Radeon RX 7900 XT vs GeForce RTX 4090: one sfida gives 100 TFLOPs?

The battle between AMD and NVIDIA as for the next generation video cards it could reserve us a “high altitude” confrontation, about 100 TFLOPs. According to the latest rumors of the leakers Greymon55 and Kopite7kimi, the power of the future top of the range could dwarf the proposals of the current generation.

The current Radeon RX 6900 XT achieves FP32 computing power of around 23 TFLOPs, while the hypothetical Radeon RX 7900 XT GPU-based Navi 31 could get to touch the 92 TFLOPs. AMD’s new solution is expected to offer a maximum of 15360 stream processor thanks to an MCM type design and working at a frequency close to, if not higher than, i 3 GHz.

As for the hypothetical instead GeForce RTX 4090 basata su GPU AD102the leakers do not go so far as to indicate a particular number, but they believe that it will have what it takes to overcome the wall of 100 TFLOPs. For reference, the GeForce RTX 3090 Ti hits 40 TFLOPs.

The leap forward of the AMD and NVIDIA proposals would therefore be respectively equal to 4 times in the case of the transition from 6900 XT to 7900 XT and 2.5 times in that of the RTX 3090 Ti to the RTX 4090.

The FP32 power parameter does not say that it scales linearly in gaming performance, and is only one piece of a decidedly more complex picture that will see solutions collide on many fronts, from ray tracing performance to upscaling technologies up to consumption that really promises to be. high.

To reach the power of 100 TFLOPs or more, NVIDIA’s AD102 GPU would have to operate at over 2.7 GHz. The chip should have 18432 CUDA cores in its full form, but NVIDIA’s top-of-the-line probably makes only partial use of the full GPU potential. It is currently unclear whether the “100 TFLOPs and beyond” hypothesis refers to the fully active GPU or not.

Is NVIDIA testing a 900 Watt GeForce RTX 4000? Or maybe the TITAN RTX is back?