That makes no sense.CUs are hardware and thus it is hardware accellerated ray tracing.
I can only assume you made this comment by comparison to Nividia's PC GPUs. The basic of an RT core is simple faster data access. More efficient, yes.Absolutely needed, not really.
Tensor cores are basically for faster AI learning.
It will basically boil down to the usual outcome of PC has a bit more.
Except the difference there is that RT cores in Nvidia GPU's are dedicated and were designed from the beginning to running RT Ops. CU's were not and are not. Are they mixing in rasterization work with it? No one knows at this point.
By your thinking, decoding/encoding H264 video on the CPU is hardware video decoding too since the CPU is a piece of hardware? Yea.....no.
Last edited: