Microsoft and Nvidia Announce NDv2: The World's Largest GPU-Accelerated Cloud-Based Supercomputer

Microsoft's commitment to the cloud leaves no room for doubt. Azure takes center stage, the streaming game arrives with Project xCloud, an alternative to Google Stadia that wins by a landslide on paper and now we know that Microsoft and Nvidia have announced NDv2, the cloud-based supercomputer and world's largest GPU-accelerated drive

The announcement took place at the SC19 event held in Denver dedicated to supercomputing. It has been the joint work of Microsoft and Nvidia that has made possible the arrival of the new NDv2 supercomputer.

Super Cloud Computing

There were two more announcements at the event, where Nvidia brought to the stage the Nvidia Magnum IO, a suite of software for help scientists and researchers working with artificial intelligence to process massive amounts of data in minutes. Along with it, the company introduced a design platform through which companies can more quickly and efficiently create GPU-accelerated Arm-based servers. And next to them the new NDv2 supercomputer.

For those who don't know them, NDv2 equipment is systems designed and created for very specific tasks They seek to execute loads of most demanding distributed tasks of HPC (High Performance Computing), AI as well as machine learning tasks.

The NDv2 is the world's largest GPU-accelerated cloud-based supercomputer Inside it is packed with 8 NVIDIA Tesla V100 GPUs NVLink interconnects, each with 32 GB of HBM2 memory, 40 non-hyperthreaded cores of the Intel Xeon Platinum 8168 processor, and 672 GiB of system memory. This supercomputer features 100 Gigabit EDR InfiniBand from Mellanox allowing it to scale up to 800 NVIDIA V100 Tensor Core GPUs interconnected with Mellanox InfiniBand.

With NDv2 Bert was tested, a conversational AI model and the power was demonstrated in the fact that it only took three hours to exploit its functions. It is an example of the power that these equipments offer.

This movement by Nvidia and Microsoft comes on the one hand when Intel sees its dominance in terms of data centers threatened and also with AMD as an alternative to Intel CPUs thanks to the SoC Arm.But it is also that and almost in parallel, Amazon Web Service reported that it would launch some of its EC2 instances in the cloud supported by AMD EPYC Rome processors or that Intel would reveal its Pone Vecchio GPU for data centers.

More information | Nvidia