…Both Intel and AMD optimized their software stacks to get massive speedups in generative AI which has seen AMD’s RTX 7900 XTX get higher performance per dollar than an NVIDIA RTX 4080 in generative AI (specifically Stable Diffusion with A111/Xformers). Considering Stable Diffusion accounts for the vast majority of non-SaaS, localized generative AI right now - this is a major milestone and finally offers some competition to NVIDIA.
Even more reason for GPUs for PC gaming to continue increasing in price and scarcity. Awesome!
You love to see it. AMD has been notorious for being bad at AI, hopefully with improvements they can be a real competitor to Nvidia.
I just want better support from developers. I’m fine with AMD being slower, but a ton of projects just don’t even run with AMD hardware. I wish there was a psuedo-vulkan style interface for AI, so that GPU makers and developers could target that in the same way game developers target vulkan/opengl.
Blame the dominance of cuda in college cs.
I haaaaaaaate when things say 900% instead of 9x. It’s so transparent that they’re trying to make it seem bigger than it is.
Just be honest. 9x is great. Call it what it is.
Technically I believe 900% increase is 10 times (compare that 100% increase is doubling = 2x). To me this makes these hundreds of percents reporting even more annoying.
And then here I am, still unable to run StableDiffusion on my 6650xt because pytorch apparently hates it.
TBF llama.cpp runs like a wonder on ROCm.
All aside the improvements in the AI ecosystem by AMD are appreciated and overall tangible. Let’s hope they’ll keep it up.
RTX 7900 XTX
Isn’t this whole article comparing apples with oranges? You can also use Microsoft Olive for Nvidia GPUs. But it isn’t used in this comparison.
Nvidia at least talked about a 2x increase in performance when using Microsoft Olive on their GPUs. So that would leave the 7900 XTX in the dust again.
Tech news should be objective, doesn’t matter what team you root for.
There’s a plugin for Blender I saw a while ago that uses Stable Diffusion to create dynamic textures for objects, the example that was shown was a low-poly AC unit. The GPU generate option was greyed out for AMD GPUs because the Nvidia option uses CUDA, I wonder if this update would change that? Because that’s something I really wanna be able to use for prototyping
Do Olive models work with LoRA yet?
