those that have used both, which would you say is faster and by how much
I have not used both, but I’m confident that the 4090 is faster. Probably by about 10% to 20%.
I do not have the RTX 4090, but I do have the RTX 5080. My experience with the RTX 5080, paired with my Ryzen 7 9800X3D, is neither the GPU nor the CPU max out. VRAM for most AI models is only using about 4GB. The bottleneck seems to be the software itself. I would image the same results would be true for the 4090 and the 5090. Until further AI model optimizations are made, maxing out the hardware is not possible… There is one close exception… The Rhea and Rhea XL models task my GPU to 80 to 85%, VRAM to 12GB. CPU is at 40 to 45%. (approx 4.5fps for 480 upscale to 4x)
IFor comparison… Proteus and Iris, the GPU is at 60%, VRAM at 4GB, CPU is at 65%. (approx 20fps for 480 upscale 2X and 4X)
