#nvidia

Nvidia unveiled the Vera CPU, focused on max single-threaded performance at scale for AI and HPC workloads. The chip aims to eliminate bottlenecks in large-scale data processing and model training by improving single-core execution speed. AI innovators are adopting Vera to boost throughput and reduce latency.

The article argues that the current AI industry is fueled by unsustainable hype and investment, leading to a market bubble reminiscent of past tech crashes. It warns that the massive financial burn and lack of real-world utility could result in a harsh correction. The author suggests letting the AI bubble burst to clear out speculative excess.

First in a series on optimizing matrix multiplication for NVIDIA's Blackwell GPU. Covers architecture features (GB202/GB203 dies, new SM partitioning, enhanced Tensor Cores) and methodology using CUDA and low-level assembly tuning for peak performance.

NVIDIA's NVCF (NVIDIA Cloud Functions) is a framework for deploying and routing GPU-accelerated AI workloads at scale. It enables serverless execution of AI inference and data processing tasks across distributed cloud environments, optimizing resource usage and reducing latency for production AI applications.

CanopyWave's analysis compares Nvidia's B300 and H200 GPUs across architecture, memory bandwidth, and compute performance. The B300 offers significant improvements over the H200 in AI training and inference workloads, including higher flop rates and faster memory. The article provides detailed benchmark data and spec comparisons to help data centers choose the optimal GPU for their needs.

The article explores the capabilities of NVIDIA's DGX Station and its role in running frontier AI models locally, addressing questions about memory requirements, performance, and practical deployment for local AI workloads.

Thieves stole $1.3 million worth of Nvidia and AMD GPUs from a San Francisco data center by posing as legitimate contractors, highlighting the vulnerability of AI infrastructure to cargo theft and the lack of security protocols around high-value hardware.

This article introduces a series on matrix multiplication optimization for NVIDIA's Blackwell architecture, explaining the importance of efficient matrix math for AI workloads and outlining the hardware advancements in Blackwell that enable faster computation compared to previous architectures.

In 2003, Nvidia released a humorous video mocking itself over a flawed product launch, acknowledging the launch issues of its GeForce FX 5800 Ultra graphics card. The parody video, titled "Nvidia own mockery of a bad product release (2003)," shows the company poking fun at the card's high noise levels and performance problems that had drawn criticism from reviewers and consumers.

Singaporean authorities have seized a mansion worth $42 million linked to an alleged scheme to smuggle Nvidia chips to China. The operation is part of an ongoing investigation into unlawful exports of advanced semiconductors subject to U.S. trade restrictions.

Nvidia is offering startup customers the option to access its computing power in exchange for a share of their future revenue, rather than paying upfront. The program aims to support early-stage AI companies by reducing immediate costs while Nvidia benefits from their potential success.

NVIDIA Research presents a method called "World from Motion" that reconstructs dynamic 3D Gaussian scenes from a single monocular video. The approach generates a fully editable, real-time renderable 4D representation of dynamic scenes without requiring multi-view or depth data.

A crypto miner reflects on a decade of watching Nvidia evolve from a GPU maker for gamers and miners to a trillion-dollar AI powerhouse, tracing the shift from general-purpose CUDA cores to specialized AI hardware and the cultural change from mining mania to the AI boom.

An analysis of the best-performing stocks over the last century found that nearly all top performers are technology companies, including Apple, Nvidia, Tesla, and SpaceX, highlighting the dominant role of tech in long-term investment returns.

NVIDIA GQE is an open-source library that helps developers build GPU-accelerated SQL query engines. It provides building blocks for query compilation, execution, and data management on GPUs, enabling higher throughput and lower latency for database and analytics workloads compared to CPU-only engines.

Nvidia is bringing back older GPU models like the RTX 3060 to address rising RAM demands and help stabilize graphics card prices in the market.

The article compares NVIDIA's DGX Spark (a compact AI supercomputer) against Apple's Mac Studio and other "Halo" AI PCs, examining their performance, price, and suitability for local AI development and inference tasks.

Taiwanese authorities raided Super Micro Computer's office in Taiwan as part of an expanding probe into alleged smuggling of Nvidia chips to China, which could violate export controls on advanced AI semiconductors.

Wall Street analysts are increasingly bullish on Micron Technology, predicting it could be the next big winner in AI, following in Nvidia's footsteps. The company's memory and storage solutions are seen as critical for AI data centers and high-performance computing. This optimism is driven by surging demand for its HBM (high-bandwidth memory) chips used in AI processors.

BlackRock, Nvidia, and Temasek are among major investors pouring billions into quantum computing, betting that the nascent technology will eventually revolutionize industries by solving problems beyond the reach of classical computers. The heavy influx of capital signals growing confidence that quantum computing is moving closer to practical, commercial applications.

Nvidia CEO Jensen Huang compared the startup Fireworks to TSMC, positioning it as a key provider of AI inference infrastructure for enterprises, similar to how TSMC manufactures chips. He highlighted Fireworks' role in running AI models efficiently and reliably for business applications.

This article explains the detailed process of launching and executing a CUDA GPU kernel, covering steps from CPU invocation and driver interaction to thread block scheduling on streaming multiprocessors (SMs), warp execution, memory access, and synchronization.

Nvidia partner Lambda is developing a home AI data center called LambdaCube, priced at $150,000. Designed for personal use, the compact system would contain Nvidia's powerful GPUs and fit in a user's yard, bringing enterprise-grade AI computing to residential settings.

The Reddit post asks about what is known regarding Nvidia's "Feynman" architecture in 2026. The thread likely discusses rumored specs, performance expectations, and potential release timelines for Nvidia's next-generation GPU architecture following Blackwell.

The article benchmarks non-reasoning Tiny LLMs (small language models) on the Jetson Orin Nano Super 8GB. It evaluates models like Phi-3, Gemma, and Llama variants on this edge device, measuring performance metrics such as tokens per second.

NVIDIA has released v0.1.0 of the nvidia-nvswitch-setup utility, which automates the configuration of NVIDIA NVSwitch hardware in multi-GPU nodes. The tool handles firmware updates, driver loading, and system setup to streamline the deployment of NVSwitch-based systems.

An analysis of the best-performing stocks over the past century shows that nearly all top investments are technology companies, including Apple, Nvidia, Tesla, and SpaceX, highlighting the dominant role of tech in long-term market growth.

An open-source project introduces VDCores, a new programming model for NVIDIA GPUs that decouples compute and memory, enabling asynchronous memory operations. It reports a 12% performance improvement over existing state-of-the-art and a 67% reduction in kernel code.

Intel's upcoming Arc B580 and B570 graphics cards aim to deliver strong 1440p gaming performance at aggressive $219-$249 price points, filling a budget-gamer gap left by Nvidia and AMD's recent focus on higher-end offerings.

Nvidia's new 45°C liquid cooling design for AI data centers can reduce water consumption to near zero. The approach uses warm water cooling, eliminating the need for water-intensive evaporative cooling systems. This could help AI factories operate more sustainably while managing the heat generated by high-performance computing.