论文部分内容阅读
NVIDIA首款基于Volta架构GPU技术的产品终于正式登场。在这次的GTC大会上,该公司推出了无论在性能亦或是价位上都相当强大的资料中心用显卡:NVIDIATesla V100。做为P100的升级款式,其使用上了NVIDIA第七代的最新架构、采用12nm制程,搭载5120 CUDA核心、16GB HBM2内存与NVLink 2.0的连接界面,带来为更精密的AI与深度学习技术当后盾的15teraflops运算能力。在最新运算核心发布的同时,NVIDIA也宣布了直接搭载8组Tesla V100、要价美金14.9万
NVIDIA’s first product based on Volta-based GPU technology finally debuted. At this GTC conference, the company introduced NVIDIATesla V100, a graphics card for data centers that is both fairly powerful and affordable. As an upgrade to the P100, it uses the latest NVIDIA Gen 7 architecture with a 12nm process, a 5120 CUDA core, and a 16GB HBM2 memory interface with NVLink 2.0 for more sophisticated AI and deep learning Backed by 15teraflops computing power. At the same time as the latest computing core was released, NVIDIA also announced a direct shipment of eight Tesla V100s with a asking price of $ 149,000