Measured wall-draw during sustained generation, priced at Korean residential rates (mid-tier ~150 KRW/kWh) and US average (~$0.16/kWh). 24/7 duty, one month = 720h.
| Card | Sustained W | USD/month 24-7 |
|---|---|---|
| RTX 5060 Ti 16GB | ~150 | ~$17 |
| RTX 5070 Ti | ~250 | ~$28 |
| RTX 4090 | ~350 | ~$39 |
| M5 Max (idle-heavy use) | ~40 avg | ~$4.5 |
Electricity is noise compared to the GPU price unless you run 24/7 idle-serving — which is exactly where Apple silicon’s 4W idle wins. For bursty chat/coding use, every card costs under $15/month. The real cost is the hardware, and it’s rising.
Methodology
Apple M5 Max 128GB unified memory, MLX/Ollama, tok/s = sustained generation over a 2K-token prompt, measured on this machine. Korea street prices from Danawa (multi-vendor median), converted at ~1,380 KRW/USD. Rows marked “public record” cite published benchmarks reproduced where possible; “measured” rows are from our hardware. Updated weekly — check the date in the title.
👉 Does it run on YOUR card? Check the VRAM Fit Matrix — measured estimates for every model above.
댓글 남기기