---
title: "Case airflow vs long GPU clearance: lessons from real setups"
url: https://memory.wiki/L6iIGIBW
updated: 2026-11-18T21:56:08.182Z
source: "memory.wiki"
---
# Case airflow vs long GPU clearance: lessons from real setups

## The question behind "Case airflow vs long GPU clearance"

Mock-fit the card; check PCIe power cable bend radius.  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 [device-to-model matching](https://microaione.com/match/) 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.

## Questions to ask before you buy

- Will the model you need still fit after the operating system, drivers and runtime take their share?
- How long are the documents people really paste, and how many people paste them at the same time?
- Does the room tolerate the noise and heat of a sustained load during working hours?
- Who replaces a failed fan or PSU, and how long can the team live without the box?
- Is there a clear upgrade path, or will the next step require a new platform?

If any answer is unclear, rent or borrow time on similar hardware first. A week of real use teaches more than a month of reading spec sheets, and it keeps your purchase grounded in your own workload.

See also [5000 CNY budget tier](https://microaione.com/hardware/budget-5000/).

## Quick reminders

- Cap concurrent sessions before lowering quantisation.
- Use one models directory shared by every runtime.
- Test a restore, not just a backup.
- Write the date and owner on every change record.

*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
Effective local LLM hardware selection requires defining the specific workload and memory requirements before choosing a platform. Prioritize sufficient VRAM headroom and system stability over individual GPU specs to ensure consistent performance.

## Themes
- Local LLM hardware planning
- System integration strategies
- Memory budget management
- Workload-first procurement

## Key takeaways
- Define the workload including model families, context length, and concurrent users before selecting hardware.
- Calculate VRAM requirements based on free memory after a cold boot to ensure headroom for context and runtime buffers.
- Prioritize platform stability by ensuring PSU capacity handles transient spikes and PCIe slots support future expansion.
- Verify physical clearance for GPU length and power cable bend radius before finalizing case selection.
- Test hardware with a burn-in period and a custom evaluation set before declaring a build production-ready.

## Insights
- Hardware selection should be driven by specific model requirements and concurrent user load rather than GPU marketing tiers.
- System memory needs must account for OS overhead, runtime buffers, and KV cache growth rather than just raw model weight size.
- Physical constraints like cable bend radius and thermal airflow are as critical to system stability as raw compute power.

## Open questions / gaps
- What specific metrics define a successful burn-in test beyond temperature and log monitoring?

