The Maverick AI Server - Water Cooled
We started this company because we believe that AI compute will migrate from "white gloved" data centers out to less formal environments. Not all computing needs high-availability (99.999% uptime) or extensive disaster recovery. At least 25% of your AI compute is probably devoted to trial-and-error, debugging, testing, or just experimentation. That compute should not be run on expensive rented machines and doesn't need to be housed in a Triple A data center.
 
Our machines are designed to be run in your office, in your basement, or in a modular data center in the middle of an oil field if you really want to optimize for low energy costs. TEP wants to build AI compute for this market segment.
 
This is our prototype of a water cooled tower that houses 6 GPUs and is designed to sit quietly (relatively) in your office. The components were choosen to maximize FLOPs per dollar. We validated the builds based on where the GPU rental market is robust and liquid, so whether you want to rent out our server or use it for your own, it is a high performing server for a good value.
 
If you want us to customize this layout for your application, please contact us. We are here to build the machines that you need.
 

Specifications:
Graphics: 6 x 24GB NVIDIA GeForce RTX 4090
Processor: Intel Xeon W7-3465X, 28C/56T, 2.5GHz - 4.8GHz
Memory: 256GB (8x32GB) DDR5 ECC 4800MHz
Storage: 2TB Samsung 980 PRO NVMe PCIe 4.0 M.2 SSD
Motherboard: ASUS PRO WS W790E-SAGE SE
Power Supply: SuperNOVA G+ 2000W Fully Modular 80+ Gold PSU x2
Chassis: Corsair Obsidian Series 1000D
Operating System: Ubuntu 22.04
 
This server is built to meet the rigorous demands of AI and HPC environments, providing the computational backbone necessary for cutting-edge research and development.