NVIDIA’s Hopper architecture GH100 GPU chip area set a record: up to 1000mm²
It is rumored that Nvidia’s Hopper architecture is its first GPU based on a multi-chip module design (MCM), which will be manufactured using TSMC’s 5nm process and CoWoS advanced packaging, support for HBM2e, and other connectivity features, aimed at the data center, is expected to debut in mid-2022, competing against Intel’s Xe-HP architecture and AMD’s CDNA 2 architecture.
The largest GPU currently in production is the GA100 from Nvidia’s Ampere architecture, measuring 862mm². However, Twitter user @kopite7kimi revealed that the GH100 based on the Hopper architecture has a chip area of about 1000mm², easily surpassing its predecessors by more than 100mm², which will be the largest GPU ever. Considering the design of the introduction of MCM, the new top-level computing card will have two computing modules, which means that the die area is doubled, with a total of 288 units of SMs. In addition, Twitter user @greymon55 said that the GH100 still adopts the traditional design. As for the truth, it will be revealed by NVIDIA.
It is rumored that Nvidia also has a secret weapon, which is the COPA solution based on the Hopper architecture. Nvidia researchers previously published an article detailing how Nvidia is exploring how to deploy multi-chip designs for future products, including a reference to related designs. Nvidia will have two designs based on the same architecture, but for the high-performance computing (HPC) and deep learning (DL) segments. HPC will use the standard scheme, DL will have a huge independent cache connected to the GPU.
It is understood that Nvidia may unveil the Hopper architecture GPU at GTC 2022, and we can learn more about it at that time.
