Local Ai Home Server Build at Super Low $150 Budget Price

When you click on links to various merchants on this site and make a purchase, this can result in this site earning a commission. Affiliate programs and affiliations include, but are not limited to, the eBay Partner Network. As an Amazon Associate I earn from qualifying purchases. #ad #promotions

The Dell 7050 Mid Tower is the cheapest dedicated way to get started in both homelab and locally hosted Ai that I can put together that is running off a very common system. It is also very decent in overall wattage efficiency. Based off the Dell Mid Tower 7050 and either 2x bus powered GPUs Nvidia M2000 4GB gpus or a single Nvidia Tesla P4 GPU with cooling fan. It provides an effective way to carve up resources if you run multiple small LLMs or other services on the same machine by using LXC and or Docker for your GPU intensive workloads.

Dell 7050 Ai Server SPECS

PRICE CATEGORY: $150

VRAM: 8

PRICE/GB/VRAM: $20.63

SETUP REVIEW VIDEO: https://youtu.be/VV30CMHc-kY

GPU ALTERNATIVES: A Single Tesla P4 is a great cheap option also but does require an additional cooling shroud and fan. One P4 comes out around the price of 2 M2000’s. The temps are very stable once you have that provided but will instantly thermally throttle without the forced air cooling. The shroud and fan is usually a $25 addition.

 

ITEM DESC QTY PRICE ($) SUBTOTAL ($) Link
CHASSIS Dell 7050 Mid Tower 1 75 75 https://geni.us/OptiPlex7050MT
RAM included
MOBO included
CPU included
CPU COOLER included
GPU M2000 2 45 90 https://geni.us/Quadro_M2000
PSU included
NVMe/SSD included
ACCY na
TOTAL 165

Review Video

$150 Ai Server Tips and Tricks

  • For the CPU I have been able to use an i7-7700K without issues with the stock cooler. Even under intense processing it does not overheat as the rear fan near it can pull a lot of air out. I do not have a 5.25″ bay insert in mine but something to facilitate airflow like a 6x 2.5″ ssd caddy enclosure would be a cool potential.
  • These cases can look pretty cool with some paint and modifications.
  • The Tesla P4 option is what I am using in mine now and it is all around better. I do plan to drill or cut out the intrusive 2.5″ cages at a minimum. They are really in the way. Will allow for me to mount a larger fan off the front.
  • The internal bays take a lot of space up for caddy holders/cage holders. Have seen folks cut that out and have a lot better cooling.
  • Laying it on its side works well for airflow but drilling holes in the base bottom to provide “intake” is one option that might be nice to see and provide for inlet air to cool PCIe cards directly would mean setting it up.
  • It is a small footprint machine even in the Mid Tower design.
  • The idle states are very good and you can disable USB if not needed to save up to an extra few watts.
  • I suspect a PSU swap out can be made to allow for a larger PSU/better GPU support is doable with an adapter and moderate grinding.

Local LLM Performance Benchmarks

Mini-CPM Vision 8B – Q4 – 8192 ctx

Dell Optiplex 7050 (4.5 t/s, CPU only) https://geni.us/OptiPlex7050MT

Quadro M2000 4GB (11 t/s, better option) https://geni.us/Quadro_M2000

Quadro K2200 4GB (7 t/s, get M2000 instead)

Quadro P2000 5GB (20 t/s, or follow the 350 build👇) https://geni.us/Quadro_P2000

Tesla P4 8GB ( 28 t/s, needs extra cooling fan and shroud https://geni.us/TeslaP4
JOIN AS A YOUTUBE MEMBER

Table of Contents

Hardware Review

Local Ai Setup Guide