Pricing
Overview
The GPU Trader platform allows you to rent high-performance GPUs on an hourly basis or through reserved instances. With reserved instances, suppliers offer lower hourly rates for renters who commit to daily, monthly or multi-month rental terms. This flexibility lets you select the best option for your workload and budget, whether you need on-demand, short-term, or cost-saving longer reserved instances.
How Pricing Works
- Supplier-Set Rates: The rental rate for each GPU is determined by the individual supplier. Rates vary based on GPU specifications, bandwidth features, and geographic location, offering a range of options to fit various budget and performance requirements.
- Pricing Transparency: Before confirming a rental, the Platform provides a detailed cost breakdown that includes the supplier’s rate for the selected GPU and any additional features or fees.
- Billing: Your rental period starts once the instance is successfully deployed. If renting hourly, you are billed hourly until you choose to stop or pause the instance. Charges reflect the actual hours used. If renting a reserved instance, you are responsible for paying for the entire reserved period, but are billed on a monthly basis, in advance.
Factors Influencing Rates
Because pricing is set by individual suppliers, rates may vary based on:
- GPU Model and Performance: Higher-performance GPUs, such as those with more CUDA cores, higher memory, or faster clock speeds, generally have higher hourly rates.
- Bandwidth Options: Suppliers may charge extra for features like Infiniband, RoCE, Ethernet, NVLink, or NVSwitch, depending on the bandwidth needed for your workload.
- Storage: Some rentals include SSD storage on the host device, with pricing reflecting the storage type and size, which is particularly useful for data-intensive tasks.
- Location: Suppliers may adjust rates based on the GPU’s physical location. Renting a GPU closer to your geographic area can reduce latency, which is beneficial for location-sensitive applications.
On-Demand vs. Reserved Pricing
- On-Demand Rentals: Hourly rates for on-demand rentals are set by the supplier. This option provides flexibility to rent GPUs for as long as needed and is ideal for shorter tasks or uncertain workloads.
- Reserved Instances: Suppliers also offer reserved instances at reduced hourly rates for renters who commit to longer usage terms. Reserved instances are available with daily, weekly, or monthly commitments, making them a cost-effective choice for projects with predictable or extended GPU requirements.
Comparing Pricing Options
| Pricing Model | Commitment | Pricing | Best For |
|---|---|---|---|
| On-Demand | No commitment | Hourly rates as set by supplier | Short-term tasks, unpredictable workloads |
| Reserved | Monthly commitment | Lowest hourly rate | Long-term, high consumption tasks |
Choosing the Right Rental Option
- On-Demand Rentals: For flexible, pay-as-you-go access, on-demand rentals provide an hourly rate with no long-term commitment. Use this option if you need the GPU for variable periods or for testing purposes.
- Reserved Instances (Monthly): For longer projects, monthly reserved instances offer the greatest savings. These are ideal for high-consumption projects like model training or video rendering.
Finalizing Your Rental
After selecting a GPU and choosing between on-demand or reserved pricing, review the final costs. Your rental will begin upon deployment, and charges will accrue according to your chosen pricing model. Reserved rentals automatically apply the supplier’s discounted rate based on your commitment duration. By understanding these options, you can select a rental plan that optimizes costs for your specific needs. For more help with choosing the best rental plan, contact our support team.