If optimized AI workflows are like a superbly tuned orchestra — the place every element, from {hardware} infrastructure to software program libraries, hits precisely the appropriate observe — then the long-standing concord between NVIDIA and Microsoft is music to builders’ ears.
The most recent AI fashions developed by Microsoft, together with the Phi-3 household of small language fashions, are being optimized to run on NVIDIA GPUs and made out there as NVIDIA NIM inference microservices. Different microservices developed by NVIDIA, such because the cuOpt route optimization AI, are frequently added to Microsoft Azure Market as a part of the NVIDIA AI Enterprise software program platform.
Along with these AI applied sciences, NVIDIA and Microsoft are delivering a rising set of optimizations and integrations for builders creating high-performance AI apps for PCs powered by NVIDIA GeForce RTX and NVIDIA RTX GPUs.
Constructing on the progress shared at NVIDIA GTC, the 2 firms are furthering this ongoing collaboration at Microsoft Construct, an annual developer occasion, happening this 12 months in Seattle by Could 23.
Accelerating Microsoft’s Phi-3 Fashions
Microsoft is increasing its household of Phi-3 open small language fashions, including small (7-billion-parameter) and medium (14-billion-parameter) fashions much like its Phi-3-mini, which has 3.8 billion parameters. It’s additionally introducing a brand new 4.2-billion-parameter multimodal mannequin, Phi-3-vision, that helps pictures and textual content.
All of those fashions are GPU-optimized with NVIDIA TensorRT-LLM and out there as NVIDIA NIMs, that are accelerated inference microservices with a normal utility programming interface (API) that may be deployed wherever.
APIs for the NIM-powered Phi-3 fashions can be found at ai.nvidia.com and thru NVIDIA AI Enterprise on the Azure Market.
NVIDIA cuOpt Now Out there on Azure Market
NVIDIA cuOpt, a GPU-accelerated AI microservice for route optimization, is now out there in Azure Market through NVIDIA AI Enterprise. cuOpt options massively parallel algorithms that allow real-time logistics administration for delivery companies, railway programs, warehouses and factories.
The mannequin has set two dozen world data on main routing benchmarks, demonstrating the very best accuracy and quickest occasions. It might save billions of {dollars} for the logistics and provide chain industries by optimizing automobile routes, saving journey time and minimizing idle durations.
By way of Azure Market, builders can simply combine the cuOpt microservice with Azure Maps to assist teal-time logistics administration and different cloud-based workflows, backed by enterprise-grade administration instruments and safety.
Optimizing AI Efficiency on PCs With NVIDIA RTX
The NVIDIA accelerated computing platform is the spine of contemporary AI — serving to builders construct options for over 100 million Home windows GeForce RTX-powered PCs and NVIDIA RTX-powered workstations worldwide.
NVIDIA and Microsoft are delivering new optimizations and integrations to Home windows builders to speed up AI in next-generation PC and workstation purposes. These embrace:
- Quicker inference efficiency for big language fashions through the NVIDIA DirectX driver, the Generative AI ONNX Runtime extension and DirectML. These optimizations, out there now within the GeForce Recreation Prepared, NVIDIA Studio and NVIDIA RTX Enterprise Drivers, ship as much as 3x sooner efficiency on NVIDIA and GeForce RTX GPUs.
- Optimized efficiency on RTX GPUs for AI fashions like Steady Diffusion and Whisper through WebNN, an API that permits builders to speed up AI fashions in internet purposes utilizing on-device {hardware}.
- With Home windows set to assist PyTorch by DirectML, hundreds of Hugging Face fashions will work in Home windows natively. NVIDIA and Microsoft are collaborating to scale efficiency on greater than 100 million RTX GPUs.
Be a part of NVIDIA at Microsoft Construct
Convention attendees can go to NVIDIA sales space FP28 to satisfy developer consultants and expertise dwell demos of NVIDIA NIM, NVIDIA cuOpt, NVIDIA Omniverse and the NVIDIA RTX AI platform. The sales space additionally highlights the NVIDIA MONAI platform for medical imaging workflows and NVIDIA BioNeMo generative AI platform for drug discovery — each out there on Azure as a part of NVIDIA AI Enterprise.
Attend periods with NVIDIA audio system to dive into the capabilities of the NVIDIA RTX AI platform on Home windows PCs and uncover easy methods to deploy generative AI and digital twin instruments on Microsoft Azure.
And join the Developer Showcase, happening Wednesday, to find how builders are constructing modern generative AI utilizing NVIDIA AI software program on Azure.
