I saw in the 2020-09-01 Official Launch Event for the NVIDIA GeForce RTX 30 Series that the RTX IO in the Nvidia GeForce RTX 30 Series allows the GPU to directly access the storage (SSD) without having to go through the CPU, as shown in the flowcharts below.
In the Nvidia GeForce RTX 20 Series and before, the information from the storage goes through the CPU before reaching the GPU (flowchart from the Official Launch Event linked above):
In the Nvidia GeForce RTX 30 Series, the information goes straight from the storage to the GPU:
How ever I see on the same slide "20x lower CPU utilization":
Why does the RTX IO in the Nvidia GeForce RTX 30 Series use any CPU at all when information flow from the storage to the GPU, since according to the flowchart the information goes straight from the storage to the GPU?


