Closest would be prompt processing from https://www.phoronix.com/review/nvidia-rtx5090-llama-cpp/2 https://www.phoronix.com/review/nvidia-rtx5090-llama-cpp/2, a

Closest would be prompt processing from https://www.phoronix.com/review/nvidia-rtx5090-llama-cpp/2 https://www.phoronix.com/review/nvidia-rtx5090-llama-cpp/2, and it's barely +20%