ERROR: cuMemAllocHost() 2

[rico]

I guess you could benchmark one or two at a time instead of them all at once and observe memory usage:

Device #1: GeForce GTX TITAN X, 12288MB, 1076Mhz, 24MCU
Device #2: GeForce GTX TITAN X, 12288MB, 1076Mhz, 24MCU
Device #3: GeForce GTX TITAN X, 12288MB, 1076Mhz, 24MCU
Device #4: GeForce GTX TITAN X, 12288MB, 1076Mhz, 24MCU

So the command below will benchmark Device #1 & #2:

cudaHashcat64.exe -b --gpu-devices 1,2

Might help narrow it down.

EDIT: good idea to monitor power usage too.