NVIDIA GeForce RTX 4080 SUPER — Benchmarks & Specs

Bottom line: how fast is the NVIDIA GeForce RTX 4080 SUPER?

SpecPicks median: 102 tokens/sec across 5 cited single-user runs of 7-9B dense models at 4-bit (Q4) under llama.cpp (range 54.4–148 tok/s), from llama.cpp GitHub, Hardware Corner (2 runs), MyAIHardware, LocalScore.ai. At the $1,644.05 Amazon price of the cheapest linked listing (checked 2026-10-20 16:14 UTC; prices change), that is a SpecPicks Value Index of 62.0 tok/s per $1,000. At 1440p Ultra, the NVIDIA GeForce RTX 4080 SUPER averages 119 fps in Tom's Hardware GPU Hierarchy rasterization suite (geomean), per Tom's Hardware. For local LLM inference it generates 62.0 tokens/sec running qwen3:14b at q4_K_M under llama.cpp, per Hardware Corner. In 3DMark Fire Strike it scores 55,495 points, per LanOC Reviews. 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, MXFP4), generating 139.1 tokens/sec under llama.cpp, per Hardware Corner.

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.