Yeah and that's the 1D10 variant. The 1D12 variant in T480s is even slower. If it wasn't for the Linux driver headaches I still might have gone for it to have

Yeah and that's the 1D10 variant. The 1D12 variant in T480s is even slower. If it wasn't for the Linux driver headaches I still might have gone for it to have another way to check if CUDA code was running properly without having to ssh to the central server.