Intel Xe-HPG DG2 (Intels GPU )


Well-Known Member
in older news /rumor they were going to implement RTX with different approach
Underlying approaches of every vendor is completely different. In fact the same hardware schedulers are being used for both mesh shading and RT and a more generalized work (compute) graph scheduling in future.

The only commonality is whether the RT approach is single hop or multihop. NVIDIA used multihop where there are acceleration units for ray triangle intersection and also ray traversal.

Others so far seem to use a single hop approach where triangle intersection is done in hardware but traversal uses the standard execution units(shaders). It has both drawbacks and benefits. The main drawback is that you just can't dispatch a shading call the moment you have a ray triangle intersection if you care about performance which you can in the multi hop approach. The main advantage is smaller silicon size and lower power consumption.