June 2, 2026
Swap Til You Drop
Use your Nvidia GPU's VRAM as swap space on Linux
People are turning spare graphics memory into emergency laptop memory — and reactions are split
TLDR: A Linux project lets people use spare memory on Nvidia graphics cards as emergency overflow space, mainly for laptops that can’t be upgraded. Commenters are split between calling it a clever lifesaver and a gloriously cursed way to use “really expensive” memory.
A new Linux hack is making the internet do a full double take: instead of letting your laptop choke and dump everything onto the slow drive, this tool grabs unused memory from your Nvidia graphics card and uses it as backup space. In plain English, it turns that idle chunk of gaming muscle into an emergency overflow bin for your system. The creator says it’s especially for laptops with memory soldered in place, where upgrading is basically a fantasy. And yes, people immediately had feelings.
The comment section is a perfect little tech soap opera. One camp looked at it and went, absolutely not. The bluntest reaction came fast: “I mean, cool, but I’d rather not?” That pretty much set the mood for the skeptics, who see this as a chaotic stunt involving very expensive parts. Another commenter joked it means swapping from pricey memory to really expensive memory, which is both a roast and, somehow, a compliment. Still, the defenders were ready: if your graphics card has 8GB or even 32GB sitting around doing nothing, why not put it to work instead of hammering your storage drive?
Then the dreamers arrived. One person basically pitched the plot twist sequel: if the data is already sitting in graphics memory, maybe this opens the door to wild future uses like speeding up databases. So the vibe is clear on this project: part clever lifesaver, part mad scientist energy, and 100% the kind of idea that makes the comments more entertaining than the announcement itself.
Key Points
- •nbd-vram uses NVIDIA GPU VRAM as Linux swap space by exposing CUDA-allocated memory through the kernel’s Network Block Device interface.
- •The article reports testing on an RTX 3070 Laptop system with 16 GB RAM and 8 GB VRAM, allocating 7 GB of VRAM for swap.
- •The documented memory overflow order is RAM first, then VRAM-backed swap, then zram, and finally SSD swap.
- •The project avoids NVIDIA’s P2P and direct BAR1 mapping approaches because the article says they do not work on consumer GeForce GPUs in this use case.
- •The tool installs as a systemd-managed service with configurable VRAM size and swap priority, plus optional power-aware management and test scripts.