How to comprehensively evaluate the performance of an arm CPU?

I would like to conduct a comprehensive performance evaluation of the newly launched Nvidia Grace Superchip, including the following three aspects:



* Capabilities at the microarchitecture level, such as integer computing capability, floating-point computing capability, etc.

* Higher level capabilities, such as memory access latency, memory bandwidth limits, frequency fluctuations, etc.

* Actual operation situation, typical benchmark operation of computationally intensive or memory intensive workload






I wonder if there are sound and authoritative test suites, workloads, or methodologies for these three aspects?