---
title: "24GB build motherboard and PSU notes: a short field guide"
url: https://memory.wiki/A5Qk134M
updated: 2026-11-20T21:59:31.903Z
source: "memory.wiki"
---
# 24GB build motherboard and PSU notes: a short field guide

## The question behind "24GB build motherboard and PSU notes"

Do not bottleneck a good GPU with a weak platform.  Most hardware threads about local LLMs jump straight to a card model. A better starting point is the workload: which model families you actually need, the quantisation you can live with, the context length your users paste, and how many people will hit the box at once. Only then does a shopping list make sense.

## Work out the memory budget first

Model weights are only part of the picture. The KV cache grows with context length and concurrent sessions, the runtime keeps its own buffers, and the desktop or display server may already be holding a slice of VRAM before you load anything. Write down free memory after a cold boot, not the number on the box. If the model plus a realistic context does not fit with headroom, it will spill to system RAM and feel slow in a way that no setting fixes. A [hardware reviews](https://microaione.com/hardware/) is useful here because it lets you reason in parameter counts and quant levels instead of marketing tiers.

## Platform matters as much as the card

A good GPU on a weak platform is a common regret. Check:

- PSU capacity and quality, with real headroom for transient spikes, not just the nominal TDP.
- Physical clearance: card length, thickness, and the bend radius of the power cable.
- PCIe slot wiring if you plan to add a second card later.
- Airflow: intake and exhaust paths that still work when the side panel is on and the room is warm.
- System RAM: enough to hold the OS, the inference server, and an embedding model without swapping.

## New, used, or unified memory

Used cards can be good value, but inspect them: missing screws, replaced fans, discoloured connectors and odd thermals under a short load test are reasons to walk away. Apple Silicon machines take a different route: unified memory means the model and the system share one pool, so leave clear headroom for macOS and anything else you run. Mini PCs are fine for small models and kiosks if you accept their thermal limits. Whatever you choose, record serial numbers, invoices and photos on install day.

## A sane buying sequence

1. Write the workload in one paragraph: models, context, users, hours.
2. Map it to memory needs with headroom.
3. Choose the platform (PSU, board, case, RAM), then the GPU.
4. Burn in for at least half an hour while watching temperatures and logs.
5. Run your own small eval set before declaring the build done.

## A worked example without fake numbers

Imagine a five-person office that wants a private assistant for drafting and document search. They write down the model family they tested, the quantisation that still answered Chinese and English questions well, and the longest documents they paste. From that they estimate memory with a comfortable margin and decide that one card with enough VRAM is safer than two smaller cards that complicate the setup. They choose a quality PSU with headroom, a case with clear front intake, and enough system RAM to run an embedding model alongside the main one. After assembly they run a burn-in, check temperatures and logs, and only then run their evaluation prompts. Every figure they share afterwards comes from their own measurements, with the method written next to it.

See also [custom local-AI builds](https://microaione.com/builds/).

## Quick reminders

- Test a restore, not just a backup.
- Photograph cable routing before closing the case.
- Use one models directory shared by every runtime.
- Log the model name and version with every request.

*Published by [MicroAI One (本地AI 1号网)](https://microaione.com/), a Shenzhen team working on local LLM hardware, software and deployment.*

---

## Summary
To build an effective local LLM system, define your specific workload and memory requirements before selecting hardware components. Prioritize a robust platform with sufficient power, cooling, and system RAM to ensure the GPU performs optimally without bottlenecks.

## Themes
- LLM hardware planning
- system architecture balance
- VRAM capacity management
- workload-first deployment

## Key takeaways
- Workload requirements including model family, quantisation, and context length must be defined before selecting hardware.
- PSU capacity must account for transient power spikes beyond nominal TDP ratings.
- System RAM must be sufficient to host the OS, inference server, and embedding models to prevent swapping.
- A burn-in period and evaluation set are necessary to validate system stability and performance before deployment.
- Physical constraints such as cable bend radius, card thickness, and airflow paths are critical for long-term reliability.

## Insights
- System performance is dictated by the total memory budget including KV cache and OS overhead rather than just model weight size.
- A high-end GPU is often bottlenecked by inadequate power delivery or poor thermal management in the host platform.
- Unified memory architectures like Apple Silicon require different capacity planning compared to discrete GPU setups due to shared resource pools.

## Open questions / gaps
- What specific metrics or tools are recommended for measuring the KV cache memory footprint during runtime?

