Everything tagged gpu, newest first. Tags come from the classifier reading each item's summary; 394 tags used 25 times or more have their own page.
Nscale has formed a multi-year strategic partnership with Figure to support physical AI development by deploying up to 100,000 NVIDIA GPUs, with the first rollout planned for the second half of 2027 at Nscale’s UK site.
Amazon plans to add 2 million more Nvidia GPU chips to its data centers over the next two years, signaling a deeper partnership that goes beyond simply purchasing hardware.
The rapid expansion of AI continues, with massive investments in data centers and GPUs making compute the largest expense for AI product development, yet pricing and hedging strategies for compute costs remain unclear.
French startup Kog challenges the misconception that GPUs are unsuitable for agentic workflows, suggesting they may be more capable than previously thought.
Vultr is expanding its AI infrastructure by introducing the AMD Instinct MI455X GPU and supporting AMD Helios architecture, targeting AI and high-performance computing applications, with pre-orders available for Q4.
Meta's Generative Ads Recommendation Model (GEM) now trains at LLM scale, achieving doubled training efficiency and quadrupled FLOPs using thousands of advanced GPUs.
A $400 million loan secured by chips indicates a significant upcoming trend in AI infrastructure investments, highlighting the increasing financial focus on AI technology development.
This beginner's course, developed by Pat Verra, introduces graphics programming fundamentals using p5.js, teaching vertex and fragment shaders, and culminating in creating an animated fractal heart effect, while explaining the efficiency of shaders on GPUs compared to CPUs.
To optimize large language model inference, techniques like KV cache and paged attention from VLLM can significantly reduce latency and costs by efficiently managing GPU memory during the pre-fill and decode phases.
Crea 2 is a fast, open-source image generator capable of producing photorealistic and diverse art styles offline, compatible with most consumer GPUs, and easily installed using Comfy UI.
AI infrastructure planning is shifting from a focus on training clusters to prioritizing inference workloads, necessitating changes in infrastructure behavior and design.
The RX 9070, a high-performance product, was launched over a year ago with a price tag of $549.
Intel's Arc B390 integrated GPU delivers remarkable performance enhancements in laptops, showcasing its capabilities in the competitive GPU market.
Ryan discusses with MinIO co-founders Garima Kapoor and Anand Babu Periasamy about overcoming storage bottlenecks, their collaboration with NVIDIA on STX architecture, and the trend towards S3-compatible object storage in AI infrastructure.
Upgrading an old PC with a high-end GPU can enhance its performance and compatibility with modern gaming and applications.
Ryan and AMD CTO Mark Papermaster discuss AMD's AI silicon strategy, addressing heterogeneous CPU/GPU computing, AI workload management, and the role of AI agents in accelerating chip innovation.
Google's new Gemma 4 model, released under an Apache 2.0 license, is a highly capable and compact AI model that can be run locally on standard GPUs, offering significant performance even on edge devices.
Famous, a startup that sought to revolutionize web development with a GPU-accelerated CSS property, failed due to browser advancements, complex API, and unsustainable business practices, despite initially raising $30 million.
Meta has open-sourced RCCLX, an enhanced version of RCCL, to improve AI model communication and innovation across various backends, fully integrated with Torchcomms.
AFL, a fibre optic cable manufacturer, will showcase their solutions for reducing GPU-to-GPU latency at Data Centre World 2026 in London, highlighting the importance of bandwidth for AI clusters.
The course on pipeline parallelism teaches how to accelerate AI model training by distributing models across multiple GPUs, using assembly line processing, and building distributed training systems from scratch, with prerequisites in PyTorch and Python.
The Free Code Camp podcast episode features an interview with software engineer Schumit Shaha, community news about a UC Berkeley hackathon documentary, new educational courses on Kubernetes and GPUs, and a song highlight, while encouraging donations to support their charitable mission.
GPU manufacturers are focusing on producing more profitable models, while large SSDs are becoming increasingly scarce in the market.
Dave introduces Dell's GB10, a powerful chip combining GPU and CPU capabilities, demonstrating its practical applications in AI, gaming, and video monitoring, emphasizing its unified memory and efficient design for real-world tasks.
Gordon Johnson predicts that by 2026, the data centre industry will shift towards hybrid cooling solutions due to increasing rack densities and GPU power demands, surpassing traditional air-cooling capabilities.
The video demonstrates how to use the free, open-source tool Persona Live to transform into various characters in real-time using a consumer-grade GPU, highlighting its ease of use and limitations with non-humanlike characters.
The video explores the creation of a powerful AI supercomputer using four Mac Studios with enhanced networking capabilities, aiming to determine if clustering local AI can be efficient and cost-effective compared to previous attempts.
A Raspberry Pi, despite its limited PCIe capabilities, can effectively utilize multiple GPUs for tasks like running large language models, offering a cost-effective and energy-efficient alternative to traditional PCs for certain applications, though it struggles with gaming due to driver and software limitations.
"Redstone" combines traditional and innovative concepts, yet Nvidia maintains a significant lead in the industry.
The author is excited about testing NVIDIA A4000 GPU on a Raspberry Pi CM5, achieving full GPU acceleration for AI tasks despite no display output, and plans to explore compatibility with other ARM systems and graphics cards post-Thanksgiving.
Ryan and Jared Quincy Davis discuss efficient resource allocation and GPU utilization in AI, debunk GPU shortage myths, and explore economic changes with new scheduling strategies.
Cisco's unified platform offers a modular, configurable solution with compute nodes, built-in router, and GPU support to address security, connectivity, and AI needs on-site.
Intel Arc GPUs are now somewhat stable on Raspberry Pi, requiring a patch for broader support, with additional steps for optimal performance, including firmware installation and manual Mesa compilation.
The introduction of new components enhances functionality and power, while maintaining the product's unique and unconventional nature.
The Minisforum MS-R1, featuring a 12-core Arm CPU, offers impressive hardware but falls short in performance and efficiency compared to competitors like Apple's M1, with issues in OS compatibility and power consumption.
Cisco's AI pods offer a scalable, secure, and simplified architecture integrating high-performance GPUs, networking, and storage, enabling rapid deployment and management of AI solutions with real-time monitoring and customizable dashboards for optimal performance and ROI.
Building effective AI systems requires understanding and integrating key components of the AI technology stack, including models, infrastructure, data, orchestration, and application layers, to solve complex, meaningful problems efficiently and safely.
The NVIDIA DGX Spark is a compact AI supercomputer with impressive specs, including a GB10 Grace Blackwell superchip and 128 GB of unified memory, but it struggles to outperform a dual 4090 AI server in speed, though it excels in running multiple tasks simultaneously.
Nvidia's continuous innovation has led to significant financial success, achieving $41.1 billion in data center sales in the latest quarter.
A new AI tool generates steerable 3D-like videos, offering advanced visual effects but necessitating powerful GPU resources for optimal performance.
Cisco's IT team rapidly developed a cutting-edge AI infrastructure using existing resources and new facilities, overcoming power and latency challenges to support over 25 business use cases.
DE-CIX has launched the world's first AI Internet Exchange (AI-IX), connecting over 50 AI-related networks to make its global internet exchanges AI-ready.
The rapid adoption of AI in enterprises demands robust networking and operational foundations to prevent bottlenecks and ensure efficient, ROI-driven deployments.
Digital Connexion's MAA10 facility in Chennai has been certified as an NVIDIA DGX-Ready Data Center, highlighting its capability to support AI and GPU-intensive computing workloads.
Nvidia has announced support for gamers utilizing Windows 10's extended updates, ensuring continued compatibility and performance for their gaming hardware.
AI is transforming industries, requiring AI-ready data centers with high-performance computing and low-latency networking, and Cisco, in partnership with NVIDIA, offers solutions to overcome traditional data center limitations.
AFL has launched DENALI, a modular optical fibre platform designed to support high-growth GPU environments and scalable infrastructure for hyperscale and AI workloads.
GPUhammer is a pioneering tool that successfully manipulates bits in GPU memory, setting a precedent for future developments.
The podcast discusses Arista's market share, a critical Adobe vulnerability, Arista's acquisition of VeloCloud for SD-WAN expansion, and CoreWeave's purchase of a cryptominer for GPU and electricity access.
Cisco, in collaboration with NVIDIA, offers AI Ready Infrastructure Solutions to address fragmented infrastructure and scalability issues, enabling partners to deliver scalable AI solutions and earn rewards through the APJC Partner Winner Circle 2025.
