I've been trying to get away from using CPAI and large power hungry gpu cards.
Tried the ONNX AI with just my cpu and that's a no go. amd 2200g. Also got an intel i7 7th gen.
I had an old firepro 4100 card, self powered from the slot hanging around in a box.
Thought I'd give it a go, blew off the dust and hey voila getting inference speeds of about 130ms. Not bad for such an old card. A keeper for now.
Had a migraine trying to get it working though on the amd pc as the iommi groups are not split like intel.
Had to split the amd card with the igpu to pass though only the card (proxmox vm). blacklists, kernal arguments, etc etc
Broke it a few times but now running smoothly.
Are some drawbacks.
Would be better off with the firepro on the intel i7 pc but has very few room for disks. Will need some splitters.
Project for another time.
My cheap ryzen pc - boot disk on zfs pool mirrored 2 sata ssds.
Fast storage in the nvme card. and another 500gb for the pbs server.
Tried the ONNX AI with just my cpu and that's a no go. amd 2200g. Also got an intel i7 7th gen.
I had an old firepro 4100 card, self powered from the slot hanging around in a box.
Thought I'd give it a go, blew off the dust and hey voila getting inference speeds of about 130ms. Not bad for such an old card. A keeper for now.
Had a migraine trying to get it working though on the amd pc as the iommi groups are not split like intel.
Had to split the amd card with the igpu to pass though only the card (proxmox vm). blacklists, kernal arguments, etc etc
Broke it a few times but now running smoothly.
Are some drawbacks.
Would be better off with the firepro on the intel i7 pc but has very few room for disks. Will need some splitters.
Project for another time.
My cheap ryzen pc - boot disk on zfs pool mirrored 2 sata ssds.
Fast storage in the nvme card. and another 500gb for the pbs server.