Method and system for GPU (Graphics Processing Unit) server IO (Input/Output) resource distribution

The invention discloses a method for GPU (Graphics Processing Unit) server IO (Input/Output) resource distribution. The method comprises the following steps that: through POST self inspection, scanning PCI (Peripheral Component Interconnect) equipment connected to a GPU server; according to scanned parameter information, judging whether the PCI equipment is a video card of a preset model or not; and if the PCI equipment is the video card of the preset model, stopping distributing IO resources to the PCI equipment. By use of the method, according to the scanned parameter information, whether the connected PCI equipment is the video card of the preset model or not is judged; if the connected PCI equipment is the video card of the preset model, IO resources are not distributed to the PCI equipment, so that a consumption-level video card does not consume the own IO resources of the BIOS (Basic Input / Output System) of the limited GPU server, a situation that the consumption of IO resources is avoided by limiting