c0deJedi

c0deJedi / nbd-vram

Public

Use your NVIDIA GPU's VRAM as swap space on Linux. Built for laptops with soldered memory and no upgrade path. If you have an RTX card sitting there with 8GB of VRAM and you're getting swapped to SSD, this puts that VRAM to work

www.seanlobjoit.com cuda gpu laptop linux memory
15
0
85% credibility
Found Jun 01, 2026 at 15 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
C
AI Summary

nbd-vram is a tool that repurposes unused NVIDIA GPU memory as extra swap space on Linux laptops. It creates a small background program that allocates VRAM via the CUDA driver, then exposes it as a swap device that the kernel can use like normal memory. This gives laptops with limited soldered RAM access to additional fast memory through their graphics card, improving performance before hitting slower SSD storage. The project works without requiring any special kernel modules or NVIDIA kernel symbols, making it stable across driver and kernel updates.

How It Works

1
💡 Discover your GPU has untapped memory

You learn that your laptop's NVIDIA graphics card has memory sitting idle while your computer struggles with limited RAM and slow SSD swap.

2
📦 Install the swap system

You download and run the installer, which automatically sets up the VRAM swap service to start automatically every time you turn on your computer.

3
⚙️ Configure how much memory to use

You decide how much of your GPU's 8GB of memory to repurpose as fast swap space, setting it up so your computer uses the fast VRAM before slower storage.

4
🚀 Watch your memory triple overnight

Your laptop now has access to your normal RAM, plus your GPU's memory as fast swap, plus compressed memory and SSD storage as backup layers -- tripling what you can use.

5
Choose your testing path
🧪
Run the smoke test first

You do a quick 1MB write-and-read check to confirm the VRAM swap is working correctly before relying on it.

⚡
Start using your computer normally

You trust the setup and jump straight into using your apps, with the expanded memory working silently in the background.

🎉 Your laptop now has triple the memory

Everything is running smoothly with your GPU's fast memory handling overflow from your RAM, making your computer feel more responsive.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 15 to 15 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is nbd-vram?

nbd-vram turns your NVIDIA GPU's VRAM into swap space on Linux. Built in C, it runs a daemon that allocates GPU memory via CUDA and serves it as a block device using the NBD protocol. The kernel connects to it over a Unix socket, and from there it's treated like any other swap partition. The project targets laptops with soldered RAM and no upgrade path -- if you have an 8GB RTX card sitting idle while your system hits SSD swap, this puts that VRAM to work.

Why is it gaining traction?

The clever part is how it sidesteps the NVIDIA P2P API, which is blocked on consumer GeForce cards. Instead, it uses standard cuMemcpy calls that work on any CUDA GPU. No kernel module means no fighting driver updates -- the setup survives across kernel and NVIDIA driver upgrades without recompilation. The performance is surprisingly usable at around 1.3 GB/s sequential throughput, faster than NVMe since data travels over PCIe to the GPU rather than to storage. The priority system lets you layer VRAM swap above zram and SSD swap, so overflow hits the fastest tier first.

Who should use this?

Linux laptop users with RTX/GTX cards who are stuck with 8GB or 16GB of soldered RAM and keep running into OOM. Developers running memory-heavy workloads locally -- compilers, Docker, VMs -- who want more headroom without opening their machine. Anyone willing to trade a chunk of GPU VRAM for tripled addressable memory on a budget laptop.

Verdict

This is a genuinely clever hack with a solid architectural approach, but the 15 stars and single-user authorship mean it's early-stage and lightly tested. The documentation is thorough and the no-kernel-module design is smart, but you'd be wise to run the smoke test before trusting it in production. The 0.85% credibility score reflects that maturity gap -- promising and well-documented, but not yet battle-hardened by a community. Try it on a non-critical machine first.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.