- Skrive om SM-SM interconnect i clusters
- Clustering begrensninger
- API for å anvende Thread Block Clusters
Hopper Power Efficiency
ARM and Grace
Announced for 2023, Nvidia´s ARM-based Grace CPU seems also noteworthy. USA´s Los Alamos National Laboratory and the Swiss National Computing Centre have already announced plans for Grace-based supercomputers \cite{pressure}. ARM, originally a UK IP-only company is now owned by the Japanese SoftBank Group Corp (SBC). Nvidia also tried buying ARM from them in a Fall 2020 agreement, but the deal was terminated in Feb. 2022 due to "regulartory challenges" in both US and China.
Intel still dominates the top500.org list of the world´s largest supercmputers, but in 2021 Japan’s ARM-based Fugaku supercomputer (> seven million cores, running at 442 petaflops) took the top spot.
The
Grace CPU Superchip https://www.nvidia.com/en-us/data-center/grace-cpu/\cite{intro} features two Grace cores connected via the NVLink-C2C technology thus providing
up to 144 Arm v9 CPU cores. It claims to be the World’s first CPU using LPDDR5x memory with ECC and 1TB/s total bandwidth. Its 900 GB/s coherent interface is 7X faster than PCIe Gen 5.
Nvidia´s Grace Hopper Superchip combines the Grace CPU and Hopper GPU architectures using Nvidia´s NVLink-C2C technology to deliver a coherent CPU+GPU memory model. The systeim is targeting accelerated AI and high performance computing (HPC) applications. The system can provida a 30X higher aggregate system memory bandwidth to GPU compared to DGX A100.
Both the Grace and Grace Hopper superchips will run Nvidia´s software stacks, including NVIDIA HPC, NVIDIA AI, and NVIDIA Omniverse™ .