How to speed up training

My computer has an i9-14900KF CPU, an RTX 5090 GPU with 32GB of VRAM, 128GB of RAM, and a 2TB SSD. I'm using the kohya_ss v25.2.1 trainer and want to train a Flux.1 model. How should I configure it to utilize my computer's full performance for training? The problem I'm encountering today is that the VRAM usage is around 31.3 MHz, but the GPU utilization is only around 30%, and it's very slow. Does kohya_ss not support the RTX 5090 graphics card?