Analysis updated 2026-05-18
Create extremely fast, temporary scratch storage backed by GPU memory.
Compress files directly on the GPU using the optional nvCOMP integration.
Compute hash values for files stored on the VRAM disk using GPU parallelism.
Experiment with using GPU memory as general-purpose file storage.
| activetk/vramdisk | rightnow-ai/auto | yarlabs/hyperspace-db | |
|---|---|---|---|
| Stars | 116 | 116 | 113 |
| Language | Rust | Rust | Rust |
| Setup difficulty | hard | hard | hard |
| Complexity | 4/5 | 4/5 | 5/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Windows, a compatible NVIDIA GPU, and installing the WinFsp filesystem driver before mounting a drive.
VRAMDISK is a Windows application that lets you turn a portion of your graphics card's video memory, known as VRAM, into a mountable disk drive that appears alongside your normal drives. Instead of storing files on your hard drive or SSD, they are held in an object-storage-style layout inside the GPU's own memory, which means the GPU's fast parallel processing can be used directly on the files stored there, for example to compress them, calculate hash values, or encode them. Everything stored on the drive disappears once you unmount it or close the application, since it never leaves the GPU's memory. To use it, your computer needs Windows 10 or 11 and an NVIDIA graphics card newer than the Maxwell generation, with CUDA 12.8 support. Setup involves installing a filesystem driver called WinFsp, downloading the prebuilt release archive, and then running the included program, where you pick which GPU to use, choose a drive letter or folder to mount it to, set a size, and click mount. An optional extra step lets you install NVIDIA's nvCOMP runtime, which unlocks a file compression feature that runs directly on the GPU once the drive is mounted. The README also lists fixes for two common Windows errors people may hit during setup, involving missing Visual C++ or WebView2 runtime components. Building the project from source requires a Visual Studio developer environment, and the app itself is built with Tauri, combining a Rust backend with a web-based interface, installed and built through a couple of PowerShell scripts. The author points anyone wanting deeper technical details on the internal design to a separate development notes file, suggesting it is meant to be read by an AI assistant rather than a person. The project is released under the MIT license, which allows free use and modification, including for commercial purposes. This suits Windows users with a compatible NVIDIA GPU who want extremely fast, temporary storage for tasks like scratch space during heavy computation, or who are curious about using GPU memory and parallel processing for everyday file operations like compression and hashing.
A Windows app that mounts part of your GPU's video memory as a disk drive, letting you use GPU speed for compression and hashing on those files.
Mainly Rust. The stack also includes Rust, Tauri, CUDA.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated hard, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.