AMD Radeon RX 9070 — Benchmarks & Specs

Bottom line: how fast is the Radeon RX 9070?

SpecPicks median: 99.4 tokens/sec across 5 cited single-user runs of 7-9B dense models at 4-bit (Q4) under llama.cpp (range 81.7–114 tok/s), from KnightLi GPU Benchmark Scoreboard, llama.cpp GitHub (2 runs), vachsark.com, localaimaster.com. At the $699.99 Amazon price of the cheapest linked listing (checked 2026-10-21 20:20 UTC; prices change), that is a SpecPicks Value Index of 142.0 tok/s per $1,000. At 1440p Maximum, the Radeon RX 9070 averages 126 fps in Final Fantasy XIV Dawntrail, per Gamers Nexus. For local LLM inference it generates 54.9 tokens/sec running qwen3-coder:14b at q4_K_M under llama.cpp, per vachsark.com. In 3DMark Solar Bay it scores 106,882 points, per Notebookcheck. Its 16 GB of VRAM is the binding constraint for local inference: the largest model on file at 4-bit quantization on this card is gpt-oss:20b (20B parameters, q4_K_M), generating 91.9 tokens/sec under ollama, per GitHub hirokuze/local-llm-benchmark-rx9070xt.

Every figure above is a row in the tables below, and each row links out to the review or public benchmark database the number was taken from; the SpecPicks median is computed here from the runs it names, each a row in the AI table, and the per-$1,000 figure divides it by the dated listing price. SpecPicks aggregates published measurements; it does not report first-party benchmark runs.