NVIDIA GeForce RTX 3080 — Benchmarks & Specs

Bottom line: how fast is the NVIDIA GeForce RTX 3080?

SpecPicks median: 98.7 tokens/sec across 11 cited single-user runs of 7-9B dense models at 4-bit (Q4) under llama.cpp (range 55.7–140 tok/s), from llama.cpp GitHub (2 runs), knightli.com, XiongjieDai/GPU-Benchmarks-on-LLM-Inference (3 runs), LocalScore.ai, Hardware Corner (3 runs), Kunal Ganglani LLM Benchmarks. At the $759.00 Amazon price of the cheapest linked listing (checked 2026-10-20 22:15 UTC; prices change), that is a SpecPicks Value Index of 130.0 tok/s per $1,000. One of those runs: llama3:8b at q4_K_M generates 106.4 tokens/sec under llama.cpp, per XiongjieDai/GPU-Benchmarks-on-LLM-Inference. At 1440p High with DLSS Quality upscaling, the NVIDIA GeForce RTX 3080 averages 103 fps in The Finals with a 61 fps 1% low, per NoobFeed. In PassMark G3D Mark it scores 25,024 pts, per PassMark. It has 10 GB of VRAM, and the llama3:8b run above (8B parameters) is the largest model on file at 4-bit quantization on this card. A second published run of the same setup measured 109.6 tokens/sec, per XiongjieDai/GPU-Benchmarks-on-LLM-Inference again, so expect 106.4-109.6 tokens/sec. Larger models have no sourced run on file.

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.