the-pi-guy said:
I'm a little confused on what the distinction between the two would be. I'm assuming the distinction is that "ray-tracing acceleration" could be that the ray-tracing could be a mix of software/hardware. Or at the very least, it's more ambiguous than the latter. |
There is a very thin line interpreting this sentence "ray-tracing acceleration in the GPU hardware", whether an engineer tells it or whether a pr speaker tells it. Since it likely comes from an engineering standpoint, it means there are additional features implemented in hardware. The first suspects are additional scratch pad memories that hold intermediate results (and in ray tracing, there are a lot of those required). In its simplest form, we could look at generic raytracing code performed with gpgpu, with additional memory scratchpads. In the most extrem form, the ray tracing code is hard-coded ("Hardware-Level ray tracing") with transistors (lots and lots and lots of them) and memory scratchpads. Both implementations have their advantages and disadvantages (basically big&fast vs flexibility), there is no clear winner.
