Hardware / two extremes

Machines I've built

From a dual-GPU AI workstation made to run heavy local AI, to a computer smaller than a credit card — built, not bought.

Build 01 · workstation

A dual-GPU box for local AI

I wanted to run large AI models locally — the kind that power Raven — without paying cloud bills forever. The trick is VRAM. So I built an AM4 machine that pairs a normal gaming GPU with a second-hand NVIDIA Tesla P40, a datacentre card that gives 24GB of VRAM for a fraction of the price of a modern equivalent. The result is a quiet, single-tower workstation with serious memory for AI, built mostly from carefully-chosen used parts.

INTAKERyzen 9 5950X64GB DDR4NVMe 2TBGPU 116GB VRAMGPU 2 · passive datacentre cardTESLA P40 · 24GB1000W · 80+ PSU
Up to 40GB VRAM 16 cores / 32 threads AM4 platform Runs Raven fully local Single air-cooled tower
The numbers

Built around four numbers.

01
ProcessorAMD Ryzen 9 5950X · AM4
0cores
Why

Raven runs many things at once — scraping, processing, models and automation in parallel. More cores means none of it has to wait its turn.

How I use it

orchestrating Raven's workloads and crunching large batches of scraped business data at speed.

02
Memory64GB DDR4
0GB
Why

Big datasets and several model contexts need to sit in memory together, so nothing grinds to a halt halfway through a run.

How I use it

holding large scraped datasets and multiple live tasks in memory at the same time.

03
VRAM16GB GPU + 24GB Tesla P40
0GB
Why

VRAM is the hard ceiling on which AI models you can run locally — more of it means bigger, smarter models actually fit on the card.

How I use it

running the large local language models that power Raven, fully offline and free.

04
Storage2TB NVMe + 5TB SATA
0TB
Why

Model files are gigabytes each and scraped data piles up fast; the NVMe loads models instantly, the big SATA drive holds everything else.

How I use it

keeping a whole library of models plus datasets and project files on hand.

Full build sheet
Component breakdown
Processor
AMD Ryzen 9 5950X16-core / 32-thread, AM4. Also strong here: the 5900X, or the cracking 5900XT deal. 5900XT deal ↗
CPU cooler
Thermalright Peerless Assassin 120 SEDual-tower air cooler — near-liquid performance, silent, no pump to fail. source ↗
Motherboard
MSI MAG X570 Tomahawk WiFiA rock-solid AM4 board with the power delivery to feed a 16-core chip. Step-up: X570S Tomahawk MAX.
Memory
64GB DDR4Enough headroom to load big models into RAM alongside the GPUs. Used kits keep this affordable.
GPU 1 · primary
NVIDIA RTX 3070 10GBLeaned NVIDIA on purpose — better software and cleaner integration alongside the P40. Alternatives: RX 6800 / 6800XT (16GB).
GPU 2 · AI VRAM
NVIDIA Tesla P40 · 24GBThe whole point of the build: cheap datacentre VRAM. 24GB for the price of a budget gaming card.
Storage
2TB NVMe + 5TB SATAFast NVMe for the OS and models, big SATA drive for everything else. Value tier: 1TB NVMe + 2TB SATA.
Power supply
1000W, 80+ ratedReal headroom to drive two GPUs cleanly, from a reputable unit.
Case
Corsair 4000D AirflowMesh-front mid-tower built for airflow — exactly what a hot dual-GPU box needs. source ↗
P40 cooling · important
Bolt-on fan shroudThe Tesla P40 is passively cooled by design (it expects datacentre airflow), so it needs a dedicated fan strapped to it. Skip this and it cooks.

Why the P40

A modern 24GB card costs more than this whole machine. The P40 gives the same VRAM for AI work for around £200 — the single decision the build is designed around.

Why AM4

AM4 is mature and cheap on the used market, yet a 16-core 5950X is still seriously fast. More cores and VRAM per pound than chasing the newest platform.

Why local

Everything runs on my own machine — no API keys, no usage limits, no monthly bill. It is the same principle that makes Raven free to run forever.

Build 02 · micro

…and a computer smaller than a card

At the other extreme: a full Linux computer built around a Raspberry Pi Zero 2 W and a microSD card, for under £50 all-in. A quad-core machine with wireless built in, small enough to disappear into a project. Proof that the interesting part isn't the budget — it's knowing what to do with it at either end of the scale.

RP3A0 HDMI USB SD WiFi/BT
Under £50 all-in Quad-core WiFi + Bluetooth Runs headless Linux