DiggTop

RTX 3060 x2 vs RTX 3090 x2

Best verified output tok/s per model (single-stream generation). Hover a value for the quant.

RTX 3060 x229 verified runs · best 198.1 tok/s
RTX 3090 x26 verified runs · best 240.8 tok/s
ModelRTX 3060 x2RTX 3090 x2
Qwen3.6-27B18.9 IQ4_NL240.8 AutoRound INT4
Qwen3.8-27B219.8 AutoRound INT4 + fp8 KV
Qwen3-1.7B198.1 Q4_K_M
NVIDIA-Nemotron-3.5-Lightning-30B-A3B197.0 NVFP4
MiniCPM5-1B125.9 Q4_K_M
Trinity-Mini109.0 Q4_K_M
Phi-4-mini-reasoning108.2 Q4_K_M
gemma-4-26B-A4B-it-qat95.3 Q4_K_M
Qwen3-30B-A3B-Thinking-250794.3 Q4_K_M
gpt-oss-20b93.4 Q4_K_M
OpenReasoning-Nemotron-7B68.0 Q4_K_M
WebWorld-8B62.3 Q4_K_M
gemma-4-26B-A4B-it62.1 UD-Q4_K_M
Carnice-9b56.1 Q4_K_M
Muse-Glimmer-30B56.0 UD-Q4_K_XL
Qwen3.8-Flash-Next-REAP-256-duo53.5 Unsloth-Dynamic-Q3_K_XL
NVIDIA-Nemotron-Nano-12B-v240.4 Q4_K_M
Qwen3.8-Flash-Next40.1 Unsloth-Dynamic-IQ3_XXS
gemma-3-12b-it39.4 Q4_K_M
Qwen3-14B35.8 Q4_K_M
Nemotron-Cascade-14B-Thinking35.7 Q4_K_M
Phi-4-reasoning35.4 Q4_K_M
WebWorld-14B35.4 Q4_K_M
InternVL3-14B35.1 Q4_K_M
DeepSeek-R1-Distill-Qwen-14B35.1 Q4_K_M
phi-434.2 Q4_K_M
DeepHermes-3-Mistral-24B-Preview24.5 Q4_K_M
reka-flash-323.9 Q4_K_M
mistralai_Mistral-Small-3.2-24B-Instruct-250621.8 Q4_K_M
Tess-4-27B18.8 Q4_K_M
GLM-Z1-32B-041416.6 Q4_K_M
Olmo-3-32B-Think16.4 Q4_K_M
gemma-4-31B-it15.8 Q4_K_M

Head-to-head (models tested on both): RTX 3060 x2 0 · RTX 3090 x2 1 · tied 0 · data as of Sep 23, 2026. Full rules: methodology.

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