So, I finally maxed out my system - 4 R9700s and my old Xeon 8368 for a total of 128GB VRAM and 256GB of 8-channel DDR4. However... DS4 Flash barely runs fast now compared to the 3 card version. I'm using llama.cpp and standard fitting with a 200k context, using the bartowski mxfp4 quant. Is there a
