英伟达 GH200超级计算机：集成256个GH200芯片

英伟达 GH200超级计算机：集成256个GH200芯片

这些芯片是英伟达新推出的 DGX GH200 人工智能超级计算平台和 MGX 系统的核心组件，它们专为处理海量的生成型人工智能任务而设计。英伟达还宣布了其新的 Spectrum-X 以太网网络平台，专为人工智能服务器和超级计算集群而优化。

Grace Hopper 超级芯片是英伟达开发的基于 Arm 架构的 CPU+GPU 集成方案，它将 72 核的 Grace CPU、Hopper GPU、96GB 的 HBM3 和 512 GB 的 LPDDR5X 集成在同一个封装中，共有 2000 亿个晶体管。这种组合提供了 CPU 和 GPU 之间惊人的数据带宽，高达 1 TB / s，为某些内存受限的工作负载提供了巨大的优势。

DGX GH200 人工智能超级计算平台是英伟达针对最高端的人工智能和高性能计算工作负载而设计的系统和参考架构，目前的 DGX A100 系统只能将八个 A100 GPU 联合起来作为一个单元，考虑到生成型人工智能的爆炸式增长，英伟达的客户迫切需要更大、更强大的系统，DGX GH200 就是为了提供最大的吞吐量和可扩展性而设计的，它通过使用英伟达的定制 NVLink Switch 芯片来避免标准集群连接选项（如 InfiniBand 和以太网）的限制。

英伟达 GH200超级计算机：集成256个GH200芯片

These chips are the core components of Nvidia's new DGX GH200 AI supercomputing platform and MGX system, which are designed to handle massive generative AI tasks. Nvidia also announced its new Spectrum-X Ethernet networking platform, optimized for AI servers and supercomputing clusters.

Grace Hopper Super chip is an ARM-based CPU+GPU integration solution developed by Nvidia, which integrates 72 core Grace CPU, Hopper GPU, 96GB HBM3 and 512 GB LPDDR5X in the same package. There are 200 billion transistors. This combination provides amazing data bandwidth between the CPU and GPU, up to 1 TB/s, providing a huge advantage for certain memory-constrained workloads.

The DGX GH200 AI Supercomputing platform is Nvidia's system and reference architecture designed for the highest end AI and high-performance computing workloads, and the current DGX A100 system can only combine eight A100 Gpus as a single unit, considering the explosive growth of generative AI. Nvidia's customers are desperate for larger, more powerful systems, and the DGX GH200 is designed to deliver maximum throughput and scalability by using Nvidia's custom NVLink Switch chip to avoid the limitations of standard cluster connectivity options such as InfiniBand and Ethernet.