The biggest weak point of the RX 6000 compared to the RTX 30 is its performance against Ray Tracing. It’s just a matter of enabling it in games and seeing how the differential increases in favor of NVIDIA cards. This led AMD to make changes to the RDNA 3 architecture for ray tracing. Everything therefore indicates that this will be the most advantageous part of all for the next generation of the RX 7000.
One of the weak points that the RDNA architecture had in its early days, consisting of the RX 5000 line, was its lack of units analogous to the NVIDIA RT cores, which are responsible for performing two common tasks in Ray Tracing. The first of these is the calculation of the ray-object intersection, which occurs several billion times per second and consumes a large amount of resources. The second is traversing the data structure that represents the scene. AMD has decided to opt for a mixed solution. Where the intersection is computed through its ray accelerator units, but they don’t compute the data structure. A solution that was ultimately not the most effective.
Changes to RDNA 3 compute units for ray tracing
In its last public conference for investors and shareholders, AMD gave a quick overview of what we can see in the future RX 7000. Some changes were already known to us, such as the fact that certain models in the range have been broken down into several different chips. , as does the Ryzen desktop, and the use of TSMC’s 5nm node. However, that’s not the only change we’ll see and it looks like AMD’s commitment to Ray Tracing in RDNA 3 will be bigger than ever. Well, for what is the hybrid rendering
Let’s not forget that the calculation units are the real heart of the graphics chip because they have all the parts to carry out the different stages of the cycle of each instruction and the fact that AMD is officially announcing that it is going to change its organization is significant . The last time he did was with the move from the RX Vega to the RX 5000 and that was enough for him to start talking about a new architecture. Although the first thing we expected was better ray acceleration unit that it performs its task more efficiently and that it is at least at the level of those that exist in the RTX 30. And yes, the intersection units are found within each Compute Unit.
Double FLOPS per compute unit
The other improvement that is expected is to double the floating point computing capacity, similar to what NVIDIA did in its RTX 30. The way to do this will be place twice as many 32-bit floating point units compared to the previous generation. We don’t know this officially through AMD marketing, but we do know through fairly official information such as its own patents and graphics drivers.
In both cases, we were able to learn that double instructions can be sent to the calculation units. Thus, each SIMD unit within the compute unit and encompassing the various 32-bit floating point unitss will go from 32 elements in RDNA 2 to 64 elements in RDNA 3. Said instructions or threads can be executed as 32 double instructions or 32-bit threads or 64 single instructions or threads.