Multi-GPU x2 vs RTX 3090 x2

Best sourced output tok/s per model (single-stream generation) — engine and version, quantization, context and a source link on every row. Hover a value for the quant.

Sourced data on this pair is thin — most of the board is still partial. Submit your own run and help fill the matrix: one prompt, three steps, no downloads.

Multi-GPU x20 sourced runs · best — tok/s

RTX 3090 x27 verified runs · best 240.8 tok/s

Specs and evidence

	Multi-GPU x2	RTX 3090 x2
VRAM	24 GB	24 GB
Sourced runs on the board	0	7
Models measured	0	6
Best output tok/s	—	240.8 tok/s
Highest measured peak VRAM	—	42.9 GB
Street price (latest snapshot)	—	—

Sourced rows only. Peak VRAM is recorded by the person who ran the benchmark, so it is a floor for what the card can hold, not a specification. Price snapshots.

The measured matrix

Model	Multi-GPU x2	RTX 3090 x2
Qwen3.6-27B	—	240.8 AutoRound INT4
Qwen3.8-27B	—	219.8 AutoRound INT4 + fp8 KV
NVIDIA-Nemotron-3.5-Lightning-30B-A3B	—	197.0 NVFP4
MiniCPM-V 4.7-35B-A3B	—	172.2 int4
Qwen3.8-Flash-Next-REAP-256-duo	—	53.5 Unsloth-Dynamic-Q3_K_XL
Qwen3.8-Flash-Next	—	40.1 Unsloth-Dynamic-IQ3_XXS

6 rows · 0 models on Multi-GPU x2, 6 on RTX 3090 x2 — no verified overlap yet, each side shows its own bests · data as of Oct 17, 2026. Full rules: methodology.

Related reading: RTX 5090 vs 3090 · quantization · Value index · Full board