When it was confirmed that the PS5 Pro would have a GPU with 3,840 shaders We had the data we needed to estimate his brute strength. Sony dropped the fact that this console uses a new generation of RDNA graphics architecture, which means that we are talking about RDNA 4, successor to the current RDNA 3.

RDNA 3 uses Wave32 dual emission shadersmeaning it is theoretically capable of doubling the maximum power output of the FP32. Therefore, to calculate the GPU performance of this architecture, the final result expressed in teraFLOPS must be multiplied by two. PS5 Pro GPU performance was 16.7 TFLOPs in FP32, so adding dual emission shaders increased that number 33.4 TFLOPs.

I’ve already talked to you about this in a couple of articles, and the 16.7 TFLOPs were exactly the numbers I left you with when I estimated performance during the PS5 Pro presentation based on its estimated top speed 2.35 GHza frequency we still don’t know if it’s scalable or fixed, although it’s most likely scalable, as was the case with the PS5 GPU.

Eventually It’s a value Sony has finally confirmed in the recently leaked official spec list of this console, it contains other interesting data that I also gave you at the time, such as the inclusion of 2 GB of additional DDR5 to free up GDDR6 memory. So we have half the raw performance we would find in an equivalent GPU in shaders like the Radeon RX 7800 XT, why did this happen?

Why the PS5 Pro only has 16.7 TFLOPs

The fact that Sony gave this value raises a very important question for me, especially considering that we are talking about a GPU that is theoretically based on the RDNA 4 architecture.what happened to dual emission shaderswere they removed in said architecture? It’s clear that something important happened because the raw performance data doesn’t match what we should be doing with a traditional RDNA 3 design-based calculation.

The PS5 Pro GPU may use a design with a deeper degree of customization, or it may simply be AMD changed shader settings in RDNA 4 and that the raw power calculation in FP32 is again done in the traditional way with the architecture mentioned. Sony may also have wanted to provide a more realistic vision by removing dual-emission shaders from the equation.

To give you an idea of ​​how important this topic is, The Radeon RX 7600 already has a performance of 21.75 TFLOPs at FP32 thanks to dual-emission shaders, while the PS5 Pro GPU only achieves 16.7 TFLOPs. However, the former only has 2,048 shaders, while the GPU of Sony’s new console has 3,840 shaders.

Does this mean Radeon RX 7600 is more powerful than PS5 Pro GPU in games? Honestly, I don’t think soIt should also be considered that this new GPU brings important improvements at the architectural level and that it has support for technologies that will make an important difference in games, such as PSSR and improved ray tracing cores.

Game performance does not only depend on TFLOPs, but in the end it is a figure that allows us to make a rough estimate of GPU performance, and in this case the 16.7 TFLOPs fits much better. 45% improvement in rasterization performance compared to the PS5 that Sony was giving at the time.