Deploy Aspose.LLM for .NET on NVIDIA GPUs — select CUDA, offload all layers, configure multi-GPU split, and size VRAM for model + KV cache...., 24 GB RTX 4090 + 12 GB RTX 3080), bias the split...GPU t/s RTX 3060 (12 GB) 40-60 RTX 3090 / 4070 60-90 RTX 4080 /...