![](/cs149/fall21content/media/progmodels/images/slide_035.jpg)
Back to Lecture Thumbnails
![](/cs149/fall21content/media/users/matteosantamaria/profile.jpg)
matteosantamaria
![](/cs149/fall21content/media/users/bryu/profile.jpg)
bryu
@matteosantamaria the V100 is from several years ago, and the newer A100 has over 1500 GB/s bandwidth. Things just get faster and faster...
![](/cs149/fall21content/media/users/tcr/profile.jpg)
tcr
Appreciated how clear Kayvon made the distinction between latency and bandwidth, and especially the implications of each (i.e., when low latency matters and when low bandwidth matters).
![](/cs149/fall21content/media/users/alexder/profile.jpg)
alexder
I'm curious to know how much these distances of the memories to the processor truly make a difference in memory throughput. I'm not an electrical engineer, but I am curious how engineers go about designing the layout of such systems. Is it working within a set of physical constraints or just choosing from a subset of combinatorial possibilites
Please log in to leave a comment.
Copyright 2021 Stanford University
900 Gb/s is an mind-melting figure. I never would have imagined that our hardware is capable of reaching such a number.