More recently, rumors about AMD's flagship processor on RDNA3 architecture indicate that it will offer more than 4 times the processing power of RDNA2.NVIDIA may be making a similar decision for its upcoming top-end video processor.The AD102 GPU, based on the Ada Lovelace architecture, is expected to deliver more than 100 TFLOPs of power, 2.5 times more than the 40 TFLOPs delivered by the Geforce RTX 3090 Ti, and 2.8 times more than the Geforce RTX 3090.However, powering single precision FP32 cores does not automatically guarantee better gaming performance.Sources Greymon55, and Kopite7kimi agree that the battle for TFLOPs will only be part of the battle for the fastest graphics for desktops.There are other features in gaming, such as ray tracing acceleration, support for super-resolution technology, and other technologies that could tip the scales in favor of either architecture.To achieve 100 TFLOPs, the AD102 GPU with 18432 CUDA cores must clock in at 2.7 GHz, but almost certainly the Geforce RTX 4090 will function with the GPU partially disabled, so the clock speed will automatically be higher.If we believe Greymon55, the next generation flagship cards may come with very similar clock speeds, which in the case of the AMD Navi 31 GPU will be around 3 GHz, and expected with the use of a full graphics processor.This is not much that is known about the next generation GPUs, although some rumors have been around for months and there is still much to figure out.The number of FP32 CUDA / Stream processors is still subject to change.All the leaks lead to one thing and that is that next generation GPUs will require a lot of power.
0 Comments