---
title: "Set Up RTX Spark as Your Always-On AI Hub — Remote Access"
canonical: "https://build.nvidia.com/playbooks/ai-hub/overview.md"
---

# Basic idea

[RTX Spark compact desktops](https://www.nvidia.com/en-us/products/rtx-spark/#compact-desktop) are power-efficient PCs, which makes them well suited to staying on around the clock. This playbook shows you how to set up your RTX Spark as an always-on AI hub that you reach from your other devices. It is written for the compact desktop, but the steps apply to any RTX Spark.

The playbook has two parts:

1. **Set up remote access (this tab).** Keep RTX Spark awake and control it from another PC.
2. **Unlock a use case.** This playbook highlights different use cases where you can run agents and creator workflows differently across your hardware units, taking advantage of the “always on” nature of RTX Spark.

If something does not work, open the **Troubleshooting** tab.

# Step 1. Set up Windows on RTX Spark

For the first setup, connect a monitor, keyboard, and mouse to RTX Spark. You can remove them in Step 5.

1. **Connect power, network, and peripherals.** Use an Ethernet cable to your router if you can. For Wi-Fi, have your network name and password ready.
2. **Turn on RTX Spark** and wait for the Windows setup screen.
3. **Complete Windows setup.** Follow the on-screen prompts, and allow any updates and restarts to finish. When you see **Your PC is ready**, select **Go to Desktop**, or select **Show optional setup** for a head start on installing an agent.

# Step 2. Keep RTX Spark awake

A PC that is asleep cannot be reached remotely, so turn off sleep and hibernate.

1. On RTX Spark, open **Settings > System > Power**.
2. Expand **Screen, sleep & hibernate timeouts**.
3. Set **Make my device sleep after** and **Make my device hibernate after** to **Never**.

![Windows Settings Power page with the sleep and hibernate timeouts set to Never](https://assets.ngc.nvidia.com/products/api-catalog/rtx-spark/ai-hub-power-sleep-settings.png)

Three more settings are worth a look on an always-on device:

- **Device name.** You will use this name to connect from your other devices. Go to **Settings > System > About**, select **Rename this PC**, enter a name such as `rtx-spark`, and restart.
- **Windows Update.** Windows restarts on its own to install updates. To keep those restarts away from the hours you rely on RTX Spark, set your active hours in **Settings > Windows Update > Advanced options**.
- **Power mode.** On the **Power** page, **Power mode** sets the balance between performance and power draw. **Best Performance** favors performance and can use more power.

# Step 3. Set up remote desktop

Remote desktop lets you control RTX Spark from another PC, as though you were sitting in front of it. Choose one option:

- **Microsoft Remote Desktop** if RTX Spark runs Windows Pro. It is built in to Windows.
- **Chrome Remote Desktop** if RTX Spark runs Windows Home, or if you want a free option that works away from home with no extra setup.
- **RealVNC Connect** as another option that works on Windows Home.

To check your Windows edition, open **Settings > System > About**.

|   | Microsoft Remote Desktop | Chrome Remote Desktop | RealVNC Connect |
| :--- | :--- | :--- | :--- |
| Windows edition on RTX Spark | Pro, Enterprise, or Education | Home or Pro | Home or Pro |
| Cost | Included with Windows | Free | Free Lite plan for personal use |
| Account | Windows account with a password | Google account | RealVNC account |
| Use away from home | With Tailscale (Step 4) | Yes | Yes |
| Use with no monitor attached | Yes | May need a dummy display plug | May need a dummy display plug |
| Frame rate | 30 FPS, or 60 FPS with a [registry setting](https://learn.microsoft.com/en-us/troubleshoot/windows-server/remote/frame-rate-limited-to-30-fps) | 30 FPS | Not published |

All three let you sign in to Windows again after RTX Spark restarts. Follow the section for the option you chose.

## Microsoft Remote Desktop

1. On RTX Spark, open **Settings > System > Remote Desktop** and turn on **Remote Desktop**.

![Windows Settings Remote Desktop page with Remote Desktop turned on](https://assets.ngc.nvidia.com/products/api-catalog/rtx-spark/ai-hub-remote-desktop-settings.png)

2. On your other PC, open **Remote Desktop Connection**, enter RTX Spark’s device name, and select **Connect**.

![Remote Desktop Connection dialog with rtx-spark in the Computer field](https://assets.ngc.nvidia.com/products/api-catalog/rtx-spark/ai-hub-remote-desktop-connection.png)

3. Sign in with your Windows account password. A PIN or other Windows Hello sign-in does not work over Remote Desktop.

For other devices and more detail, see Microsoft’s guide, [How to use Remote Desktop](https://support.microsoft.com/en-us/windows/experience/connectivity-networking/how-to-use-remote-desktop).

## Chrome Remote Desktop

1. On RTX Spark, install [Google Chrome](https://www.google.com/chrome/) and sign in with your Google account. Chrome is not installed by default.
2. In Chrome, open [remotedesktop.google.com/access](https://remotedesktop.google.com/access). Under **Set up remote access**, download and install the host software, then choose a name and a PIN. Use **Remote Access**, not **Remote Support**.
3. On your other device, open the same page in Chrome, or open the Chrome Remote Desktop mobile app. Sign in with the same Google account, select RTX Spark, and enter the PIN.

For more detail, see Google’s guide, [Access another computer with Chrome Remote Desktop](https://support.google.com/chrome/answer/1649523).

## RealVNC Connect

1. Create a RealVNC account and choose the free [Lite plan](https://www.realvnc.com/en/connect/plan/lite/). It covers personal, non-commercial use for one user and up to three devices.
2. On RTX Spark, install [RealVNC Server](https://www.realvnc.com/en/connect/download/vnc/) and sign in with your RealVNC account.
3. On your other device, install [RealVNC Viewer](https://www.realvnc.com/en/connect/download/viewer/), sign in with the same account, and select RTX Spark.

## Check your connection

From your other PC, connect to RTX Spark. You should see its Windows desktop and be able to open apps.

# Step 4. Add Tailscale for access away from home (optional)

[Tailscale](https://tailscale.com) creates a private network between your own devices, so you can reach RTX Spark from outside your home without opening ports on your router. It is free for personal use.

You need Tailscale to use Microsoft Remote Desktop away from home, or to follow the Tailscale route in the **Inference** tab. Otherwise, skip to Step 5.

1. [Install Tailscale](https://tailscale.com/download) on RTX Spark and on each device you will connect from. Sign in with the same account on every device.
2. On RTX Spark, right-click the Tailscale icon in the system tray and select **Preferences > Run unattended**. Tailscale then stays connected after a restart, before anyone signs in to Windows.
3. In the [Tailscale admin console](https://login.tailscale.com/admin/machines), open the menu next to RTX Spark and select **Disable key expiry**. Otherwise, RTX Spark leaves your Tailscale network after 180 days until you sign in again.
4. On another device, run `tailscale status` to find RTX Spark’s Tailscale name, then check that you can reach it:

```powershell
tailscale ping rtx-spark
```

You can now use RTX Spark’s Tailscale name wherever this playbook asks for its address, at home or away. To limit which users and devices can connect, see [Tailscale access controls](https://tailscale.com/docs/features/access-control).

# Step 5. Go headless (optional)

With remote desktop working, RTX Spark no longer needs its own monitor, keyboard, and mouse. You can also leave them connected and use RTX Spark at a desk.

1. **Test a restart first.** With the monitor still connected, restart RTX Spark and wait a few minutes. Then connect and sign in through remote desktop, without touching RTX Spark’s keyboard.
2. **Unplug.** Shut down RTX Spark and unplug the monitor, keyboard, and mouse. Leave power and the network cable connected.
3. **Reconnect.** Move RTX Spark to its permanent spot, leave room for airflow, and turn it on. Connect with remote desktop and sign in.

> [!NOTE]
> Chrome Remote Desktop and RealVNC show RTX Spark’s own display. If the screen is black or low resolution once the monitor is unplugged, connect a dummy display plug to RTX Spark or leave a monitor connected. A dummy display plug is a small adapter that you buy separately, often sold as an HDMI dummy plug or a headless display adapter. Choose one that fits a display output on your RTX Spark. Once it is plugged in, Windows acts as though a monitor is connected. Microsoft Remote Desktop creates its own display, so it is not affected.

# Step 6. Choose your use case

Remote access is ready. Now choose what RTX Spark will do, and open the matching tab.

| Tab | Agent runs on | Model runs on | Choose it if |
| :--- | :--- | :--- | :--- |
| **Agents** | RTX Spark | RTX Spark | You want one always-on agent that you reach from your phone or any PC. |
| **Inference** | Your laptop | RTX Spark | You want to keep your agent on your laptop, close to your files, and use RTX Spark to power the model. |
| **Creators** | No agent needed | RTX Spark, in ComfyUI | You want to generate images and video from any PC in your home. |

Agents can be lightweight, but running AI models can use a lot of computational power. RTX Spark is perfectly suited for running AI models optimally. Follow the **Agents** tab or the **Inference** tab based on where you want your agent or model to run. The diagram below shows the different approaches available to you.

![Where the agent and the model run. Agents tab: the agent and the model both run on RTX Spark, and your phone or other PC reaches the agent through a chat app or remote desktop, over your home network or cellular. Inference tab, Route A: the agent runs on your laptop and sends requests to PAIR on the laptop, which routes them over your home network to PAIR on RTX Spark and the model. Inference tab, Route B: the agent runs on your laptop and sends requests over a Tailscale private network, with access from anywhere, to the model on RTX Spark.](https://assets.ngc.nvidia.com/products/api-catalog/rtx-spark/ai-hub-connection-paths.png)

You can set up more than one use case. Models from different use cases share RTX Spark’s memory, so check that they fit together before you run them at the same time.