Curated by THEOUTPOST
On Wed, 20 Nov, 12:10 AM UTC
3 Sources
[1]
Microsoft Unveils NVIDIA's "Blackwell-Based" High-End Azure AI Compute Platform; HPC-Focused Azure AMD EPYC With HBM As Well
Microsoft announced some "serious" developments around its AI compute portfolio at the recent "Ignite" event, revealing NVIDIA's Blackwell integration with Azure and new AMD EPYC "Genoa" chips with custom HBM memory. It won't be wrong to say that Microsoft is a front-runner when bringing AI-oriented services to the public, given that the firm's vast AI computing arsenal allows it to have an "exclusive" position in the markets. At the recent Microsoft Ignite event, the firm revealed the utilization of NVIDIA's Blackwell AI products for the Azure platform, notably with the latest Azure ND GB200 V6 VM series, which is the first virtual machine by the company based on Blackwell. Diving a bit more into technicals, the Azure ND GB200 V6 virtual machine reportedly features two GB200 Grace Blackwell Superchips, with each of them being equipped with two high-performance Blackwell GPUs and a Grace CPU, interlinked using Team Green's NVLink interface. Interestingly, with multiple NVLink trays onboard, Microsoft can manage to provide up to 18 compute servers, ultimately providing 72 of NVIDIA's Blackwell GPUs over a single platform. Scaling isn't an issue with the newest VMs, credited to Team Green's InfiniBand fabric system. The Azure ND GB200 V6 VMs are currently in their private preview stages with selected partners, and Microsoft plans to roll out the platform for a wider audience soon. Apart from this, Microsoft also announced a "CPU-based" virtual machine on the Azure platform, called the Azure HBv5, which specifically targets memory bandwidth-intensive HPC applications. To obtain the best workload and cost-efficiency ratios, Microsoft decided to team up with AMD, utilizing their 4th Gen EPYC server CPUs. Here are the detailed specifications: Microsoft has also revealed that they have managed to upscale the performance of their newest Azure HBv5 VM by up to 20 times from previous-gen counterparts. This indeed shows the dominance of AMD's EPYC platform in the markets, which has already managed to outsell Intel as well. Overall, Microsoft's AI portfolio is undoubtedly planning to be one of the best ones out there, allowing the firm to capitalize on the AI hype to its fullest.
[2]
Microsoft launches preview of Nvidia's hot new AI chips on Azure
Microsoft (MSFT-0.55%) chief executive Satya Nadella announced the first cloud private preview of Nvidia's Blackwell AI infrastructure on its cloud computing platform, Azure, during the company's developers' conference on Tuesday. The Azure ND GB200 V6 VM series is based on the Nvidia Blackwell platform. "Blackwell is pretty amazing. It's got this 72 GPUs on a single NVLink domain, and then you combine it with InfiniBand on the backend," Nadella said during his keynote at Microsoft Ignite. "These racks are optimized for the most cutting-edge training workloads and inference workloads." New servers from Microsoft will also be equipped with its new in-house security chip, Azure Integrated HSM, or hardware security module, that uses encryption and signing keys for protection. Meanwhile, Nvidia's customers reportedly have been worried about delays for the Blackwell AI platform. The chipmaker repeatedly has asked its suppliers to change the design of its custom-designed server racks in recent months because its Blackwell AI chips are overheating when connected, The Information reported, citing unnamed people familiar with the matter. The server racks combine 72 of the AI chips in hopes that larger AI models can be trained faster with more GPUs, or graphics processing units. The redesigns are being made later than usual in the production process, The Information reported, but Nvidia could still ship the server racks on schedule, which it set at the end of the first half of next year. Nvidia previously faced production and shipping delays for Blackwell due to a design flaw, Nvidia chief executive Jensen Huang said in October, after media reports surfaced in August. "It was functional, but the design flaw caused the yield to be low," Huang said. "It was 100% Nvidia's fault." Microsoft also introduced new purpose-built AI agents in Microsoft 365 Copilot, including an interpreter agent that allows for real-time interpretation in up to nine languages in Microsoft Teams meetings, which will launch in preview early next year. It also rolled out a project manager agent, now in preview, that can automatically create plans and oversee projects. In addition, the tech giant launched the autonomous AI agent capabilities that it announced in October. Customers can build their own autonomous agents in Copilot Studio that can "understand the nature of your work and act on your behalf," Microsoft said.
[3]
NVIDIA and Microsoft Showcase Blackwell Preview, Omniverse Industrial AI and RTX AI PCs at Microsoft Ignite
Grace Blackwell now on Azure, new workflows for industrial AI, and tools and features for RTX AI PCs accelerate AI development. NVIDIA and Microsoft today unveiled product integrations designed to advance full-stack NVIDIA AI development on Microsoft platforms and applications. At Microsoft Ignite, Microsoft announced the launch of the first cloud private preview of the Azure ND GB200 V6 VM series, based on the NVIDIA Blackwell platform. The Azure ND GB200 v6 will be a new AI-optimized virtual machine (VM) series and combines the NVIDIA GB200 NVL72 rack design with NVIDIA Quantum InfiniBand networking. In addition, Microsoft revealed that Azure Container Apps now supports NVIDIA GPUs, enabling simplified and scalable AI deployment. Plus, the NVIDIA AI platform on Azure includes new reference workflows for industrial AI and an NVIDIA Omniverse Blueprint for creating immersive, AI-powered visuals. At Ignite, NVIDIA also announced multimodal small language models (SLMs) for RTX AI PCs and workstations, enhancing digital human interactions and virtual assistants with greater realism. NVIDIA Blackwell Powers Next-Gen AI on Microsoft Azure Microsoft's new Azure ND GB200 V6 VM series will harness the powerful performance of NVIDIA GB200 Grace Blackwell Superchips, coupled with advanced NVIDIA Quantum InfiniBand networking. This offering is optimized for large-scale deep learning workloads to accelerate breakthroughs in natural language processing, computer vision and more. The Blackwell-based VM series complements previously announced Azure AI clusters with ND H200 V5 VMs, which provide increased high-bandwidth memory for improved AI inferencing. The ND H200 V5 VMs are already being used by OpenAI to enhance ChatGPT. Azure Container Apps Enables Serverless AI Inference With NVIDIA Accelerated Computing Serverless computing provides AI application developers increased agility to rapidly deploy, scale and iterate on applications without worrying about underlying infrastructure. This enables them to focus on optimizing models and improving functionality while minimizing operational overhead. The Azure Container Apps serverless containers platform simplifies deploying and managing microservices-based applications by abstracting away the underlying infrastructure. Azure Container Apps now supports NVIDIA-accelerated workloads with serverless GPUs, allowing developers to use the power of accelerated computing for real-time AI inference applications in a flexible, consumption-based, serverless environment. This capability simplifies AI deployments at scale while improving resource efficiency and application performance without the burden of infrastructure management. Serverless GPUs allow development teams to focus more on innovation and less on infrastructure management. With per-second billing and scale-to-zero capabilities, customers pay only for the compute they use, helping ensure resource utilization is both economical and efficient. NVIDIA is also working with Microsoft to bring NVIDIA NIM microservices to serverless NVIDIA GPUs in Azure to optimize AI model performance. NVIDIA Unveils Omniverse Reference Workflows for Advanced 3D Applications NVIDIA announced reference workflows that help developers to build 3D simulation and digital twin applications on NVIDIA Omniverse and Universal Scene Description (OpenUSD) -- accelerating industrial AI and advancing AI-driven creativity. A reference workflow for 3D remote monitoring of industrial operations is coming soon to enable developers to connect physically accurate 3D models of industrial systems to real-time data from Azure IoT Operations and Power BI. These two Microsoft services integrate with applications built on NVIDIA Omniverse and OpenUSD to provide solutions for industrial IoT use cases. This helps remote operations teams accelerate decision-making and optimize processes in production facilities. The Omniverse Blueprint for precise visual generative AI enables developers to create applications that let nontechnical teams generate AI-enhanced visuals while preserving brand assets. The blueprint supports models like SDXL and Shutterstock Generative 3D to streamline the creation of on-brand, AI-generated images. NVIDIA's collaboration with Microsoft extends to bringing AI capabilities to personal computing devices. At Ignite, NVIDIA announced its new multimodal SLM, NVIDIA Nemovision-4B Instruct, for understanding visual imagery in the real world and on screen. It's coming soon to RTX AI PCs and workstations and will pave the way for more sophisticated and lifelike digital human interactions. Plus, updates to NVIDIA TensorRT Model Optimizer (ModelOpt) offer Windows developers a path to optimize a model for ONNX Runtime deployment. TensorRT ModelOpt enables developers to create AI models for PCs that are faster and more accurate when accelerated by RTX GPUs. This enables large models to fit within the constraints of PC environments, while making it easy for developers to deploy across the PC ecosystem with ONNX runtimes. RTX AI-enabled PCs and workstations offer enhanced productivity tools, creative applications and immersive experiences powered by local AI processing. Full-Stack Collaboration for AI Development NVIDIA's extensive ecosystem of partners and developers brings a wealth of AI and high-performance computing options to the Azure platform. SoftServe, a global IT consulting and digital services provider, today announced the availability of SoftServe Gen AI Industrial Assistant, based on the NVIDIA AI Blueprint for multimodal PDF data extraction, on the Azure marketplace. The assistant addresses critical challenges in manufacturing by using AI to enhance equipment maintenance and improve worker productivity. At Ignite, AT&T will showcase how it's using NVIDIA AI and Azure to enhance operational efficiency, boost employee productivity and drive business growth through retrieval-augmented generation and autonomous assistants and agents.
Share
Share
Copy Link
Microsoft announces integration of NVIDIA's Blackwell AI chips in Azure and new AMD EPYC-powered HPC solutions, showcasing advancements in AI computing infrastructure.
Microsoft has unveiled a significant advancement in its AI compute portfolio at the recent "Ignite" event, introducing the Azure ND GB200 V6 VM series, the first virtual machine based on NVIDIA's Blackwell AI products 1. This new platform represents a major leap in AI computing capabilities, featuring two GB200 Grace Blackwell Superchips, each equipped with two high-performance Blackwell GPUs and a Grace CPU 1.
The Azure ND GB200 V6 VM offers impressive scalability, with the ability to provide up to 18 compute servers, delivering a total of 72 NVIDIA Blackwell GPUs on a single platform 1. This scalability is made possible by NVIDIA's InfiniBand fabric system, ensuring efficient performance across multiple nodes 1.
In addition to the Blackwell-based platform, Microsoft announced the Azure HBv5, a CPU-based virtual machine targeting memory bandwidth-intensive High-Performance Computing (HPC) applications 1. This new VM utilizes AMD's 4th Gen EPYC server CPUs, demonstrating Microsoft's strategy to optimize workload and cost-efficiency ratios 1.
Microsoft claims that the Azure HBv5 VM offers up to 20 times the performance of previous-generation counterparts, highlighting the significant advancements in HPC capabilities 1.
Microsoft is bolstering the security of its new servers with the introduction of the Azure Integrated HSM (Hardware Security Module), an in-house security chip that uses encryption and signing keys for enhanced protection 2.
The company is also expanding its AI offerings with new purpose-built AI agents in Microsoft 365 Copilot. These include an interpreter agent for real-time interpretation in up to nine languages during Microsoft Teams meetings, and a project manager agent capable of automatically creating plans and overseeing projects 2.
While the announcement of Blackwell integration is significant, reports suggest that NVIDIA has faced some challenges in the development process. The company has reportedly asked suppliers to modify the design of its custom-designed server racks due to overheating issues when connecting the Blackwell AI chips 2.
Despite these setbacks, NVIDIA CEO Jensen Huang has acknowledged the company's responsibility for previous delays and design flaws, stating, "It was functional, but the design flaw caused the yield to be low. It was 100% NVIDIA's fault" 2.
NVIDIA and Microsoft's collaboration extends beyond the Azure platform. NVIDIA announced new multimodal small language models (SLMs) for RTX AI PCs and workstations, aimed at enhancing digital human interactions and virtual assistants 3.
Furthermore, Azure Container Apps now supports NVIDIA GPUs, enabling simplified and scalable AI deployment. This integration allows developers to leverage accelerated computing for real-time AI inference applications in a flexible, serverless environment 3.
NVIDIA unveiled reference workflows to help developers build 3D simulation and digital twin applications on NVIDIA Omniverse and Universal Scene Description (OpenUSD). These workflows are designed to accelerate industrial AI and advance AI-driven creativity 3.
A notable addition is the reference workflow for 3D remote monitoring of industrial operations, which connects physically accurate 3D models of industrial systems to real-time data from Azure IoT Operations and Power BI 3.
This collaboration between Microsoft and NVIDIA represents a significant step forward in AI computing infrastructure, promising to accelerate AI development and deployment across various industries and applications.
Reference
[3]
Microsoft Azure becomes the first cloud platform to integrate NVIDIA's cutting-edge Blackwell GB200 AI servers, showcasing a significant leap in cloud computing and AI capabilities.
3 Sources
3 Sources
Microsoft announces two new custom-designed chips for data centers, along with advanced cooling and power delivery technologies, to enhance AI capabilities, security, and energy efficiency in its Azure cloud infrastructure.
3 Sources
3 Sources
Nvidia's GTC 2025 showcases the company's latest AI innovations and strategies, highlighting both its dominant position and the emerging challenges in the rapidly evolving AI landscape.
27 Sources
27 Sources
Microsoft has purchased 485,000 of Nvidia's Hopper chips in 2024, more than doubling its nearest competitors' orders. This aggressive move positions Microsoft at the forefront of AI infrastructure development, outpacing tech giants like Meta, Amazon, and Google.
7 Sources
7 Sources
Nvidia announces the Blackwell Ultra B300 GPU, offering 1.5x faster performance than its predecessor with 288GB HBM3e memory and 15 PFLOPS of dense FP4 compute, designed to meet the demands of advanced AI reasoning and inference.
9 Sources
9 Sources
The Outpost is a comprehensive collection of curated artificial intelligence software tools that cater to the needs of small business owners, bloggers, artists, musicians, entrepreneurs, marketers, writers, and researchers.
© 2025 TheOutpost.AI All rights reserved