Local AI Inference GPU Economics
This repo is primarily a CI/CD and agentic deployment testbed. The main goal is to automate deployment and test agentic workflows — letting AI agents deploy software themselves and periodically verify their work online. The GPU comparison tools are the deployed content that agents use to validate end-to-end automation.
Two static tools for GPU research and buying decisions (used as deployment targets):
| Tool | What it does |
|---|---|
| Full Comparison Table | Search, filter, and sort 60+ GPUs. Compare VRAM, bandwidth, TFLOPS, prices, and LLM token speeds. Add up to 6 GPUs to a side‑by‑side compare view. |
| Price vs Bandwidth & VRAM | Interactive scatter charts. See which GPUs offer the best value for bandwidth or VRAM. Click the legend to filter by vendor. |
Coverage: NVIDIA RTX 30/40/50, data center (T4 → B200), DGX Spark, AMD ROCm, Intel Arc, plus cloud rental and used/refurb pricing.
- Consumer (used): Marktplaats.nl
- Cloud rental: Modal.com, Lambda Labs
- Purchase (data center): eBay used/refurbished
- DGX Spark: NVIDIA
Edit gpu-data.js directly — it's the single source of truth.
- Stack: Static HTML, vanilla JS, Chart.js, Google Fonts (DM Sans)
- Hosting: Works on Vercel, Netlify, GitHub Pages — no build step
- Config:
vercel.jsonfor clean URLs and trailing slash handling
© 2026. All Rights Reserved. No part of this software may be copied, reproduced, or distributed without permission.