AMD Radeon RX 5700 XT
RDNA 1 flagship. ROCm support was always experimental and is effectively defunct in 2026. Vulkan via llama.cpp is the only operator-grade path; performance is bandwidth-OK but lacks the matrix cores newer cards have. ~25-40 tok/s on 7B Q4. The card the 'I bought AMD before AI mattered' audience owns.
Affiliate disclosure: as an Amazon Associate and partner of other retailers, we earn from qualifying purchases. The verdict on this page is our editorial opinion; affiliate links never influence what we recommend.
Sub-scores sum to 306 / 1000. Headline = 306 × 0.70 (Estimated-confidence discount) = 214. This is an algorithmic performance-tier score — distinct from, and often lower than, the editorial “Our verdict” below, which weighs value and real-world fit (especially for hardware we haven’t measured yet). How scoring works →
Extrapolated from 448 GB/s bandwidth — 44.8 tok/s estimated. No measured benchmarks yet.
Plain-English: Comfortable for 7B chat.
Verdicts extrapolated from catalog VRAM + bandwidth + ecosystem flags. Hover any chip for the rationale. Want measured numbers? Submit your own run with runlocalai-bench --submit.
This card is for the operator who already owns it from a gaming build and wants to dip a toe into local inference without spending more money. It is not for anyone buying a GPU today for local AI. On 7B Q4 models via Vulkan in llama.cpp, expect 25-40 tok/s—usable for chat but not for real-time streaming. 13B Q4 models (8.5 GB weights) barely fit in 8 GB VRAM, yielding ~15-25 tok/s with heavy system RAM spillover. 30B models are out of reach entirely. The lack of ROCm and CUDA means no access to exllama, AutoGPTQ, or any optimized inference stack; Vulkan is the only path and lags in features and performance. Pass on this card if you are buying a GPU for local AI today—an RTX 3060 12 GB offers CUDA, more VRAM, and similar speed for ~$50 more used. At $200 used, it is a budget option only for the AMD-curious who already have the card; otherwise, skip it.
›Why this rating
The RX 5700 XT is usable for small models via Vulkan but lacks CUDA/ROCm, limiting software compatibility and future-proofing. Its 8 GB VRAM is tight for 13B models, and performance is mediocre. Only worth considering if acquired cheaply from an existing build.
Overview
RDNA 1 flagship. ROCm support was always experimental and is effectively defunct in 2026. Vulkan via llama.cpp is the only operator-grade path; performance is bandwidth-OK but lacks the matrix cores newer cards have. ~25-40 tok/s on 7B Q4. The card the 'I bought AMD before AI mattered' audience owns.
Some links above are affiliate links. We may earn a commission at no extra cost to you. How we make money.
Specs
| VRAM | 8 GB |
| Power draw (peak) | 225 W |
| Released | 2019 |
| MSRP | $399 |
| Backends | Vulkan |
Models that fit
Open-weight models small enough to run on AMD Radeon RX 5700 XT with usable context.
Hardware worth comparing
The closest alternatives by price, memory bandwidth, and form factor, plus a step up and down — so you can frame the buying decision against real options.
Frequently asked
What models can AMD Radeon RX 5700 XT run?
Does AMD Radeon RX 5700 XT support CUDA?
How much does AMD Radeon RX 5700 XT cost?
Where next?
Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify hardware specifications.