Dedicated GPU Server: Rack Installation and Initial Setup

## 1. Required Materials

| Name | Purpose | Note |
|---|---|---|
| Server | Main compute unit | Delivered by Parallel Cloud |
| Guide rail | Mounts to the cabinet rack to support the server | Parallel Cloud Delivery — usually shipped separately |
| GPU spoof card | Inserted into the GPU's DP port (one per GPU) | Parallel Cloud Delivery — usually included inside the server box |
| IP address | Required for network communication; each node needs at least one fixed IP | Customer-provided; plan the number of IPs in advance |
| Ethernet cable | Network communication | Customer-provided; one cable per GPU node |
| Switch port | Network connection point for each node | Customer-provided; one switch port per GPU node |

## 2. Server Rack Installation Steps

| Step | Action | Description | Image | Notes |
| :---: |---|---|---|---|
| 1 | Install guide rails or L-type brackets | Mount rails to the cabinet | ![](https://docs.paraverse.cc/media/202504/fb1dda462419430a99896fdd28b09fd01403.png) | |
| 2 | Mount the server | Slide the server onto the installed rails | ![](https://docs.paraverse.cc/media/202504/890c63747cc3480cbdde6cf0a48be4ac1919.png) | |
| 3 | Insert the power cord | Connect the server to the power supply | ![](https://docs.paraverse.cc/media/202504/e89905cc4a3a43e58a15af542135f6304486.png) | |
| 4 | Insert the Ethernet cable | Connect the server to the network switch | ![](https://docs.paraverse.cc/media/202504/b02279f735a24fcea4da40e1e560d4de6838.png) | |
| 5 | Power on | Press the power button on each GPU node | ![](https://docs.paraverse.cc/media/202504/e855d8da9139456492af1595e0bf15e86866.png) | Default credentials: `administrator` / `siton-2009` |
| 6 | Connect a monitor | Required to configure IP and run LarkXR | | Recommended resolution: 1920x1080 |
| 7 | Configure IP addresses | Each node in the PCFARM server is an independent OS that needs its own IP | ![](https://docs.paraverse.cc/media/202504/57a708c251254be2ab46242b1695eecd3424.png) | |
| 8 | Configure the management program | Management node setup | ![](https://docs.paraverse.cc/media/202504/25fee3b933f54b6e859725174b9d43998577.png) | Only one GPU node runs the management program. Locate it first and configure per the instructions in the image. |
| 9 | Configure the rendering program | Rendering node setup | ![](https://docs.paraverse.cc/media/202504/d10a4bc5abf14f3eaf7486386025a1d83149.png) | The rendering program must run on every GPU node. |
| 10 | Insert the GPU spoof card | Plug in the graphics card spoofing device | ![](https://docs.paraverse.cc/media/202504/27fd5e5a4aff4926aeef0fac3144ab5e6162.png) | Insert last — plugging it in earlier causes split-screen on connected monitors. |
| 11 | Log in to the LarkXR backend | Access `https://<center-ip>:8182` | ![](https://docs.paraverse.cc/media/202504/cb4b4bd9bedd48f791f8764afc4820ba6436.png) | Default credentials: `admin` / `123456` |
| 12 | Log in to the LarkXR frontend | Access `https://<center-ip>:8181` | | |

## 3. Notes

- Do not use Windows built-in Remote Desktop (RDP). Use **VNC** (pre-installed), **TeamViewer** (pre-installed), or **Sunflower** instead.
- After connecting the Ethernet cable, verify that the indicator lights on both the switch port and the server's NIC are blinking. No blinking indicates a connectivity issue that needs to be investigated.