Microsoft has launched the NVIDIA GB300 NVL72 supercluster, the first at-scale production cluster featuring over 4,600 NVIDIA Blackwell Ultra GPUs integrated through the advanced NVIDIA InfiniBand network. This innovative infrastructure marks a significant leap in AI capabilities, enabling Fast model training in weeks rather than months, and supports models with hundreds of trillions of parameters. The ND GB300 v6 virtual machines offer enhanced performance for agentic AI systems and multimodal generative models, boasting up to 1,440 petaflops of performance, 130 terabytes per second of NVLink bandwidth, and advanced cooling systems to maintain optimal thermal stability. Collaborations with NVIDIA ensure a fully integrated system that maximizes GPU utilization, reducing latency and overhead during training. Azure’s commitment to AI infrastructure positions it to meet the rapidly evolving demands of frontier AI, making the ND GB300 v6 VMs the new gold standard for accelerated computing. Stay tuned for further advancements and benchmarks.
Source link
