M2 16GB vs RTX 3090

For running local LLMs · 7 reports across 3 models

Side A

Vendor: apple
VRAM: 16GB
Memory: Unified

Side B

Vendor: nvidia
VRAM: 24GB
Memory: Discrete

Tokens per second by model

Model	M2 16GB	RTX 3090
Qwen3.6up to 35B	—	66.0n=5
Gemma 4up to 31B	18.0n=1	—
Qwen2.5	—	28.0n=1

More comparisons