AMD's 7900 XTX achieves better value for Stable Diffusion than Nvidia RTX 4080

&gt; Using Microsoft Olive and DirectML instead of the PyTorch pathway results in the AMD 7900 XTX going form a measly 1.87 iterations per second to 18.59 iterations per second!<p>So the headline should be Microsoft Olive vs. PyTorch and not AMD vs. Nvidia.

Been watching this quite closely. As far as I summarise, the 7900XTX is the first (and only) desktop GPU from AMD that _might_ be worth buying. (They own the console gaming space, but thats a different story).<p>Not Nvidia beating due to the CUDA issue, but a massive leap in the right direction.<p>Intel is also making _some_ progress with its ARC range.<p>Its going to be happy days for us users if&#x2F;when AMD&#x2F;Intel are competitive, and cut some of that monopoly margin off Nvidias pricing, but a way to go yet.

Well the problem is that Automatic1111 is not fast...<p>Other diffusers based UIs with PyTorch Triton will net you 40%+ performance.<p>Facebook AITemplate inference in VoltaML will be at least twice as fast as A1111 on a 3080, with support for LORAs, controlnet and such. This supports AMD Instinct cards too.<p>What I am getting at is that people dont really care about A1111 performance on a 3080 because, for the most part, its fast enough.

The comments point out that AMD in the table performing well required the use of Microsoft Olive, and someone in the article comments implies that if you use Microsoft Olive with Nvidia instead of Pytorch with Nvidia, then you&#x27;ll see the Nvidia jump in performance as well, largely rendering the supposed leap by AMD not relevant. Is that true? Can folks chime in?

Nearly bought one thinking AMD will sort itself out shortly but hard to justify vs a second hand 3090 with 24gb and no cuda hassles

Can I also interpret this as: &#x27;AMD&#x27;s pytorch support is so abysmal that inference is 10x slower than it should be&#x27;?

I&#x27;ve been running pytorch and rocm (5.6 has support for gfx1100 if you compile it yourself) for at least 3 months at 18 it&#x2F;s on a 7900 XTX. This has been possible for quite a while.<p>Could someone fill me in on what&#x27;s actually new here, other than the specific technology used?

Wait, why are they comparing Microsoft Olive on AMD to Pytorch on Nvidia? Nvidia supposedly shipped support for Olive recently, there should be no problem getting a head-to-head comparison: <a href="https:&#x2F;&#x2F;www.tomshardware.com&#x2F;news&#x2F;nvidia-geforce-driver-promises-doubled-stable-diffusion-performance" rel="nofollow noreferrer">https:&#x2F;&#x2F;www.tomshardware.com&#x2F;news&#x2F;nvidia-geforce-driver-prom...</a><p>This is a very strange comparison.

I&#x27;m actually curious if libraries like pytorch are even trying to move away from CUDA and if moving away from it is worth it. I get why newer ML toolchains would do that but do mainstream established MI frameeorks plan on sticking with nvidia exclusivity for now?

Won&#x27;t using Nvidia cards with Microsoft Olive also provide some boost in performance?

How does it compare to nvidias Jetson Orin?

Stay away from amd consumer GPUs they are not stable enough... neither hard nor software.

No it doesn’t. AMD drivers don’t support all of the extensions, optimizations, and related in things like automatic1111. There’s always stuff that breaks on AMD and works perfectly in CUDA land.