NVIDIA GeForce RTX 3090 Ti — Benchmarks & Specs

Bottom line: how fast is the NVIDIA GeForce RTX 3090 Ti?

SpecPicks median: 110 tokens/sec across 7 cited single-user runs of 7-9B dense models at 4-bit (Q4) under llama.cpp (range 85.0–172 tok/s), from llama.cpp GitHub Discussions (2 runs), Hardware Corner (2 runs), LocalScore.ai, kunalganglani.com community benchmark, MyAIHardware. At the $2,189.97 Amazon price of the cheapest linked listing (checked 2026-10-20 16:16 UTC; prices change), that is a SpecPicks Value Index of 50.2 tok/s per $1,000. At 1440p High, the NVIDIA GeForce RTX 3090 Ti averages 126 fps in Dying Light 2, per KitGuru. For local LLM inference it generates 139.6 tokens/sec running qwen3.6:35b-a3b (a mixture-of-experts model with ~3B active parameters, so dense models of similar size run far slower) at q4_0 under llama.cpp, per Giles Thomas blog. In PassMark G3D Mark it scores 29,284–29,286 pts, across two PassMark results (PassMark GPU benchmark listing, PassMark GPU benchmark listing). Its 24 GB of VRAM is the binding constraint for local inference: the qwen3.6:35b-a3b run above (35B parameters) is the largest model on file at 4-bit quantization on this card. That run used a 89,600-token context; the Will-It-Run matrix sizes each model at its listed quantization with an ~8K context, and by that measure DeepSeek-R1-Distill-Qwen 32B at Q4_K_M is the largest that fits in 24 GB.

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.