AMD·graphics card
Radeon RX 7900 XT
Based on 10 credible posts from Hacker News & Bluesky. 2 filtered out as bot-like or off-topic.
Low confidenceFew owner reviews so far, so this score can still move a lot. How we score
Scored November 11, 2026
We may earn a commission from this link. It never affects the score.
Owner reviews
Overall owner mood
Owners mostly like it, with reservations: 40% of 10 reviews are positive, 50% mixed, and only 10% negative. The most common complaint is performance (3 owners).
What owners like
Nothing two or more owners agree on yet.
What owners complain about
Performance raised by 3 owners
“Raising the context or loading the draft layers for MTP drops performance to 9-15 tok/s.”
What owners talk about
- Performance0 praise · 3 complain
Built from the reviews themselves — every quote links to the owner who wrote it.
Every credible owner review, most detailed first. Bots, off-topic comments, and ones that say nothing about using it are left out.
[...] I run mainline starting with -rc1 on my gaming pc, its been literally flawless for me since I switched to an AMD RX 7900 XT about two years ago. [...] I wasted a great deal more time backporting patches before I started doing that, than I have spent chasing new bugs since.
This is Unsloth's UD-Q4KS quantization (edit -- on llama.cpp, via the Vulkan backend, on an RX 7900 XT, with Unsloth's recommended sampler config), for "as replicable as LLMs can be" disclosure, done through llama.cpp's web UI. [...]
[...] I've got an RX 7900 XT (20GB of VRAM) and I can run glimmer with a full 128k context window with the draft model at 65-80 tok/s. [...] Raising the context or loading the draft layers for MTP drops performance to 9-15 tok/s. [...]
I've been building ANIMA, a locally-hosted persistent AI agent that maintains beliefs and memory across sessions. Qwen3-32B Q3KM instruct via llama.cpp, Vulkan on consumer AMD hardware (RX 7900 XT). [...] Some examples from the last week: belief pollution from unfiltered memory ingestion a base-model swap experiment that collapsed in hours a governance gap that allowed self-reinforcing dream loops The newest article looks at three slow failure modes persistent agents run into: echo chambers, belief crystallization, and curiosity collapse. [...]
I've got a 61mp camera, and an RX 7900XT. It takes about 15s/picture for DXO to denoise, which is a lot longer than people are willing to wait on a phone to take a photo. Topaz is even slower. [...]
[...] I think Radeon RX 7900 XT - 20 GB has been the best bang for your buck. [...]
That's what I have (RX 7900XT on Arch), and ROCm with pytorch has been reasonably stable so far. Certainly more than good enough for my experimentation. Pytorch itself has official support and things are pretty much plug & play.
I got a new PC. It is amazing! CPU: AMD Ryzen 7 9800X3D RAM: 32GB DDR5-6000 GPU: AMD Radeon RX 7900 XT Motherboard: MAG X670E TOMAHAWK WIFI Storage: WD Black SN850X SSD 1TB & 4TB Case: Lian Li Lancool 216 Monitors: LG 24" 1080p75Hz, LG Ultragear 27GN800 27" 1440p144Hz 1ms AMD Freesync Premium
In my experience using llama.cpp (which ollama uses internally) on a Strix Halo, whether ROCm or Vulkan performs better really depends on the model and it's usually within 10%. I have access to an RX 7900 XT I should compare to though.
On my machine with AMD RX 7900XT, it takes 0.17s per image. Are you using SD Turbo Scheduler node?
The claims, and the evidence
What the brand claims
No published claims on file for this product yet.
What owners report
“[...] I run mainline starting with -rc1 on my gaming pc, its been literally flawless for me since I switched to an AMD RX 7900 XT about two years ago. [...] I wasted a great deal more time backporting patches before I started doing that, than I have spent chasing new bugs since.”
after 2 yearsView on Hacker News“I got a new PC. It is amazing! CPU: AMD Ryzen 7 9800X3D RAM: 32GB DDR5-6000 GPU: AMD Radeon RX 7900 XT Motherboard: MAG X670E TOMAHAWK WIFI Storage: WD Black SN850X SSD 1TB & 4TB Case: Lian Li Lancool 216 Monitors: LG 24" 1080p75Hz, LG Ultragear 27GN800 27" 1440p144Hz 1ms AMD Freesync Premium”
“[...] I think Radeon RX 7900 XT - 20 GB has been the best bang for your buck. [...]”
“I've got a 61mp camera, and an RX 7900XT. It takes about 15s/picture for DXO to denoise, which is a lot longer than people are willing to wait on a phone to take a photo. Topaz is even slower. [...]”
Common Radeon RX 7900 XT problems
Problems most owners who mention them agree on, plus ones owners are split on.
- Performance3 owners report thisView on Hacker News
“Some examples from the last week: belief pollution from unfiltered memory ingestion a base-model swap experiment that collapsed in hours a governance gap that allowed self-reinforcing dream loops The newest article looks at three slow failure modes persistent agents run into: echo chambers, belief crystallization, and curiosity collapse.”