NBD-VRAM Enables GPU VRAM as Linux Swap Space for NVIDIA GeForce RTX Laptops
Key Takeaways
- ▸New open-source tool enables NVIDIA GPU VRAM to function as Linux swap space
- ▸Targets laptops with soldered RAM and NVIDIA GeForce RTX GPUs
- ▸Leverages NVIDIA's CUDA API and Linux NBD protocol to expose GPU memory
Summary
An MIT-licensed open-source project called NBD-VRAM has been released, enabling Linux users to leverage NVIDIA GeForce RTX GPU memory as system swap space. This addresses a specific pain point for laptop users with soldered memory who have NVIDIA consumer GPUs available but insufficient system RAM.
The tool works by running a daemon that allocates VRAM through NVIDIA's CUDA driver API and exposes it as a Linux Network Block Device (NBD) on a Unix socket. The GPU memory is then presented as a conventional Linux swap device, allowing the system to offload memory pressure to the graphics processor.
NBD-VRAM is specifically designed for consumer NVIDIA GPUs where the NVIDIA P2P API and similar alternatives don't work. It requires the official NVIDIA Linux graphics driver stack and CUDA support. Published yesterday on GitHub under an MIT license, the project provides a creative solution for users who need additional memory capacity while possessing untapped GPU resources.
- MIT-licensed project designed for consumer GPUs where P2P transfers don't work
- Available on GitHub; requires official NVIDIA drivers and CUDA support
