News from the AI & ML world
DeeperML - #blackwell
|
@futurumgroup.com
//
NVIDIA is making significant strides in the development of AI infrastructure and solutions, partnering with key industry players to advance AI capabilities across various sectors. Dell Technologies has been awarded a contract to build the next-generation NERSC-10 supercomputer, designated Doudna, powered by the NVIDIA Vera Rubin architecture. This system, named after Nobel laureate Jennifer Doudna, will integrate AI, quantum workflows, and real-time data streaming, aiming to serve 11,000 researchers in fields such as fusion energy, astronomy, and life sciences. The Doudna supercomputer will deliver 10 times the scientific output of its predecessor while improving power efficiency by 3-5 times.
NVIDIA is also addressing the China AI market with the development of export rules-compliant chips in collaboration with AMD. This move comes as the US government has placed restrictions on chip exports to China. Despite these challenges, NVIDIA CEO Jensen Huang has emphasized the importance of maintaining a presence in China, where a significant portion of AI developers are located. Huang advocates for the adoption of American technology stacks, highlighting the strategic importance of NVIDIA's CUDA development platform in the Chinese market. These efforts underscore NVIDIA's commitment to navigating trade regulations while continuing to foster AI innovation globally. Furthermore, Dell is expanding its AI offerings with the Pro Max Plus workstation laptop, featuring a discrete enterprise-grade NPU (Neural Processing Unit). This device enables enterprises to run large AI models locally, bypassing the cloud for enhanced control, privacy, and cost-efficiency. The Pro Max Plus can support models with up to 109 billion parameters on-device, catering to AI developers, engineers, and data scientists working with proprietary data. This initiative reflects a growing demand for secure, high-performance AI environments that can be deployed in portable, flexible form factors, further solidifying NVIDIA's role in driving advancements in AI infrastructure and accessibility. Recommended read:
References :
Aaron Klotz@tomshardware.com
//
NVIDIA has recently announced a significant breakthrough in AI inference, achieving a new world record with its DGX B200 Blackwell node. This node, equipped with eight Blackwell GPUs, surpassed the 1,000 tokens per second (TPS) per user barrier while running Meta’s Llama 4 Maverick large language model. According to a report by Artificial Analysis, the DGX B200 node achieved 1,038 TPS/user, outperforming previous record holders like SambaNova, who achieved 792 TPS/user. This advancement showcases the immense capabilities of the Blackwell architecture and sets a new standard for AI performance.
NVIDIA achieved this record-breaking performance through extensive software optimizations, utilizing TensorRT and Eagle-3 techniques for speculative decoding. These optimizations resulted in a 4x performance uplift compared to Blackwell's prior best results. Further enhancements involved using FP8 data types, Attention operations, and the Mixture of Experts (MoE) AI technique. These improvements not only boosted speed but also maintained response accuracy. NVIDIA's Blackwell GPUs reached 72,000 TPS/server at their highest throughput configuration. In addition to AI performance, NVIDIA is also revolutionizing AI data center infrastructure through a collaboration with Navitas Semiconductor. They are introducing a new 800V HVDC architecture designed to replace the aging 54V systems currently in use. This new architecture is expected to deliver up to 5% better power efficiency and 70% lower maintenance costs. The transition to 800V power enables a 45% reduction in copper wire thickness, significantly lowering material use and weight, while reducing heat and fewer losses. Recommended read:
References :
@blogs.nvidia.com
//
Nvidia is significantly expanding its AI infrastructure initiatives by introducing NVLink Fusion, a technology that allows for the integration of non-Nvidia CPUs and AI accelerators with Nvidia's GPUs. This strategic move aims to provide customers with more flexible and customizable AI system configurations, broadening Nvidia's reach in the rapidly growing data center market. Key partnerships are already in place with companies like Qualcomm, Fujitsu, Marvell, and MediaTek, as well as design software firms Cadence and Synopsys, to foster a robust and open ecosystem. This approach allows Nvidia to remain central to the future of AI infrastructure, even when systems incorporate chips from other vendors.
Nvidia is also solidifying its presence in Taiwan, establishing a new office complex near Taipei that will serve as its overseas headquarters. The company is collaborating with Foxconn to build an "AI factory" in Taiwan, which will utilize 10,000 Nvidia Blackwell GPUs. This facility is intended to bolster Taiwan's AI infrastructure and support local organizations in adopting AI technologies across various sectors. TSMC, Nvidia's primary chip supplier, plans to leverage this supercomputer for research and development, aiming to develop the next generation of AI chips. Furthermore, Nvidia is working with Taiwan's National Center for High-Performance Computing (NCHC) to develop a new AI supercomputer. This system will feature over 1,700 GPUs, GB200 NVL72 rack-scale systems, and an HGX B300 system based on the Blackwell Ultra platform, all connected via Quantum InfiniBand networking. Expected to launch later this year, the supercomputer promises an eightfold performance increase over its predecessor for AI workloads, providing researchers with enhanced capabilities to advance their projects. Academic institutions, government agencies, and small businesses in Taiwan will be able to apply for access to the supercomputer to accelerate their AI initiatives. Recommended read:
References :
@blogs.nvidia.com
//
Cadence has unveiled the Millennium M2000 Supercomputer, a powerhouse featuring NVIDIA Blackwell systems, aimed at revolutionizing AI-driven engineering design and scientific simulations. This supercomputer integrates NVIDIA HGX B200 systems and NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, coupled with NVIDIA CUDA-X software libraries and Cadence's optimized software. The result is a system capable of delivering up to 80 times higher performance compared to its CPU-based predecessors, marking a significant leap forward in computational capability for electronic design automation, system design, and life sciences workloads.
This collaboration between Cadence and NVIDIA is set to enable engineers to conduct massive simulations, leading to breakthroughs in various fields, including the design and development of autonomous machines, drug molecules, semiconductors, and data centers. NVIDIA's founder and CEO, Jensen Huang, highlighted the transformative potential of AI, stating that it will infuse every aspect of business and product development. Huang also announced NVIDIA's plans to acquire ten Millennium Supercomputer systems based on the NVIDIA GB200 NVL72 platform to accelerate the company’s chip design workflows, emphasizing the importance of this technology for NVIDIA's future endeavors. In related news, the open-source OpenSearch software has launched its 3.0 version, which includes GPU acceleration to enhance AI workloads through its new OpenSearch Vector Engine. This update leverages NVIDIA GPUs to improve search performance with large-scale vector workloads and reduce index build times, aiming to address scalability issues common in vector databases. OpenSearch 3.0 also supports Anthropic PBC’s Model Context Protocol, facilitating the integration of large language models with external data. The Millennium M2000 Supercomputer harnesses accelerated software from NVIDIA and Cadence for applications including circuit simulation, computational fluid dynamics, data center design and molecular design. Recommended read:
References :
@blogs.nvidia.com
//
References:
NVIDIA Newsroom
, blogs.nvidia.com
,
NVIDIA's Blackwell platform is set to revolutionize data center cooling with a focus on water efficiency. The new platform introduces direct-to-chip liquid cooling, which dramatically reduces water consumption compared to traditional air-cooled systems. NVIDIA claims this innovative approach offers a 300x improvement in water efficiency. This is crucial as the increasing compute power required by AI and HPC applications is driving a global shift towards larger, more power-hungry data centers and AI factories.
This cooling solution addresses the escalating energy demands and environmental concerns associated with AI infrastructure. Historically, cooling has accounted for up to 40% of a data center's electricity consumption. The Blackwell platform's liquid cooling technology captures heat directly at the source, cycling it through a coolant distribution unit and liquid-to-liquid heat exchanger, before transferring it to a facility cooling loop. This method allows data centers to operate effectively at warmer water temperatures, reducing or eliminating the need for mechanical chillers in many climates, leading to significant cost savings and reduced energy consumption. The NVIDIA GB200 NVL72 and GB300 NVL72 systems, built on the Blackwell platform, utilize this direct-to-chip liquid cooling technology. Unlike evaporative or immersion cooling, this is a closed-loop system, meaning the coolant doesn't evaporate or require replacement due to loss from phase change, further conserving water. These rack-scale systems are designed to handle the demanding tasks of trillion-parameter large language model inference, making them ideal for running AI reasoning models while efficiently managing energy costs and heat. This advancement not only ensures optimal performance of AI servers but also promotes a more sustainable and environmentally friendly AI infrastructure. Recommended read:
References :
@hothardware.com
//
References:
hothardware.com
, insideAI News
,
Nvidia's Blackwell architecture is making significant strides in both the AI and gaming sectors. The GeForce RTX 5060 Ti, priced at $429, brings the Blackwell architecture to mainstream gamers, targeting 1440p gaming with power efficiency and overclocking headroom. Reviews indicate that the RTX 5060 Ti is built around the GB206 GPU and is the full implementation of the chip, with roughly 21.9 billion transistors manufactured on TSMC's 4N node. It is arranged into 3 GPCs. It is the first of the Blackwell cards to target under $500 and should see peak performance realized in advanced AI rendering techniques like DLSS4 with multi-frame generation.
CoreWeave, a GPU cloud platform, is among the first to bring Nvidia's Grace Blackwell GB200 NVL72 systems online at scale. Companies like Cohere, IBM, and Mistral AI are leveraging these systems for model training and deployment. Cohere is using Grace Blackwell Superchips to develop secure enterprise AI applications, with reported performance increases of up to 3x in training for 100 billion-parameter models. IBM is scaling its deployment to thousands of Blackwell GPUs on CoreWeave to train its Granite open-source AI models for IBM watsonx Orchestrate. Mistral AI is also utilizing the Blackwell GPUs to build the next generation of open-source AI models, reporting a 2x improvement in performance for dense model training. However, Nvidia faces challenges due to U.S. government restrictions on exports to China. The company is writing off $5.5 billion in GPUs as the U.S. government chokes off supply of H20s to China, highlighting geopolitical impacts on the tech industry. The U.S. government's concern stems from the potential use of these processors in Chinese supercomputers. In response, Nvidia is reportedly working to onshore the manufacturing of chips and other components, as well as the assembly of systems, to the United States. Recommended read:
References :
NVIDIA Newsroom@NVIDIA Blog
//
Nvidia has announced a major initiative to manufacture its AI supercomputers entirely within the United States. The company aims to produce up to $500 billion worth of AI infrastructure in the U.S. over the next four years, partnering with major manufacturing firms like Taiwan Semiconductor Manufacturing Co (TSMC), Foxconn, Wistron, Amkor, and SPIL. This move marks the first time Nvidia will carry out chip packaging and supercomputer assembly entirely within the United States. The company sees this effort as a way to meet the increasing demand for AI chips, strengthen its supply chain, and boost resilience.
Nvidia is commissioning over a million square feet of manufacturing space to build and test Blackwell chips in Arizona and assemble AI supercomputers in Texas. Production of Blackwell chips has already begun at TSMC’s chip plants in Phoenix, Arizona. The company is also constructing supercomputer manufacturing plants in Texas, partnering with Foxconn in Houston and Wistron in Dallas, with mass production expected to ramp up within the next 12-15 months. These facilities are designed to support the deployment of "gigawatt AI factories", data centers specifically built for processing artificial intelligence. CEO Jensen Huang emphasized the significance of bringing AI infrastructure manufacturing to the U.S., stating that "The engines of the world’s AI infrastructure are being built in the United States for the first time." Nvidia also plans to deploy its own technologies to optimize the design and operation of the new facilities, utilizing platforms like Omniverse to simulate factory operations and Isaac GR00T to develop automated robotics systems. The company said domestic production could help drive long-term economic growth and job creation. Recommended read:
References :
NVIDIA Newsroom@NVIDIA Blog
//
Nvidia has announced plans to manufacture its AI supercomputers entirely within the United States, marking the first time the company will conduct chip packaging and supercomputer assembly domestically. The move, driven by increasing global demand for AI chips and the potential impact of tariffs, aims to establish a resilient supply chain and bolster the American AI ecosystem. Nvidia is partnering with major manufacturing firms including TSMC, Foxconn, and Wistron to construct and operate these facilities.
Mass production of Blackwell chips has already commenced at TSMC's Phoenix, Arizona plant. Nvidia is constructing supercomputer manufacturing plants in Texas, partnering with Foxconn in Houston and Wistron in Dallas. These facilities are expected to ramp up production within the next 12-15 months. More than a million square feet of manufacturing space has been commissioned to build and test NVIDIA Blackwell chips in Arizona and AI supercomputers in Texas. The company anticipates producing up to $500 billion worth of AI infrastructure in the U.S. over the next four years through these partnerships. This includes designing and building "gigawatt AI factories" to produce NVIDIA AI supercomputers completely within the US. CEO Jensen Huang stated that American manufacturing will help meet the growing demand for AI chips and supercomputers, strengthen the supply chain and improve resiliency. The White House has lauded Nvidia's decision as "the Trump Effect in action". Recommended read:
References :
NVIDIA Newsroom@NVIDIA Blog
//
Nvidia has announced plans to manufacture AI supercomputers entirely within the United States for the first time. The company is working with manufacturing partners to design and build factories that will produce NVIDIA AI supercomputers – or "AI factories" – on U.S. soil. This initiative includes a projected investment of up to $500 billion over the next four years and aims to establish a comprehensive domestic supply chain for AI infrastructure. Nvidia's partners include industry giants such as TSMC, Foxconn, Wistron, Amkor, and SPIL, deepening their ties with NVIDIA while expanding their global footprint and enhancing supply chain resilience.
The company has already commissioned over a million square feet of manufacturing space in Arizona and Texas to build and test NVIDIA Blackwell chips and assemble AI supercomputers. NVIDIA Blackwell chips have started production at TSMC’s chip plants in Phoenix, Arizona. NVIDIA is building supercomputer manufacturing plants in Texas, with Foxconn in Houston and with Wistron in Dallas. Mass production at both plants is expected to ramp up in the next 12-15 months. Amkor and SPIL will handle packaging and testing operations in Arizona. This move marks the first time Nvidia will be building AI supercomputers entirely in the US. Jensen Huang, founder and CEO of NVIDIA, emphasized the strategic importance of this initiative. "The engines of the world's AI infrastructure are being built in the United States for the first time," Huang stated. "Adding American manufacturing helps us better meet the incredible and growing demand for AI chips and supercomputers, strengthens our supply chain, and boosts our resiliency." NVIDIA also intends to deploy its own technologies, such as Omniverse and Isaac GR00T, to optimize factory operations and automate manufacturing processes. Manufacturing NVIDIA AI chips and supercomputers for American AI factories is expected to create hundreds of thousands of jobs and drive trillions of dollars in economic security over the coming decades. Recommended read:
References :
@www.intel.com
//
NVIDIA's Blackwell platform has dominated the latest MLPerf Inference V5.0 benchmarks, showcasing significant performance improvements in AI reasoning. The NVIDIA GB200 NVL72 system, featuring 72 Blackwell GPUs, achieved up to 30x higher throughput on the Llama 3.1 405B benchmark compared to the NVIDIA H200 NVL8 submission. This was driven by more than triple the performance per GPU and a 9x larger NVIDIA NVLink interconnect domain. The latest MLPerf results reflect the shift toward reasoning in AI inference.
Alongside this achievement, NVIDIA is open-sourcing the KAI Scheduler, a Kubernetes GPU scheduling solution, as part of its commitment to open-source AI innovation. Previously a core component of the Run:ai platform, KAI Scheduler is now available under the Apache 2.0 license. This solution is designed to address the unique challenges of managing AI workloads that utilize both GPUs and CPUs. According to NVIDIA, this will help in managing fluctuating GPU demands, which traditional resource schedulers struggle to handle. Recommended read:
References :
@www.intel.com
//
NVIDIA is making strides in both agentic AI and open-source initiatives. Jacob Liberman, director of product management at NVIDIA, explains how agentic AI bridges the gap between powerful AI models and practical enterprise applications. Enterprises are now deploying AI agents to free human workers from time-consuming and error-prone tasks, allowing them to focus on high-value work that requires creativity and strategic thinking. NVIDIA AI Blueprints help enterprises build their own AI agents.
NVIDIA has announced the open-source release of the KAI Scheduler, a Kubernetes-native GPU scheduling solution, now available under the Apache 2.0 license. Originally developed within the Run:ai platform, the KAI Scheduler is now available to the community while also continuing to be packaged and delivered as part of the NVIDIA Run:ai platform. The KAI Scheduler is designed to optimize the scheduling of GPU resources and tackle challenges associated with managing AI workloads on GPUs and CPUs. Recommended read:
References :
Jaime Hampton@BigDATAwire
//
NVIDIA's GTC 2025 showcased significant advancements in AI, marked by the unveiling of the Blackwell Ultra GPU and the Vera Rubin roadmap extending through 2027. CEO Jensen Huang emphasized a 40x AI performance leap with the Blackwell platform compared to its predecessor, Hopper, highlighting its crucial role in inference workloads. The conference also introduced open-source ‘Dynamo’ software and advancements in humanoid robotics, demonstrating NVIDIA’s commitment to pushing AI boundaries.
The Blackwell platform is now in full production, meeting incredible customer demand, and the Vera Rubin roadmap details the next generation of superchips expected in 2026. Huang also touted new DGX systems, highlighting the push towards photonic switches to handle growing data demands efficiently. Blackwell Ultra will offer 288GB of memory. NVIDIA claims the GB300 chip brings 1.5x more AI performance than the NVIDIA GB200. These advancements aim to bolster AI reasoning capabilities and energy efficiency, positioning NVIDIA to maintain its dominance in AI infrastructure. Recommended read:
References :
|
BenchmarksBlogsResearch Tools |