User guide

Rent a GPU & connect in 5 minutes

Everything you need to get started: rent a machine, connect over SSH, open the browser desktop, run Jupyter and move files in/out.

Rent a GPU now Open GPU Manager

1. Rent a GPU machine

Top up your CKEY balance (Deposit section) — machines are billed hourly from your balance.
Open the Rent GPU page, pick a card (RTX 4090, A100…) and a configuration that fits your workload.
Click Rent / Build and wait a few minutes for provisioning. The machine is ready when its status turns Online.

Machines that expire or run out of balance are stopped. Renew before expiry to keep your data.

The most common way to use your rental. Open the GPU Manager → select your machine to see the SSH command and password:

Copy the SSH command shown on the management page, it looks like: ssh root@<ip-máy> -p <port>
Paste it into a terminal: use Terminal on macOS/Linux, PowerShell (built-in) or PuTTY on Windows.
On first connect, type yes when asked about the fingerprint.
Enter the password (shown right on the management page) when prompted. Typing shows no characters — just type and press Enter.

Tip: Windows 10/11 already ships the ssh command in PowerShell — nothing to install.

If your machine has a graphical desktop, the management page shows a Desktop button (noVNC):

Open the GPU Manager → select your machine.
Click the Desktop button/link to open the graphical desktop right in your browser — nothing to install.
Log in with the password shown on the management page if asked.

Your GPU machine has a port forward table on the management page — each internal port is mapped to a public port:

SSH into the machine, start a service listening on one of the listed ports (e.g. Jupyter): jupyter notebook --ip=0.0.0.0 --port=8888 --no-browser --allow-root
Check the port forward table on the management page: find the public port mapped to internal port 8888.
Open http://<machine-ip>:<public-port> in your browser and paste the Jupyter token printed in the terminal.

Only ports listed in the port forward table are reachable from the internet.

Use scp or sftp with the same host, port and password as SSH:

Upload a file to the machine:

scp -P <port> ./file.zip root@<ip-máy>:/root/

Download a file to your computer:

scp -P <port> root@<ip-máy>:/root/ket-qua.zip ./

For large files, consider archiving first (tar/zip) or using rsync to resume interrupted transfers.

SSH says Connection refused / timeout?

Check the machine is still within its rental period and Online in the GPU Manager. Out-of-balance or expired machines are stopped and unreachable.

Forgot the password?

The password is always shown on the machine management page (next to the SSH command).

Cannot open a web app (Jupyter…)?

Make sure the service listens on a port listed in the port forward table, and use the matching public port.

Lost data after rebuild / expiry?

Rebuilding or leaving a machine expired too long wipes its data. Always back up important files locally (scp) or to cloud storage.

Back to home