Servidor de IA recondicionado com 64 GB de VRAM

Refurbished AI server with 64 GB of VRAM

VRAM determines the model. Dense Qwen 27B–32B, quantized Llama 70B, Mixtral 8×7B — these require 32 to 64 GB, not the 24 GB of a 4090. Engineer Donato Capitella, with Bargain Hardware, built a refurbished Supermicro with four data center GPUs (64 GB total) for around £2,000 and benchmarked V100, P100 … Read more

ova2xva: migre VMs do VMware/VirtualBox para XCP-ng com um comando

ova2xva: migrate VMs from VMware/VirtualBox to XCP-ng with one command

Moving a VM from VirtualBox or VMware to XCP-ng (or XenServer) usually becomes a ritual: export OVA, convert the disk by hand, and wrestle with xe. ova2xva is a Go CLI (MIT) that handles the trip in both directions in a single binary — converting an OVA or standalone disk … Read more

llama-model: gerencie modelos GGUF e o llama-server sem o daemon do Ollama

llama-model: manage GGUF models and llama-server without the Ollama daemon

If you run AI models locally with llama.cpp / llama-server, you know the pain: download GGUF by hand, edit the configuration, restart the service, and hope the model starts. llama-model solves this — a Go CLI, one static binary, that manages your model collection and the llama-server service without needing the daemon from … Read more

Monitorando Servidores Linux com Prometheus e Node Exporter

Monitoring Linux Servers with Prometheus and Node Exporter

A server without monitoring only tells you it is down when someone complains. With Node Exporter, each host exposes Linux metrics; Prometheus collects and stores them as time series for queries and alerts. This guide uses Ubuntu packages where available, avoids exposing metrics to the internet, and validates each step before continuing. The … Read more