aicomputer

This is a guide on how you can build a low-cost (<1,300€) AI computer by yourself to run your AI flows locally.

Ideal for students on a low budget. The build can, of course, also be expanded and constructed in a way to offer the best performance for enterprises, for example.

Table of Contents

• Parts List
• Assembly Guide
• Installation
• Remote Access Setup

Parts List

Component	Model
GPU	RTX 3090 FE (used)
CPU	Intel i5 12400F
Motherboard	Asus PRIME B660M-K D4
RAM	Corsair Vengeance LPX 32GB (2x16GB)
Storage	WD Green SN3000 500GB NVMe
PSU	MSI MAG A750GL PCIE5 750W
CPU Cooler	ARCTIC Freezer 36
Case Fans	4x ARCTIC P12 PWM
Case	ASUS Prime AP201 MicroATX

Note: Links are to German retailers but parts are available worldwide

Assembly Guide

1. Unpack all components
2. Install CPU on motherboard
3. Mount CPU cooler
4. Install RAM modules
5. Install M.2 SSD
6. Mount motherboard in case
7. Connect front panel cables
8. Install power supply
9. Install exhaust fan:

   

10. Install GPU
11. Close case

Installation

1. Connect:
   • Ethernet cable to router
   • Keyboard to motherboard USB
   • Monitor to GPU (not motherboard)
2. Download Ubuntu Server
3. Create bootable USB
4. Insert USB into motherboard port
5. Power on (PSU switch to 'I')
6. Enter BIOS/boot menu:

   

7. Select USB boot device:

   

8. Complete Ubuntu installation with SSH server enabled

Remote Access Setup

After installation, you can SSH from any device on your network:

ssh username@192.168.2.x  # Replace with your machine's IP

How to get the IP

1. Connect the monitor along with the mouse and keyboard.

2. Input the Wi-Fi password via the GUI and connect to Wi-Fi.

3. Run ip a in the terminal and find the IP (it will look like 192.168.x.x or 10.0.x.x).

4. From my end, run ssh username@thisip

Usage

Ready to use with GPU accelerated models from Ollama and other AI frameworks.

Need Help?

Reach out to me on X for build assistance or questions!