okuha-logo-small-white

Stable Diffusion Requirements – Hardware & Software

stable-diffusion-hardware-requirements-feature-image

If you click on a link and make a purchase, I may receive a small commission. As an Amazon affiliate partner, I may earn from qualifying purchases.
Read our disclosure.

My Experience

“You can run Stable Diffusion on your laptop, but don’t expect it to be fast and generate large-resolution images. Getting a proper NVIDIA RTX GPU with 10GB of VRAM makes the image generation process a lot faster and more enjoyable. Reserve storage space (preferably SSD) for new models you might like to download from Civitai or from other sources”.

  • VRAM is the most important component for running Stable Diffusion and diffusion models in general.
  • 8GB of VRAM is the minimum to make image generation enjoyable.
  • NVIDIA GPUs are preferred over AMD.
  • There are WebUI like Sygil that are optimized for low VRAM and bigger generations.

Minimum System Requirements for Stable Diffusion

  • CPU: Any AMD or Intel processor.
  • RAM: At least 16GB, preferably the latest DDR memory available.
  • GPU: Any GeForce RTX GPU that has at least 8GB memory.
  • Storage: Preferably any SSD drive with at least 200GB of storage space.
  • OS: Mac, Windows, or Linux

Stable Diffusion Hardware Requirements

Stable Diffusion (SD) is hard on VRAM; thus, you should focus on that when building a computer for SD. Using PCIe to process and transfer data between CPU and RAM (utterly simplified) is very slow compared to using data inside VRAM, so that’s the reason why SD uses VRAM instead of RAM.

GPU and VRAM

Stable Diffusion can be run on older NVIDIA and AMD graphics cards, but they are more prone to problems. Any NVIDIA RTX GPU card is a safe bet for running Stable Diffusion without a hitch. The most important thing is to have a lot of VRAM in the card.

While 8GB is a minimum requirement (for adequate results), it’s preferred that you get a graphics card that has at least 10GB of VRAM.

Stability.ai (creator of the Stable Diffusion base model) recommends NVIDIA chipset rather than AMD (AMD graphic cards can also run diffusion models as pointed out in this Reddit thread). Stable Diffusion requires 8GB of VRAM to run properly. However, the more VRAM your graphics card has, the better.

Bestseller No. 1
MSI Gaming GeForce RTX 3060 12GB 15 Gbps...
  • NVIDIA GeForce RTX 3060 12GB GDDR6...
  • 1710 MHz GPU clock speed and 1807 MHz...
  • DisplayPort x 3 (v1.4a) and HDMI 2.1 x 1...
  • 7680 x 4320 maximum display resolution
Bestseller No. 2
ASUS Dual NVIDIA GeForce RTX 3050 6GB OC...
  • NVIDIA Ampere Streaming Multiprocessors:...
  • 2nd Generation RT Cores: Experience 2X...
  • 3rd Generation Tensor Cores: Get up to...
  • Axial-tech fan design features a smaller...
  • A 2-slot Design maximizes compatibility...

If you have 4-6GB of VRAM available, it is possible to reduce VRAM usage with some modifications:

  • Edit webui-user.bat (after installing Stable Diffusion locally)
    • After COMMANDLINE_ARGS= (enter your desired parameters):
    • Example: COMMANDLINE_ARGS=--medvram (check the image below for correct formatting).
    • If you have 4GB-6GB VRAM, use --medvram.
    • If you have 2GB VRAM, use --lowvram.
    • You can also try --xformers, as many users have seen great results with that, some reporting 4x faster results than with --medvram.
  • If you are getting ‘Out of memory’ errors on either of these, add --always-batch-cond-uncond to the other arguments.
stable-diffusion-commandline-parameters-example
Image showing the modified COMMANDLINE_ARGS.

If you get a green or black screen instead of generated images, you might have a card that doesn’t support half-precision floating point numbers (a known problem on 16xx cards), use --precision full --no-half in addition to other flags (this enables the model to take more space in VRAM).

If you are using a .vae file, you can try out --no-half-vae parameter. You can also disable hardware acceleration in your browser and close any other app or task that is occupying VRAM.

RAM

You should have at least 16GB of RAM (Random Access Memory) installed on your computer. The more RAM you have, the more tasks and apps (Chrome browser, Adobe Photoshop, OBS, etc. apps) you can operate at the same time.

CPU

CPU is not as important as GPU. Any modern AMD or Intel CPU is sufficient to run Stable Diffusion as its main usage is VRAM and not CPU or RAM.

Storage

While the base model and other software needed to run Stable Diffusion is around 10GB, you want to get at least 200GB of storage space (preferably SSD) to your computer.

Stable Diffusion is only one model that you can download, but to make the most out of SD, you want to download different AI art models found from sites like Civitai and Hugging Face.

The latent diffusion models you can find from Civitai vary between 2GB-7GB in size, and downloading 10 of them already puts a weight in the storage space.

Already a simple Windows install needs roughly 64GB of storage space, and you also have to account for the images you generate and how much space they require.

Generating 500 images per day is nothing unheard of, and when upscaled, they will start to take space from your computer.

Stable Diffusion Software Requirements

The Stable Diffusion (SD) base model file is only one part of running SD locally on your computer. You also need Git, Python, WebUI, and preferably other models besides SD V1.5 or V2.1.

To run Stable Diffusion, you need the following files:

Recommended tutorial on downloading and installing needed files: Stable Diffusion 1.5 – Windows Installation Guide – tutorial covers how to download SD, install Python and Git, and how to configure the webui-user.bat file, etc.

Operating system requirements

You can run Stable Diffusion on Mac, Windows, and Linux. To generate images with Stable Diffusion, you need a WebUI (browser user interface) and models (Stable Diffusion base model, checkpoint models, LoRAs, etc.).

Stable Diffusion installation guides for different operating systems:

WebUI

stable-diffusion-user-interface
Image showing how the AUTOMATIC1111’s WebUI looks like.

The user interface I suggest you get for Stable Diffusion is AUTOMATIC1111’s Stable Diffusion WebUI. Currently, it’s the most popular WebUI to run SD.

Another popular Stable Diffusion WebUI is from Sygil-Dev. Sygil UI follows what many commercial AI art generators use and is optimized for VRAM usage and bigger generations.

Git

To run Stable Diffusion on Windows, you need to install Git for Windows. Git for Windows is a software package that brings Git, a version control system, to Windows. It includes Git BASH, which emulates the UNIX command line, and Git GUI, a graphical interface for Git commands. It also provides shell integration for easy access.

Python

To run Stable Diffusion on Windows, you need to install Python, either the latest version or 3.10.6, which worked for me.

Python is a powerful and user-friendly object-oriented programming language. Python is ideal for prototyping and ad-hoc programming tasks. It has an extensive standard library supporting everyday programming tasks like web server connections and file manipulation.

Python runs on various platforms and is free to download and use. It supports basic data types, object-oriented programming, modular code organization, and exception handling.

Models

realistic-vision-v2-stable-diffusion-model
Realistic Vision V2 checkpoint model from Civitai.

As the “last” thing for generating images with the WebUI, you need models. Stable Diffusion models mostly refer to the official models released by StabilityAI but also to custom user-generated checkpoint models.

Models can be found in the old file format .ckpt (checkpoint) or the new file format .safetensors.

Check out: The Best Stable Diffusion Models.

Models are the thing that makes SD shine. You’ve probably seen what Midjourney can do, but in all honesty, SD can generate as beautiful images as Midjourney, but it all comes down to what AI art models you use to generate the images.

Guide: How to use models.

Feature image credits.

Search
artist-profile-picture-avatar

Okuha

Digital Artist

I’m a digital artist who is passionate about anime and manga art. My true artist journey pretty much started with CTRL+Z. When I experienced that and the limitless color choices and the number of tools I could use with art software, I was sold. Drawing digital anime art is the thing that makes me happy among eating cheeseburgers in between veggie meals.

More Posts

Thank You!

Thank you for visiting the page! If you want to build your next creative product business, I suggest you check out Kittl!