Can a PC be set up to only run CUDA computing on a Quadro 2000, and run all all other graphics on the integrated graphics device? We are running into an issue where our new Win 10 IoT Enterprise OS is using more GPU dedicated memory than the previous WES7 OS, and our image reconstruction algorithm is no longer able to get enough dedicated memory to run on the Quadro 2000. One solution we are considering is to only run the CUDA based reconstruction on the Quadro 2000, and all other graphics on the IGD. We are supporting several generations of PC’s and NVida Quadro GPU’s in the field, but only the Quadro 2000 with 1GB of video memory is having the issue. Below is some more detail on the two supported configurations.
OS: Windows 10 IoT Enterprise 2019 LTSC
Quadro Driver: 377.83 (to support all GPU’s listed below)
Single Monitor, Single GPU per Computer
CPU: Intel COre i7-2600
MB: Itox MicroATX Intel Q67 Chipset
GPUs (1 per PC): Quadro 2000, K2000, or K2200
Boot: BIOS / MBR
CPU: Intel Core i7-6700
MB: Contec Mini-ITX Intel Q170 Chipset
GPU: P1000 v2
Boot: UEFI / GPT
My questions are:
Is it possible to set up these configurations to do only the CUDA computing on the Nvidia GPUs, and all other graphics (and the display) on the IGD?
If so, what are the basic steps?
Is there a better way to approach reducing the windows 10 GPU footprint?