I have been trying to install nvidia drivers and cuda on an embedded platform we have. It has a 32-bit ARM processor, integrated Root Complex, 1 GB System memory etc. and comes with an older Linux kernel(3.9-xx). PCIe x4 is supported through a slot.My interface to the system is through a terminal and there’s no monitor attached to the GPU(GeForce GT640).
lspci etc. commands work and can recognize the device with BARs allocated. But when I try to load the driver, the kernel hangs with no output on the terminal. Only way to get things back up is to power cycle the device. My question is related to whether the system memory (1 GB) too limited for this GPU or if its got to do with kernel version etc.
Any advice is greatly appreciated.