I'm looking at using either thunderbolt 3 to PCIE x16 or m.2 to PCIE x16 adapters to connect two extra RTX 3090s for deep learning.
Benchmarking has shown the chipset bandwidth/latency drop doesn't affect the load from deep learning much.
I have the z790 Maximus Hero, which has 3x m.2 pcie x4 slots (1 cpu, 2 chipset), and then a chipset 2xnvme pcie x4 card that supports an x4/x4 mode.
Anyway, if I don't care about the latency/speed drop using chipset x4 lanes for GPUs, will ASUS z790's support GPUs in these (suggested) nvme m.2 slots with m.2 -> pcie x16 adapters?
