Why is the Memory Usage in HARDWARE DASHBOARD so constant?

Hi,

Can anyone tell me why is the Memory Usage in HARDWARE DASHBOARD so constant? It always keeps 6 - 7 GiB each pod no matter there is an operation or not. The memory limit of each pod is 16GiB. When I use kubectl top pods command, I can see the MEMORY(bytes) is actually around 14000Mi. Why they are so different. Which one accurately measures the memory usage of each pod?

Thanks,
Derek