Stable Diffusion Webui Forge vs Automatic1111 Speed Test

In this video, we're putting Stable Diffusion Webui Forge head-to-head with Automatic1111 in a simple speed test. Webui Forge claims to increase generations speeds for an 8GB VRAM NVidia GPU to be 40-45%. Even more for a 6GB VRAM GPU at 60-75%. Subscribe to see the results and stay up to date with all the latest in AI tech.

Specs used in the video Ryzen 5800x, 32GB system RAM, NVidia 3060Ti 8GB VRAM.

How to install Forge via @OlivioSarikas

Download the easy installer

Tiled VAE Extension for A1111

⏲Time Stamps
0:00 Webui Forge on a 3060Ti Preview
0:50 What to expect for 6-8GB GPU's
1:37 Who developed this?
1:57 Specs & Settings used A1111 vs Forge
2:50 Comparison 30 steps
3:31 Comparison 40 steps
3:55 Comparison 50 steps
4:26 Why is Realistic vision so horny?
4:44 SD1.5 results 632x840, 30 steps

**Disclaimer Affiliate Links Below**

📸 Gear I use

🔦 Find us on:

[MonzonMedia]

Have you tried Forge yet? Let me know your system specs, GPU and times you are getting.

[EskaronVokonen]

I actually like the fact the UI is similar to A1111. Makes it very easy for the users to seamlessly switch to it, and made the development much faster as the developer could focus on improving the image generation engine, and keep compatibility.

[vVinchi]

This is exactly what I wanted to know about Forge, you are a mindreader! Seems worth installing, I'm waiting for it to arrive on Pinokio tho.

[bahethelmy]

Thanks! I am also using Forge and I like it. Nvidia RTX3050 8GB.

[RajeshJustaguy]

i'm just going to try now and will get back to you

[sergiykravchuk]

Recommend a platform (Midjourney, A1111, Fooocus e.t.c.), and which model is best to use to generate realistic photos of people with all body parts, cars, and buildings?

[AndyHTu]

Instead of getting someone who hire these guys, we s hould be helping to donate to his cause and let him live that financial freedom ;)

[rememberme]

Great video! Unfortunately I think for a 4090 theres not a big difference, I'll have to stay on Automatic1111

[hmm-m]

what improvement can it give on my GTX 1080 8GB? normally I use Fooocus, can I also speed it up?

[rudesforges]

Bro, help me with the choice. To generate SDXL images in webui, which card is best to take?
3060 12gb / 4060 8gb/ 4060 ti/ 3070 8 gb?

[RajeshJustaguy]

I just started using Stable Diffusion, and using rtx2060, so I seem to have some issues using SDXL, would Forge solve the speed problem?

[stevehaney]

I tried it on my rtx2060 6 gig vram and saw no noticeable improvement. Certainly not the 60-75% they claim for a 6 gig card.

[irangel]

LOL, "why is this model so damn horny", I'll be thinking that now every time one of my models goes off the rails hahaha, thank you for the comparison. Forge has been awesome for me and my 6gig machine. Don't even have time to make tea between file generations. Guess some sacrifices must be made. BTW I'm running an Nvidia RTX A2000, only fit for a small form factor machine.

[pennybillson]

Models render about 0.5 to 1% faster on my 1060 6GB. SDXL is still far far too slow. Surprisingly SDXL renders are about 15 times faster in Fooocus.

[DavidAsche]

Why would you want a different interface? People know a1111 why learn a new interface when you can just improve what we know.

[andrejsshewchenko]

It's faster on rtx 3080 in my case, specially xl models..

[BabylonBaller]

Tiled VAE for txt2img? never thought of using that. cool. Only use it for upscaling at img2img