Quote from: davidm on Yesterday at 14:32:36Fast RAM benefits all those scenarios. In an ideal world, RAM would be at the same speed as the CPU.
Yes, but the Deep Learning AI inferencing I use benefits from the following in order:
0) Idle background activity,
1) enough RAM (up to 64 GB) and VRAM (up to 3 GB),
2) choice of Nvidia library versions so that they work well together,
3) tuning of the major AI parameters,
4) speed of all GPU cores together,
5) single thread CPU speed,
6) number of CPU cores,
7) multi thread CPU speed,
8 ) RAM speed (because the AI objects are small),
9) RAM bandwidth.