HOSTKEY Company blog

Mumble is simple, fast, and reliable for small groups. Here’s how to deploy it on Ubuntu with Docker and keep it backed up.

Testing the RTX PRO 2000 (70W power consumption, 16GB GDDR7 memory) with Ollama, ComfyUI, and Blender. We’ve checked what this card is actually capable of and whether it’s worth the price.

Is your service experiencing performance drops during peak periods? Hybrid architecture helps stabilize loads and avoid unnecessary expenses. Find out when this approach works best.

Learn when it’s time to switch to a dedicated server and what benefits you’ll reap from doing so. We provide data-driven insights and an overview of the technical aspects involved in this transition.

How did we create our LLM benchmark for GPU servers using Ollama? We developed a script, tested it with DeepSeek R1, and configured the necessary contexts. We identified some patterns and compared the performance of different GPUs, all of which are now available on GitHub.