Today One Stop Systems (OSS) announced the availability of a new OSS PCIe 4.0 value expansion system incorporating up to eight of the latest NVIDIA V100S Tensor Core GPUs. As the newest member of the company’s AI on the Fly product portfolio, the system delivers data center capabilities to HPC and AI edge deployments in the field or for mobile applications.
The 4U value expansion system adds massive compute capability to any Gen 3 or Gen 4 server via two OSS PCIe x16 Gen 4 links. The links can support an unprecedented 512 Gpbs of aggregated bandwidth to the GPU complex.
Our new compute acceleration expansion platform demonstrates how we continue to lead the industry in delivering the latest in high-performance technology for mission critical edge applications,” said OSS CEO Steve Cooper. “By combining our proprietary technology and designs with the latest NVIDIA GPUs, we lead the market in Gen 4 expansion and powering AI on the Fly applications.”
The expansion system features 10 PCIe 4.0 slots and 4,000 watts of load sharing power. In conjunction with V100S Tensor Core GPUs, it delivers up to 1,040 teraFLOPS of tensor performance and 65.6 teraFLOPS of double precision performance, accelerating both computational science and data science.
The NVIDIA V100S Tensor Core GPU brings CUDA Cores and Tensor Cores in a unified architecture to enable mixed-precision computing. This feature is especially useful for AI training, where operations are run in FP16 precision, and results are accumulated in FP32 precision, which delivers significant speedups while preserving accuracy. In addition, NVIDIA V100S GPUs offer FP64 precision for scientific computing applications like simulations, and INT8 precision for AI inference. The OSS expansion system provides this GPU capability for military, automotive, aerospace and industrial edge applications.
In this video from SC19, Jaan Mannik from One Stop Systems describes how the company delivers “AI on the Fly.”