Asus' new 5K gaming monitor is so high-end it doesn't even officially support RTX 40-series GPUs β€” the XG27JCG is a 5K 180Hz beast with 330Hz 1440p dual-mode support

Asus is betting on overpriced GPUs in this ultra high-res and high refresh rate pairing, dubbed the "XG27JCG," a new 5K 180 Hz monitor that can only run on Nvidia's RTX 50-series, officially, or AMD's RX 7600 and up. It supports all the gaming features you could ever want, along with solid color accuracy and HDR support.

Microsoft, Google, OpenAI, and Anthropic join forces to form Agentic AI alliance, according to report β€” organization backed by the Linux Foundation is set to create open source standards for AI agents

The world's leading AI firms are collaborating on a new Agentic Artificial Intelligence Foundation managed by the Linux Foundation to build open standards around AI agents. The move will focus on three key open source tools to begin with, sharing findings on technical problems.

Framework puts Dell and Apple on blast over egregious RAM prices β€” modular laptop maker will be forced to increase memory prices, but won't "gouge customers" like other vendors

Memory prices are rising everywhere, and the latest to be hit in the aftermath of this crisis is modular laptop manufacturer Framework, which just announced its own price hike in a very interesting fashion. The company replied to a tweet showing Dell's allegedly insane markup on RAM upgrades that later turned out to be incorrect.

Nvidia decries 'far-fetched' reports of smuggling in face of DeepSeek training reports β€” unnamed sources claim Chinese company is involved in Blackwell smuggling ring

DeepSeek is allegedly involved in a "phantom data center" smuggling scheme to get Blackwell GPU servers into China as part of training its newest LLM generation. While Nvidia refutes the claims as "farfetched", some proof indicates otherwise.

Microsoft promises to make Windows 11 the best operating system for gaming β€” says it will focus on background workloads, power and scheduling, graphics stack, and drivers

Microsoft says Windows 11 will receive major gaming-focused enhancements in 2026, including broader support for Xbox Full Screen Experience, Auto Super Resolution for handhelds, expanded Advanced Shader Delivery, and more.

Intel's upcoming Core Ultra X9 388H is up to 8.7% faster for 1T perf than Ryzen AI Max+ 395 β€” Panther Lake gains significant ground on Strix Halo in early Geekbench leak

Intel's Panther Lake family is still awaiting release, but the possible flagship chip of the bunch has been spotted on Geekbench. The Core Ultra X9 388H scores 3,057 points in the single-core test and 17,687 points in the multi-core test, racing past its Arrow Lake-H predecessor and AMD's Strix Halo.

This GitHub script claims to wipe all of Windows 11's AI features in seconds β€” "RemoveWindowsAI" can disable every single AI feature in the OS, from Copilot to Recall and more

If you've been unhappy with the direction Microsoft has taken Windows, offering no meaningful improvements beyond AI and aesthetics, then, well, not much can be done about that. But, at least you can disable all the AI features that seem to have populated every corner of the OS, with a simple script from GitHub.

ASML under fire for selling DUV equipment to Chinese firm with military ties, says the machines are not subject to export controls β€” fears grow that 'old technology' will bolster Beijing's quantum effort

Dutch chip manufacturing design firm, ASML, has been accused of selling chip fabrication hardware to a Chinese defence firm tied to the ruling party. Although it claims the technology was old, it has raised concerns over its potential use in developing new quantum technologies, particularly with military applications.

Scavenger scores 14900KS PC with 64GB of DDR5 for less than the cost of RAM alone β€” $2500+ machine sells for just $600

A pawn shop in Portugal clearly didn't know the worth of what they had and ended up selling a high-end PC worth at least $2,500 with current market pricing, for just $600 to a lucky buyer. The PC includes flagship components, including two 32 GB sticks of 6000 MT/s DDR5 memory that alone costs almost $700.

Nvidia reportedly wins H200 exports to China β€” US Department of Commerce set to ease restrictions for full Hopper AI GPU

The U.S. government is reportedly preparing to let Nvidia ship its H200 accelerators to China, a move that could restore Nvidia’s influence in the Chinese AI market and reinforce CUDA’s dominance, but the question is if Beijing agrees to accept this hardware.

Research commissioned by OpenAI and Anthropic claims that workers are more efficient when using AI β€” Up to one hour saved on average, as companies make bid to maintain enterprise AI spending

OpenAI and Anthropic claim in a pair of reports released today and earlier in the month that the use of enterprise AI tools increase productivity and corporate ROI. These studies may be damage control to counter those released by MIT and Harvard in August claiming the opposite.

AMD's B650 chipset isn't going anywhere, according to a new rumor β€” rising memory costs and softening DIY demand mean the transition to B850 may take longer than expected

AMD is apparently no longer telling vendors to stop B650 production in favor of B850 (and B840) due to the current component crisis. Skyrocketing DDR5 prices have caused a steep decline in new motherboard sales, which has reportedly influenced the Red Team to continute to make and sell B650 boards for now, helping consumers still select the cheaper option.

The Senate's new SAFE bill is set to curb access to advanced chips to China, but that won't slow down the AI war β€” training workloads still heavily rely on Nvidia, while alternatives remain inefficient

Even as U.S. and Chinese lawmakers make it harder for Nvidia to sell its chips to Chinese customers, the alternatives aren't able to offer a competitive product. With smuggling and obfuscated shipment routing making it possible to get around blocks and barriers, it may be that Nvidia remains the dominant training hardware provider in the region.

Geekbench leak sees Intel's upcoming Core Ultra 7 270K comfortably ahead of Core Ultra 265K β€” alleged result shows Arrow Lake refresh chip 5.6% ahead of the 265K

The Arrow Lake refresh is expected to fill in the gap till Nova Lake finally launches. The Core Ultra 7 270K Plus has just appeared on Geekbench and it achieved great scores that show a decent improvement over the outgoing 265K Plus it'll replace, despite running on relatively slow memory.

Asus RTX 5070 gaming laptop hits an all-time low price of $1,099 β€” this 16-inch model comes with a Core i7 CPU, plus expandable 16GB RAM and 1TB SSD storage

If you've been looking to upgrade to a more powerful machine but don't want to spend the big bucks on nice-to-haves like an OLED display, the Asus TUF F16 is the perfect pick for you. With its performant hardware combo and plenty of connectivity, along with a sleek design, it's a solid laptop that can do it all: gaming, productivity, and media consumption.

CPU air cooler runs ice-cold water through its heatpipes to liquid-cool a GPU β€” negative-temp DIY mod sees up to 17% performance uplift

Watch as a crazed modder grinds down the heatpipes of an excellent air cooler, then injects ice-cold water through them with retrofitted tubes, and finally puts the entire apparatus on two GPUs to see how far he can push performance, witnessing massive clock-speed uplifts across the board.

AMD's imminent Ryzen 7 9850X3D chip shows up on Geekbench with 5.6 GHz boost clocks β€” Scores slightly lower than 9800X3D in multi-core tests, higher in single-core

Two new Geekbench listings for the upcoming Ryzen 7 9850X3D chip have been uncovered, both from mid-November, that show the CPU posting higher single-core numbers than the existing 9800X3D while being beat in multi-threaded results. The chip boosted up to 5.6 GHz as expected, but was running on slow 4800 MT/s memory.

Huawei Ascend NPU roadmap examined β€” company targets 4 ZettaFLOPS FP4 performance by 2028, amid manufacturing constraints

Huawei has unveiled its Ascend NPU roadmap featuring Ascend 950, 960, 970 processors and massive SuperClusters with over a million of processors and up to 4 ZettaFLOPS FP4 performance in 2028, shifting from chip scaling to system-level scaling, amid U.S. sanctions and manufacturing constraints.

❌