2x RTX 3090 24 GB

48 GB · 24 GB each · dual-card layer split · $3,000 Card · $4,200 Full rig

Quick decision verdict

2x RTX 3090 24 GB outputs 111.3 tok/s at $0.082/M tokens (rank #13 of 28 measured rigs on unit cost), breaking even in 598 days at full load.

  • Generation speed rank #3 of 29 · System power rank #29 (850 W).
  • For a lower upfront build cost, compare against RTX 4090 ($4,199)
  • If you need faster interactive token speed, compare against RTX PRO 6000 Blackwell (171.1 tok/s)

Can I Run It? → · Best → · Method →

How you'll use it

Fits?Source
2x RTX 3090 24 GB
24 GB each · dual-card layer split
48 GB2,137.6111.3Tight850 W$4,20065.5$0.082$3.06$10.08598 daysMeasured

Selected hardware

2x RTX 3090 24 GB

48 GB · 24 GB each · dual-card layer split · Layer split across two cards

Read tok/s
2,137.6
Write tok/s
111.3
Power
850 W
Output/day
65.5
Cost/M tok
$0.082
Power/day
$3.06
Saved/day
$10.08
Payback
598 days

Details · See source

FAQ

How fast does 2x RTX 3090 24 GB run local LLMs?
2x RTX 3090 24 GB (48 GB) measures 111.3 tokens/s generation and 2,138 tokens/s prompt reading on Qwen3.8-Flash-Next (snapshot 2026-10-07).
Is 2x RTX 3090 24 GB worth buying over cloud LLM APIs?
At $0.082/M tokens local cost, 2x RTX 3090 24 GB breaks even in 598 days at 24h/day utilization, or 3,588 days at 4h/day — making daily utilization the main deciding factor.