NVIDIA today introduced NVIDIA HGX-2, a unified computing platform for both artificial intelligence and high performance computing.
Announced by NVIDIA CEO Jensen Huang at the GPU Technology Conference Taiwan Wednesday, the new "building block" cloud-server platform will let server manufacturers create more powerful systems around NVIDIA GPUs for high performance computing and AI.
"Computing demand is greater than ever, so more than ever, we need this computing performance to continue to extend, we need to extend Moore's law," Huang told a packed house of more than 2,000 technologists, developers, researchers, government officials and media in Taipei.
The HGX-2 cloud server platform, with multi-precision computing capabilities, provides flexibility to support the future of computing. It allows high-precision calculations using FP64 and FP32 for scientific computing and simulations, while also enabling FP16 and Int8 for AI training and inference.
At the HGX-2's heart is the NVIDIA Tesla V100 GPU - equipped with 32GB of high-bandwidth memory capacity - to deliver 125 teraflops of deep learning performance.
Weave together as many as 16 Tesla V100 GPUs with NVSwitch and the result is what Huang calls "the world's largest GPU."
"Every one of the GPUs can talk to every one of the GPUs simultaneously at a bandwidth of 300 GB/s, 10 times PCI Express," Huang said. "So everyone can talk to each other all at the same time."
Huang also detailed NVIDIA's new NVIDIA DGX-2, the first system built using the HGX-2 server platform. The 350-pound machine offers 2 petaflops of computing power and 512GB of HBM2 memory.
HGX-2 has achieved record AI training speeds of 15,500 images per second on the ResNet-50 training benchmark, and Nvidia says it can replace up to 300 CPU-only servers.
The first system built using HGX-2 was the recently announced NVIDIA DGX-2.
HGX-2 comes a year after the launch of the original NVIDIA HGX-1, at Computex 2017.
Lenovo, QCT, Supermicro and Wiwynn announced plans to bring their own HGX-2-based systems to market later this year.
Additionally, Foxconn, Inventec, Quanta and Wistron, are designing HGX-2-based systems, also expected later this year, for use in large cloud datacenters.