llamaperf
Submit

RTX 3090 vs RTX 4090

For running local LLMs · 10 reports across 3 models

Side A
RTX 3090
Vendor
nvidia
VRAM
24GB
Memory
Discrete
Side B
RTX 4090
Vendor
nvidia
VRAM
24GB
Memory
Discrete

Tokens per second by model

ModelRTX 3090RTX 4090
Qwen3.6up to 35B72.9n=680.0n=2
Gemma 4up to 31B—149.6n=1
Qwen2.528.0n=1—

More comparisons