# aigearwatch.com

> Daily price tracking for AI-capable hardware, plus sourced guides on
> building a local AI server at home. Every spec traces to a vendor page
> that was actually fetched; unpublished figures are left blank rather
> than estimated, and no throughput number is published unless measured.

Catalog last updated: 2026-08-17T02:25:20.000Z
Products tracked: 15

## Machine-readable exports

- [Catalog JSON](https://aigearwatch.com/hardware.json): full specs with inline sources
- [Catalog CSV](https://aigearwatch.com/hardware.csv): flat spec table
- [News RSS](https://aigearwatch.com/rss.xml)

## Key pages

- [Hardware](https://aigearwatch.com/hardware/): every product, sortable on memory and bandwidth
- [Compare](https://aigearwatch.com/compare/): up to four boxes side by side
- [What can it run](https://aigearwatch.com/can-it-run/): model size crossed with quantization
- [Deals](https://aigearwatch.com/deals/): computed from confirmed price history only
- [Methodology](https://aigearwatch.com/methodology/): how prices and specs are sourced

## Guides

- [Buying used AI hardware safely](https://aigearwatch.com/guides/buying-used-safely/): The used market is where the value is, and where the traps are. What to check before you pay, what protection you actually have, and what to test the day it arrives.
- [Multi-GPU: tensor parallel, layer split, and when NVLink matters](https://aigearwatch.com/guides/multi-gpu-splitting/): Two GPUs do not automatically mean twice the speed. What each split mode actually does, which one to pick, and when your interconnect becomes the bottleneck.
- [Power, cooling, and noise](https://aigearwatch.com/guides/power-cooling-and-noise/): The constraints that bite after you have already bought the hardware: what the wall can deliver, what a passive card needs, and what it will sound like in the room you sit in.
- [Q4, Q5, Q8, FP16: what quantization actually costs you](https://aigearwatch.com/guides/quantization-explained/): The numbers in every model filename. What bits-per-weight means, what you give up going lower, and which one to pick for the box you own.
- [Rig builds by budget](https://aigearwatch.com/guides/rig-builds-by-budget/): What to buy at four price points, chosen on memory capacity and bandwidth rather than on benchmark charts we cannot verify.
- [Setting up llama.cpp, Ollama, and vLLM](https://aigearwatch.com/guides/llama-cpp-ollama-vllm-setup/): Three runtimes, three different jobs. Which one to install, the commands that actually work, and the memory knobs that keep them from falling over.
- [What can you actually run at home?](https://aigearwatch.com/guides/what-can-you-run-at-home/): Memory capacity decides whether a model loads. Memory bandwidth decides how fast it answers. Everything else is a distant third.

## News

- [Build a 48GB local AI server for under $5,000](https://aigearwatch.com/blog/2026-08-16-build-a-48gb-local-ai-server-under-5000/) (2026-08-16): A beginner-safe dual-RTX 3090 build that keeps 70B-class quantized models on GPU, plus the power, PCIe, cooling, and software details most parts lists omit.

## Catalog

- [AMD Radeon RX 7900 XTX](https://aigearwatch.com/hardware/amd-rx-7900-xtx/): 24 GB, 960 GB/s, 38.75 USD/GB
- [Apple Mac mini (M4 Pro, 64GB)](https://aigearwatch.com/hardware/apple-mac-mini-m4-pro-64gb/): 64 GB, 273 GB/s, price unconfirmed
- [Apple Mac Studio (M3 Ultra, 512GB)](https://aigearwatch.com/hardware/apple-mac-studio-m3-ultra-512gb/): 512 GB, 819 GB/s, 14.26 USD/GB
- [Apple Mac Studio (M4 Max, 128GB)](https://aigearwatch.com/hardware/apple-mac-studio-m4-max-128gb/): 128 GB, 546 GB/s, 25.78 USD/GB
- [Framework Desktop (Ryzen AI Max+ 395, 128GB)](https://aigearwatch.com/hardware/framework-desktop-ryzen-ai-max-395-128gb/): 128 GB, bandwidth unpublished, 21.87 USD/GB
- [GMKtec EVO-X2 (Ryzen AI Max+ 395, 128GB)](https://aigearwatch.com/hardware/gmktec-evo-x2-128gb/): 128 GB, bandwidth unpublished, 27.34 USD/GB
- [NVIDIA A100 80GB PCIe](https://aigearwatch.com/hardware/nvidia-a100-80gb-pcie/): 80 GB, 1935 GB/s, 186.87 USD/GB
- [NVIDIA DGX Spark](https://aigearwatch.com/hardware/nvidia-dgx-spark/): 128 GB, 273 GB/s, 36.33 USD/GB
- [NVIDIA GeForce RTX 4090](https://aigearwatch.com/hardware/nvidia-rtx-4090/): 24 GB, 1008 GB/s, 140.83 USD/GB
- [NVIDIA GeForce RTX 5090](https://aigearwatch.com/hardware/nvidia-rtx-5090/): 32 GB, 1792 GB/s, 105.91 USD/GB
- [NVIDIA H100 80GB PCIe](https://aigearwatch.com/hardware/nvidia-h100-80gb-pcie/): 80 GB, 2000 GB/s, 350.00 USD/GB
- [NVIDIA Jetson AGX Orin 64GB Developer Kit](https://aigearwatch.com/hardware/nvidia-jetson-agx-orin-64gb/): 64 GB, 204.8 GB/s, 50.77 USD/GB
- [NVIDIA RTX 6000 Ada Generation](https://aigearwatch.com/hardware/nvidia-rtx-6000-ada/): 48 GB, 960 GB/s, 152.85 USD/GB
- [NVIDIA RTX PRO 6000 Blackwell Workstation Edition](https://aigearwatch.com/hardware/nvidia-rtx-pro-6000-blackwell/): 96 GB, 1792 GB/s, 150.42 USD/GB
- [Tenstorrent Blackhole p150a](https://aigearwatch.com/hardware/tenstorrent-blackhole-p150a/): 32 GB, 512 GB/s, 43.72 USD/GB