Calculate allocated GPU-hours and cloud cost, reverse-solve wall time with scaling efficiency, or estimate training duration from samples, epochs, and throughput.

Allocated GPU-hours count reserved devices, including idle time. Throughput and scaling estimates assume a constant rate; cloud billing rules and other charges may differ.

Repeated runs and compute price

Blank = one run.

Blank = $0 for usage-only estimates; enter your actual compute rate for cost.

Related Calculators

GPU Hours Formula

The following formula is used to calculate the GPU hours for a given number of GPUs and hours per GPU.

H = N × h

Variables:

  • H is the total GPU hours
  • N is the number of GPUs
  • h is allocated wall-clock hours per GPU, including idle time. For repeated runs, sum their wall time.

To calculate the total GPU hours, multiply the number of GPUs by the hours per GPU.

What are GPU Hours?

GPU hours refer to the total amount of time that a GPU (Graphics Processing Unit) is used for processing tasks. This metric is often used in cloud computing and high-performance computing environments to measure the usage and cost associated with GPU resources. GPU hours can be calculated by multiplying the number of GPUs by the number of hours each GPU is used. This helps in understanding the computational resources consumed and can be useful for budgeting, resource allocation, and performance analysis.

How to Calculate GPU Hours?

The following steps outline how to calculate the GPU Hours.

  1. First, determine the number of GPUs (N).
  2. Next, determine the hours per GPU (h).
  3. Finally, calculate the total GPU hours using the formula H = N * h.
  4. After inserting the values and calculating the result, check your answer with the calculator above.

Example Problem : 

Use the following variables as an example problem to test your knowledge.

Number of GPUs (N) = 4

Hours per GPU (h) = 5; total allocated usage =4 ×5 =20 GPU-hours.