RTX 3090 Beats 64GB Mac Studio with 140 TPS Local AI Speed
TL;DR
Running local AI models such as Qwen 3.6, a 35-billion-parameter system, means balancing performance, memory capacity and cost. The Stack compares a custom-built PC with an RTX 3090 GPU against a 64GB Mac Studio with the M5 Max chip. The PC reaches around 140 tokens per second and leads on speed, while the Mac keeps advantages in memory and power draw.
Nauti's Take
A used RTX 3090 delivers surprisingly strong local AI speed and is a cheap entry point for tinkerers. The catch is limited VRAM, high power draw and noise, so larger models still hit memory limits.
Those who want raw speed should lean toward the PC, while those who value quiet and large context windows should look closer at the Mac.
Video
Sources