Developers: | Mellanox Technologies |
Last Release Date: | 2020/11/16 |
Technology: | Network Health Monitoring - Monitoring of network or management of health performance of IT Infrastructure, SCS, Data processing centers are technologies for DPC |
2020: Presentation of NVIDIA Mellanox 400G InfiniBand
On November 16, 2020 the NVIDIA company provided the next generation of the NVIDIA Mellanox 400G InfiniBand platform which provides to solutions AI developers and researchers high network performance for the solution of the most difficult global tasks.
As requirements to calculations continue to grow exponential in such areas as development of drugs, researches of climate and genomics, NVIDIA Mellanox 400G InfiniBand accelerates this work due to large increase of performance on the only platform in the world with support of intra network calculations.
The seventh generation of Mellanox InfiniBand guarantees ultralow delays, doubles capacity of data to 400 GB / with and adds algorithms of network calculations of NVIDIA for additional acceleration.
World leading manufacturers of the equipment are going to integrate NVIDIA Mellanox 400G InfiniBand into the corporate solutions. These obligations are supplemented with extensive support from the leading partners in storage systems, including DDN and IBM Storage.
The most important work of our clients is based on the artificial intelligence and harder and harder applications requiring faster, intelligent and scalable networks. Huge capacity and intelligent mechanisms of acceleration NVIDIA Mellanox 400G InfiniBand allocate the systems of high-performance computing, artificial intelligence and hyper scalable cloud solutions with unsurpassed performance at smaller costs and complexity, - Gilad Shainer, the senior vice president for network NVIDIA technologies speaks |
Technical characteristics and product availability
Transfer of computing loading in network is crucial for problems of AI. The NVIDIA Mellanox SHARP technology of third generation allows to unload and accelerate transactions of deep learning in InfiniBand network that leads to acceleration of tasks AI by 32 times. The combination to a program stack of NVIDIA Magnum IO provides acceleration of scientific calculations directly from a box.
Edge-switches based on architecture of Mellanox InfiniBand have cumulative bidirectional capacity in 51.2 Tbit / with with a record performance more than 66.5 billion packets per second. The modular switches based on Mellanox InfiniBand will provide total bidirectional capacity in 1.64 petabits per second that 5 times higher, than at the previous generation.
The architecture of Mellanox InfiniBand is based on the industry standards guaranteeing the backward and future compatibility and protection of investments into data processing centers. Solutions on the basis of this architecture will become available in the second quarter 2021.