FROZEN EDITION — ISSUE 252 / 2026-08-25. THE CURRENT SHEET IS AT GPUDECK.COM →
Price says what the clock costs. Goodput says what the paid hour finishes.
● = this model also has measured goodput cells on the observatory
Every card, its own page →MEASURED ON ONE DISCLOSED FLEET · NEVER A RANKING
RENTAL RATE × GPU-HOURS ÷ COMPLETED RENDERS
↓
COST / 1,000 COMPLETED RENDERS
$24.95 — $79.65
3.2× SPREAD / ONE DISCLOSED FLEET
workload: wan_v2.2-14b-fp8_i2v_lightx2v
9 PRESENT NOW / 39 OBSERVED / 96 WORKERS
CONTEXT ONLY — NEVER WEIGHTED INTO GOODPUT · FULL LIFELINES →
SUPPRESSED WHEN UNDER-REPLICATED · NO SPONSORS · NO RANKINGS · REBUILD FROM CSV
The operations layer this record is measured with — pre-emptive memory recycling that saves paid jobs, earning-layer truth checks, incidents that heal themselves. Productizing next.
Join the waitlist →Two GPUs at the same hourly price can finish very different amounts of work — the measured spread on this fleet is printed above. One page, one picture, dated evidence.
Read the method →