A recent leak leaves us with some very interesting information about the Radeon RX 7000, the next generation of graphics cards that AMD will launch, in theory, in the middle of next year. Their base will be the RDNA 3 architecture, and they will be manufactured in 5nm process. It is important to contextualize something in this case, and that is that AMD has been using the 7nm process in its graphics cards for a long time, so a manufacturing node reduction makes a lot of sense.
Information It has been distributed among several sources, but made a compilation for you with the most interesting that we know, so far, of those new Radeon RX 7000s. First of all, we must bear in mind that this could become a generation focused on ray tracingThat is, your performance improvements will prioritize the acceleration of that technology. This is something very important, and I am going to explain why.
When NVIDIA developed the Turing architecture it introduced RT cores and tensor cores. The former specialized in ray tracing acceleration, and the latter in artificial intelligence. Those RT cores are used to accelerate both the ray-triangle intersections like the BVH cross intersections, and the box bounding intersections. With Ampere, he also added the possibility of interpolating the position of the triangle in time, which allows generating ray tracing with motion blur.
This approach is more complex and consumes more space in the package, but offers higher performance and totally relieves shaders of the most important burden that ray tracing represents. The RT cores also have an asynchronous pipeline next to the tensor cores, which allows the concurrence of different simultaneous operations. In contrast, the RDNA 2 architecture uses a shared resources approach that is less effective, since:
- Ray tracing units work with ray-triangle intersections and box anchors, but BVH cross intersections are handled by shaders.
- Those ray tracing units share resources with texturing units, which means that they cannot work simultaneously, as we explained at the time when analyzing the Xbox Series X SoC.
- They lack that asynchronous operation capability which is present in Turing (RTX 20) and Ampere (RTX 30).
What can we expect from the RDNA 3-based Radeon RX 7000s?
A reduction in the manufacturing process thanks to the jump to 5 nm of TSMC, as we have already told you. This should allow higher energy efficiency (performance per watt) and thermal (lower operating temperatures). We can also hope that this new architecture will maintain the division of 64 shaders for each compute unit, but I don’t think we’ll see a major jump in raw performance, at least not as much as the Radeon RX 6000 versus the Radeon RX 5000.
If AMD decides to focus on improving ray tracing performance, it will have to introduce a major redesign at the silicon level that allows you to have everything you need to accelerate, in a more efficient way, the entire workload that this technology represents. With this in mind, I wouldn’t be surprised to see a set of specialized ray-tracing cores built into RDNA 3 GPUs with a closer approach (though not the same) as NVIDIA has been adopting, plus an asynchronous system to fully free up. to the shaders of this workload, and to achieve full specialization.
Regarding its specifications, at the moment everything seems to indicate that the Navi 31/33 GPU will have up to 80 CUs, which would translate into a maximum of 5,120 shaders, the same as the Radeon RX 6900 XT. It is a very curious fact, since it reinforces everything that I have said previously, that the Radeon RX 7000 could bring minor changes in gross power compared to the RX 6000, and that they could become (almost) a “lost generation” to improve its position working with ray tracing. Its arrival would be scheduled for the second half of 2022, and the first to arrive would be the most powerful solutions within the Radeon RX 7000 series, that is, the Radeon RX 7900 XT and RX 7800/7800 XT.