Virtualization of professional video cards NVIDIA launched on older models

A group of modders managed to implement the function of dividing the resources of GeForce series video cards into virtual copies. A special hack forces proprietary drivers to create an interface that is required to emulate multiple virtual cards on a single physical device. Previously, this feature was only available on professional graphics cards NVIDIA Quadro and Tesla costing several thousand dollars.

The gpu_unlock module, published on GitHub, overrides the value of the current PCI video card ID in the system. Accordingly, GPUs are referred to as Quadro or Tesla cards. Currently, video cards based on graphics cores GP102, GP104, TU102, TU104 and GA102 are supported. The table shows the correspondence of real cards to professional models.

The hack can only be implemented on Linux using KVM (Kernel-based Virtual Machine), a software virtualization solution for Linux. It doesn’t work on other systems and other software. The developers warn that their solution is untested and should be used with caution.

Donald Westbrook

Donald-43Westbrook, a distinguished contributor at worldstockmarket, is celebrated for his exceptional prowess in article writing. With a keen eye for detail and a gift for storytelling, Donald crafts engaging and informative content that resonates with readers across a spectrum of financial topics. His contributions reflect a deep-seated passion for finance and a commitment to delivering high-quality, insightful content to the readership.