I assume we are talking about real time ray tracing or hardware accelerated ray tracing because ray tracing already exists on A series chips with Metal.
"Ray Tracing" has two main parts: building a structure, generally called a BVH, from the geometry, that allows for rapid ray intersection tests; and then performing tests of the rays against the structure.
"Accelerating ray tracing" or "HW ray tracing" is a vague term that can mean anything from
- using the CPU to construct the BVH and existing GPU HW to perform the intersection testing TO
- providing GPU HW to construct the BVH and augmenting GPU HW to allow performing multiple intersection testing in one GPU "core cycle" to
- providing even better GPU HW that can make minor changes to the BVH based on minor changes to the geometry from to frame (so that we don't have to build the full BVH every time geometry changes).
Right now Apple is (as far as we know) at the first of these levels. If you look at the Metal Feature Sets PDF that's essentially what's meant by what they offer ("Ray tracing in render pipelines").
BUT it's worth noting that the Metal Feature Set tables only go up to A15/M2... So who knows what was planned (and what is available) in A16?
There's also the larger point that you have to remember that
- the journalists who report these things are not tech experts AND
- the engineers they spoke to are not making formal reports.
We did not have the engineering team give a presentation to some journalist about "Postmortem on what went wrong with the A16". Instead we have someone talking to someone else in vague terms and, quite possibly, half way through realizing "OMG I've said too much" and starting to redirect the conversation to cover their tracks and obscure the issue.
It is probably true that the A16 GPU is less than what was hoped. But why?
- Perhaps it was designed for N3, then investigated for N4, considered not feasible, so we revery to A15 GPU.
- Perhaps it was designed assuming some other aspect of the chip design (SLC, NoC, Coherence protocol) would be in place but they were delayed so the GPU was delayed.
I find the story as presented ("GPU was badly designed, runs far hotter than expected [implied because of ray tracing], this was only discovered late in design") to be very unlikely. More likely I think is that different facts have been put together in an attempt to create a story; the facts are kinda true, the story not at all.
Probably there is HW ray tracing in the original A16 GPU.
Probably there are multiple other features to make it faster.
Both those are obvious. But why the delay? As I said, unlikely to be the fault of the GPU per se, most likely to be because the GPU targeted a process that was not feasible at the time a shipping decision had to be made.