Estimate how many GPUs an AI translation workload needs from task volume, GPU-seconds per task, utilization, and redundancy. The result includes raw compute demand, utilization-adjusted capacity, and a practical provisioned GPU count.
Raw GPU-hours = tasks × GPU seconds per task ÷ 3,600. Base GPUs = ceiling(raw GPU-hours ÷ operating hours ÷ target utilization). Provisioned GPUs = base GPUs + redundant GPUs.
The provisioned count covers average compute demand within the selected operating window and adds explicit redundancy. Memory fit, model-loading overhead, batching efficiency, and peak traffic must be checked separately.
Planning estimate only. Validate the result against production telemetry and current provider pricing before committing capacity or budget.
120,000 tasks at 1.2 GPU-seconds each require 40.0 raw GPU-hours per day; at 70% utilization over 24 hours, the base requirement is 3 GPUs before the spare.
Use recent production telemetry or a representative benchmark on the same model, hardware, input size, and quality settings.
No. It focuses on the entered operating assumptions; add fixed staffing, development, and platform overhead separately when building a full budget.
Use the observed fraction of attempts that must be repeated, and avoid adding the same retry overhead to more than one input.
A vendor maximum is a useful ceiling, but measured sustained performance is safer for capacity planning.
Input size, batching, model version, hardware, cold starts, safety checks, and traffic bursts can all change actual results.
| Input | Purpose |
|---|---|
| Workload values | Define the size and frequency of the task. |
| Performance values | Describe measured processing behavior. |
| Planning assumptions | Add utilization, retries, price, or headroom where applicable. |