Use NVIDIA Sync to Connect Remotely to Your AI Compute

Step 1
Install NVIDIA Sync on your laptop (one time)

Your OS has been auto-detected, but you can choose a different option to download.

For Windows: After download, double-click the .exe installer and follow the instructions.

For macOS: After download, open nvidia-sync.dmg, drag and drop it into the Applications folder, then launch it from Applications.

For Debian/Ubuntu: Install from the NVIDIA APT repository.

First, configure the package repository:

Bash

curl -fsSL  https://workbench.download.nvidia.com/stable/linux/gpgkey  |  sudo tee -a /etc/apt/trusted.gpg.d/ai-workbench-desktop-key.asc
echo "deb https://workbench.download.nvidia.com/stable/linux/debian default proprietary" | sudo tee -a /etc/apt/sources.list

Then, update package lists:
Bash
```
sudo apt update
```
Finally, install NVIDIA Sync:
Bash
```
sudo apt install nvidia-sync
```

Success: A "Let's Get Started" modal opens and asks you to read and agree to the EULA.

Step 2
Agree to the EULA and select applications to launch (one time)

Click the link to read the EULA and select Agree in the "Let's Get Started" modal.
Then, NVIDIA Sync will present locally installed IDEs eligible to launch on a remote device. Select those you want configured.
Finally, select Next to proceed.

Success: A modal will open searching for your DGX Spark on the network.

NOTE

DGX Spark devices broadcast over mDNS by default. If the network allows mDNS, then you should see your device with an obvious name, e.g. spark-abcd.local. If your network blocks mDNS, then you won't see the name and must have the IP address.

Step 3
Add your DGX Spark (one time)

Select the device name if shown. Otherwise select Add a device manually.
Fill in the "Name", "Hostname or IP", "Username" and "Password" fields with relevant information for the device. Your password is used temporarily for set up and is not logged or stored.
Then select Add to proceed.
(Optional) If prompted, allow NVIDIA Sync to configure the Docker group on the device.

NOTE

Docker group configuration is not required to add the device, but if you use containers you will probably end up configuring it manually later because it's a major convenience that avoids entering a sudo password every time a docker command runs.
However, it does allow root access for applications or scripts running docker commands. Learn about the configuration and issues in the Docker user guide.

Success: A Get Started modal will appear.

Step 4
Launch applications on the remote Spark through NVIDIA Sync (repeat use)

Select Get Started to initiate the connection and open the task bar utility.
When the connection succeeds, select the pre-configured applications and start using your Spark.

Terminal launches an SSH shell on the remote.

DGX Dashboard launches a device management web application it on the remote and will open in a browser.

Resource Monitor launches an application window that shows CPU, sRAM and GPU utilization on the device.

Success: The applications start and open.

Next steps

Use NVIDIA Sync to add Tailscale so you can connect to your device from anywhere
Use NVIDIA Sync to cluster two or more DGX Spark devices
Use NVIDIA Sync to launch a vLLM container

Use NVIDIA Sync to Connect Remotely to Your AI Compute

Step 1Install NVIDIA Sync on your laptop (one time)

Step 2Agree to the EULA and select applications to launch (one time)

Step 3Add your DGX Spark (one time)

Step 4Launch applications on the remote Spark through NVIDIA Sync (repeat use)

Next steps

Resources

Step 1
Install NVIDIA Sync on your laptop (one time)

Step 2
Agree to the EULA and select applications to launch (one time)

Step 3
Add your DGX Spark (one time)

Step 4
Launch applications on the remote Spark through NVIDIA Sync (repeat use)