Tech News
← Home  ·  All topics

Nvidia

707 GoKawiil briefs on this topic

Nintendo Switch 2 discounted to $469.99 at Woot, down from $499 list price

Retailer Woot is selling the Nintendo Switch 2 for $469.99, a $30 discount off its $499 price, which itself was a $50 increase introduced a couple of months earlier. The deal is limited to available stock or runs through the end of the month, whichever comes first.

Meta reportedly building household humanoid robot prototype codenamed Max

Meta has developed a prototype humanoid robot called Max intended to assist with household chores, according to The Information, built in about three months with help from Chinese robotics firm Fourier. The robot reportedly weighs about 106 pounds, stands roughly five feet tall, has a facial-expression display, a speaker, hands for grasping small objects, and runs on NVIDIA's Jetson Thor chips. A second prototype, Max 2, has reportedly been in development since September with design updates to the torso, hands, a projector, and added sensors for navigating stairs.

Hardware Accelerated GPU Scheduling shows little benefit on budget graphics cards, tests find

Microsoft's Hardware Accelerated GPU Scheduling (HAGS), introduced in a May 2020 Windows 10 update, shifts scheduling of graphics tasks from the CPU to dedicated hardware on the GPU. Nvidia added support that June starting with GTX 10-series cards, AMD didn't support it until December 2023 on its RX 7900, 7800 and 7700 series, and Intel limits support to its Arc B-series. Benchmarks from Tom's Hardware and Gamers Nexus found the feature produced negligible or even slightly negative performance changes, including a repeatable 2.4 percent drop in one game on a budget Intel CPU.

Open-source tool nv-uefi-unlocker boosts RTX GPU tensor speed via USB on Windows

A new open-source project called nv-uefi-unlocker lets owners of Nvidia RTX 20, 30, and 40-series GPUs remove a built-in tensor-performance limiter by booting from a USB stick, with no VBIOS flashing or driver patching required. The tool runs before Nvidia's driver loads, rebuilding firmware locally from Nvidia's public driver files, and the developer reported a 65% jump in FP16 performance on a laptop RTX 4060.

Tom's Hardware Premium rounds up AI CPU benchmarking and terminology primer coverage

Tom's Hardware's Premium subscription service published a weekly digest covering Anthropic's pre-IPO filings, Cadence's new LLM suite for EDA tools, Nvidia's AI supply chain moves, Vera CPUs in the cloud, and a TP-Link update. The issue also features CPU expert Jake Roach's early work on defining agentic AI CPU benchmarks, plus a new primer explaining AI terms like inference, quantization, context, and KV cache for locally-run models.

Nvidia in talks to guarantee debt for Nscale's West Virginia AI data center

Nvidia is reportedly negotiating to back billions in debt financing for Nscale's planned Anthropic data center in West Virginia, according to The Information and corroborated by Bloomberg. Under the proposed residual value guarantee, Nvidia would cover construction costs if Nscale defaulted, temporarily take over the site, and help find a new buyer or tenant. Anthropic has already signed a $44.6 billion contract for the first building alone, per Nscale's IPO filing.

Intel shares plunge 28% on weak Q4 forecast, cutting US stake value by $12B

Intel stock fell about 28% on Friday after the company projected fourth-quarter revenue of $15-16 billion, below analyst expectations of roughly $17 billion, and warned gross margins would drop to 37% from 43% with no positive cash flow expected. The decline cut roughly $12 billion from the value of the US government's 10% stake in Intel, acquired over a year ago, and also dented Nvidia's shares along with the wider chip sector.

Big Tech's $740 billion AI capex triggers investor pullback as Alphabet posts negative cash flow

Alphabet, Amazon, Meta and Microsoft have disclosed plans to spend over $740 billion combined on capital expenditures this year, largely tied to AI data centers and chips. Alphabet reported negative free cash flow of $9.1 billion in the third quarter, its second straight quarterly loss on that measure, even as its quarterly capex rose to $51.4 billion. The news hit markets Friday, with the Nasdaq falling 2.9% and chip stocks including Nvidia, Broadcom and Qualcomm declining sharply, following Alphabet's 12% drop the prior day that erased $500 billion in market value.

Nvidia removes RTX 5080 listings from its store after pulling RTX 5090

Nvidia has taken down GeForce RTX 5080 listings from its own marketplace, days after similarly removing the RTX 5090. Street prices for the RTX 5080 have reportedly risen to more than double its $999 MSRP as stock dwindles at US retailers, while other Blackwell cards like the RTX 5070 and 5070 Ti remain listed as out of stock rather than delisted.

Intel shares fall 30% in single-day record drop, triggering broad chip sell-off

Intel stock lost 30% of its value on Friday after the company issued a weak financial outlook, including a forecast that sales in its client computing group would drop over 20%. The slide erased nearly $150 billion from Intel's market cap and contributed to roughly $1 trillion wiped off the broader semiconductor sector, as measured by Benzinga's intraday estimates and a 9.14% drop in the PHLX Semiconductor Sector Index.

Tom's Hardware publishes glossary explaining LLM terms like weights, KV cache and quantization

Tom's Hardware has released an explainer breaking down core terminology behind large language models, covering concepts such as models, tokens, weights, KV cache, quantization and inference. The guide is aimed at readers following the growth of local AI hardware, such as Nvidia's RTX Spark, and is intended to support the publication's ongoing AI coverage.

Runware plans assembly-line factory to mass-produce AI data center pods

Romania-based Runware says it is building a second factory to manufacture its modular Sonic Inference Pod data centers on a production line, similar to automobile manufacturing. CEO Flaviu Radulescu told TechCrunch the company aims to scale to 30 megawatts of data center capacity per day by 2028, with pods installable on customer sites within 48 hours.