Amuse 2.0 beta released for easy on-device AI image generation on modern AMD hardware

[Makaveli]

Hmm another new thing to play with Ai on Radeon!!!

Requirements

• AMD Ryzen™ AI 300 series processors with 24GB of RAM or higher
• AMD Ryzen™ 8040 series processors (AMD XDNA™ Super Resolution requires the latest OEM MCDM and NPU driver update) with 32GB of RAM

RECOMMENDED SPECIFICATIONS FOR AMD GRAPHICS CARDS:​

• Radeon™ RX 7000 Series Graphics Cards

On reddit they say it works with 6000 series just 7000 is recommended.

Amuse 2.0 beta released for easy on-device AI image generation on modern AMD hardware

Ryzen AI 300-series, Ryzen 8040, or Radeon RX 7000 required, for now.

www.tomshardware.com

Introducing Amuse 2.0 Beta with AMD XDNA™ Super Resolution: a fully local, AI experience

AI image generation is one of the most popular use-cases for generative AI and a quintessential AI PC experience is simply incomplete without it. Most of this local image generation happens using Stable Diffusion models created by Stability AI – which is a pioneer in this field. Multiple...

community.amd.com

Amuse

Stable Diffusion Image and Video Generation

www.amuse-ai.com

 

[Makaveli]

New version out.

Amuse 2.1 Beta Is Now Available: Introducing FLUX.1 Support, Bug Fixes and More

Figure 1 AI image set showcasing FLUX.1 on an AMD Radeon™ RX 7900 XTX graphics. Amuse 2.1 Beta is now available for Ryzen™ AI and Radeon™ GPU users and can be downloaded from TensorStack’s website. The new Amuse 2.1 Beta brings support for the newly launched FLUX.1 Schnell model on applicable...

community.amd.com

Amuse

Stable Diffusion Image and Video Generation

www.amuse-ai.com

 

[Makaveli]

New update with

Stable Diffusion 3.5 Support​

Introducing Amuse 2.2 Beta: With Stable Diffusion 3.5 Support and AMD Ryzen™ AI Image Quality Update

Figure 1 Prompt: A prince stands on the edge of a mountain where "Stable Diffusion" is written in gold typography in the sky. A step-by-step guide on how to run Stable Diffusion 3.5 Medium, Stable Diffusion 3.5 Large and Stable Diffusion 3.5 Turbo is available here. Amuse 2.2 Beta is now avai...

community.amd.com

 

[igor_kavinski]

From the support matrix: AMD Ryzen™ AI PRO 300 Series processor (64 GB)

Strix Halo???

By the way, does the Maximum mode refuse to work with a 7900 XTX? Can anyone confirm?

 

[Makaveli]

From the support matrix: AMD Ryzen™ AI PRO 300 Series processor (64 GB)

Strix Halo???

By the way, does the Maximum mode refuse to work with a 7900 XTX? Can anyone confirm?

I'm going to assume the 7900XTX doesn't have enough VRAM for Large & Large Turbo but is fine at Medium.

I can test in abit.



*Update*
So when I load the Large model I get this warning.



You can force it to use Maximum but that wants 38GB
I got it running but its dirt slow due to lack of VRAM.

The Tiger above was generated in 63 seconds when using Medium.
The Lion below was generated in 506 seconds using Large.

 

[Shmee]

Those specs are confusing to me. Are they saying there is an RX 7900XTX with 32GB of RAM instead of 24GB, or are they talking about recommendations for system memory? In any case, looks like this program uses a lot of memory.

 

[Makaveli]

Those specs are confusing to me. Are they saying there is an RX 7900XTX with 32GB of RAM instead of 24GB, or are they talking about recommendations for system memory? In any case, looks like this program uses a lot of memory.

The 32GB is referring to system ram.

The W7800 does have 32GB of vram but that isn't a reference to that.

When it runs out of VRAM it then spills into system memory, when I was loading the Large model all my VRAM and RAM was used.

One of the reasons i'm going to 64GB of ram in my AM5 upgrade.

 

[igor_kavinski]

Wow. The lion does look more realistic!

 

[WelshBloke]

This involves a lot of downloading!

It seems to work on a 6950xt. or at least the fast setting did, as soon as I put it on performance it started downloading another 11.2gigs of data so it'll be awhile until I see if that works!

 

[Makaveli]

This involves a lot of downloading!

It seems to work on a 6950xt. or at least the fast setting did, as soon as I put it on performance it started downloading another 11.2gigs of data so it'll be awhile until I see if that works!

yup the models are all very large so fast internet will be important.

Both the Diffusion large and medium were 26GB's.

 

[Seba]

It works with RX 7600, but it is taking a lot of time.



Can't do anything else on PC during image generation, because of VRAM usage:




Generated image: