MarketAIVerse the universe of AI
guide · old hardware

You downloaded the model.
Now what?

Twenty gigabytes on disk, a card from 2017, and a VRAM calculator that told you it is not possible. It is. And the hard part is not starting it — it is keeping it running.

€25 I want it First see what you can run
An old graphics card covered in dust, on a workbench
The hardware we are talking about. Not a data centre — the card in your drawer.

Why another guide

Guides exist. We read them. They all stop at the same place: „it loaded, it works". And almost all of them are written on 24 GB cards — that is, on hardware where anything runs anyway.

What they do not cover is what happens after six hours. When speed drops and you do not know why. When the cache is thrown away on every message. When a second person joins and everything halves. When the optimisation everybody recommends makes it worse.

The eleven chapters

Seven pages, no filler. Every chapter carries at least one measured number, with the machine and the date next to it. MEASURED  Every figure in the list below is our own measurement, taken between 24 August and 12 September 2026 on the two machines described in the guide.

Their guides teach you to start it. This one teaches you to keep it running.

Proof that we know what we are talking about

Qwen3.6 35B-A3B on a GTX 1080 Ti (11 GB, from 2017)
7 hours straight, 500+ real answers: 23.86 → 23.73 tok/s
23.7 tok/s
MEASURED
Same model, no graphics card at all
And after 31.8 hours of uptime it had fallen to 2.05 — which is why the guide has a chapter on restarts
2.5 tok/s
MEASURED
These numbers are not derived from spec-sheet formulas. They happened, on our machines, on the dates written. Check what comes out for your hardware.

How to get it

the guide

The guide and the files

The full text, the measurement tables and the ready-made commands.

€25COMING
PDF, 8 pages, in English.
service

We set it up on your machine

We connect, choose, tune, and you end up with something that starts by itself.

€200COMING