Hi there. I was wondering if the number of CUDA cores from the rtx 3090ti can be compared directly with those from the rtx 2080ti in terms of the speed increase on YOLO inference ?
Specifically, would an rtx3090ti with 10496 CUDA cores be approx
10496 / 4352
times faster at yolo CNN inference ?
I ask because I was told by someone that the definition of CUDA cores changed implying that they may not be so directly comparable in terms of CNN inference.
Cheers,
Kim Hendrikse