NVIDIA founder and CEO Huang Renxun said in a keynote speech that as the computing demand increases rapidly, the method of expanding the CPU has become outdated, and the server complexity of building high-performance computing (HPC) and AI keeps rising, and it has almost reached the system. The limit of design; NVIDIA HGX-2 combined with GPU equipped with Tensor Core, through the integration of high-performance computing and AI, can provide a powerful multi-purpose computing platform in the industry to overcome the above challenges.
It is reported that the HGX-2 cloud server platform is capable of multiple precision operations and provides strong flexibility to support future computing needs. It also uses FP64 and FP32 for high-precision operations for scientific and analog computing, and can also use FP16. The Int8 format performs AI training and inference, in response to more and more high-performance computing combined with application requirements derived from AI.
In addition, the product can achieve an AI training speed of 15,500 frames per second through the ResNet-50. The performance is more than sufficient to replace the server with up to 300 CPUs. It also has an NVIDIA NVSwitchTM mesh interconnect architecture that enables smooth chaining of 16 units. The NVIDIA Tesla V100 GPU with built-in Tensor Core functions as a huge GPU, and the system built with HGX-2 is the NVIDIA DGX-2 introduced recently.
Huang Renxun pointed out that HGX-2 is part of the NVIDIA GPU acceleration server platform series. This series of products is connected to the entire data center server system and is suitable for each large market. It can recommend the most suitable GPU for different AI, high-performance computing and speeding up operations. Combined with CPU configuration. Such as HGX-T for hyperscale training and HPC; HGX-1 for large-scale inference and intelligent image analysis (IVA); and SCX-E for data center, HPC, IVA, virtual desktop infrastructure ( VDI) etc.