Guide4 sections2 min read
Start with the workload
Write down your framework, GPU memory requirement, expected runtime and preferred region. Choose from live offers rather than assuming that every GPU in a specification page is available.
Check the GPU count, CPU, RAM, storage, connectivity and rental conditions of the complete configuration. A GPU model alone does not describe the whole machine.
Prepare your account and access
Create an account in the portal, review the payment and rental conditions, and add a public SSH key. Keep the private key on your own device.
Fund the appropriate currency balance before launching. Review the price shown for the selected offer and any separately billed resources.
Launch and connect
Select an available offer, a supported environment and your SSH key. Follow the instance status in the portal, then use the displayed SSH host, port and username when it is ready.
Verify the GPU and software environment before starting a long job. Downloading models and preparing dependencies can also consume time while the instance is running.
Finish the session deliberately
Save your results before releasing the instance. Check whether the provider supports stopping without continued compute charges; otherwise terminate the instance when finished.
Review persistent volumes separately. Retained storage may remain billable after compute ends. Confirm the final resource state and consumption in the portal.