Insights &amp; Expert Guidance

Create an Ubuntu VM with KVM, virt-install, and cloud-init (2026)

Create a reusable Ubuntu 24.04 KVM guest from a cloud image with virt-install, cloud-init, private networking, QEMU guest agent, SSH, and a serial console.

Install KVM on Ubuntu: Complete libvirt Host Setup (2026)

Install KVM, QEMU, libvirt, virsh, and virt-install on Ubuntu, then configure permissions, private networking, and validate the finished virtualization host.

KVM Networking: Custom NAT Network with Static DHCP

Create a persistent libvirt NAT network, reserve a predictable DHCP address, attach a second virtio NIC, configure Netplan, and control route priority.

KVM Storage Pools and Volumes: Attach and Expand a Disk

Create a libvirt storage pool and qcow2 volume, attach it to a running KVM guest, format it, then expand the active disk and ext4 filesystem safely.

KVM Virtualization Series: Build and Manage a Linux Hypervisor Lab

Follow the complete InventiveHQ KVM series from host installation through virtual machines, storage, networking, backups, performance tuning, GPU passthrough, and migration.

Virsh Commands: Complete KVM & libvirt Cheat Sheet (2026)

Use virsh to list, inspect, start, stop, configure, and safely remove KVM virtual machines. Includes tested local and remote libvirt commands.

Technology

Why Is the Keyboard QWERTY? The Real History Behind the Myth

The story that QWERTY was designed to slow typists down is one of tech's most repeated myths. The real history involves typebars, telegraph operators, and a century of path dependence.

2026-07-02Read →

How To

Why iDRAC6 Virtual Media Is Broken on Modern Macs (And What to Do Instead)

The Dell R610's iDRAC6 virtual media requires Java 6-era native libraries that cannot run on Apple Silicon or 64-bit-only Intel Macs. Here is why every workaround fails and how PXE boot sidesteps the whole problem.

2026-06-30Read →

How To

PXE Boot Ubuntu on a Dell R610 Using a MacBook as the TFTP/HTTP Server

Turn your MacBook into a PXE boot server to install Ubuntu on old server hardware that refuses to USB boot. A complete walkthrough using macOS TFTP, Apache, and a Ubiquiti Dream Machine for DHCP.

2026-06-30Read →

The Context-Length Tax: What Going 2K to 32K Actually Costs

Going from 2K to 32K context cost essentially 0 tok/s but +1.7 GB of VRAM. We swept Qwen2.5-Coder-7B across five context sizes — the tax is memory, not speed.

More CPU Threads Made My LLM Slower: A Thread-Scaling Test

Throughput peaked near the 6 physical cores on an i7-8700, and 12 threads ran slower than 8. Here's why memory bandwidth — not core count — decides how fast a CPU runs a local model.

Flash Attention in llama.cpp: -fa Is Free Because It's Already On

We swept llama.cpp's -fa flag across 4K, 16K, and 32K context on an RTX 5060 Ti. Speed and VRAM were identical on and off — because this build already defaults flash attention on.

How Low Can You Quantize a GGUF Model Before Quality Breaks?

We swept Qwen2.5-Coder-7B from Q2 to Q8 on an RTX 5060 Ti. Output fidelity goes 21% at Q2, 57% at Q4, 80% at Q6 — the cliff is below Q4, and the sweet spot is Q4–Q5.

I Capped My GPU to 150W and Barely Lost Any Speed

We cut an RTX 5060 Ti's power limit by 17% and lost essentially zero tokens per second while gaining 16% efficiency. Local LLM decoding is memory-bandwidth-bound, not compute-bound — so watts are mostly wasted.

KV-Cache Quantization: The q4_0 Cliff Your Logs Won't Warn You About

We benchmarked f16 vs q8_0 vs q4_0 KV caches on the same model. q8_0 KV is nearly lossless (81.6% similar). q4_0 KV wrecks output quality (8.3%) while saving VRAM — and your throughput dashboard stays green the whole time.

Local LLM Benchmarks: 12 Findings From Consumer Hardware

NVIDIA publishes its inference wins on 8× DGX B300s. Almost nobody has those. So we ran the same ideas on a single consumer RTX 5060 Ti, a 2017 GTX 1080 Ti, and a CPU — and measured what actually happens. Here are all 12 experiments.

MoE on CPU: 13B-Class Answers at 3B Speed

A 30B-A3B Mixture-of-Experts model runs at dense-3B speed on an 8-year-old i7 CPU (10.6 tok/s) yet scores 8/10 on our graded set versus the 3B's 1/10. Here's why, and how to size your own box for it.

Ollama vs llama.cpp vs LM Studio: The Speed Tax, Measured

LM Studio, Ollama, and raw llama.cpp all run the same engine on the same GPU. We measured what the convenience layer costs: LM Studio adds 0.3%, Ollama adds 10%.

Artificial Intelligence

One Models Folder to Rule Them All: Stop Duplicate AI Models From Eating Your Disk

Ollama, LM Studio, EasyDiffusion, and every other local AI tool downloads its own private copy of every model — so the same weights live on your disk three and four times over. Point them all at one shared models folder on a second drive and reclaim the space. Here's exactly how, on Mac and Windows.

How Small Can a Local LLM Get Before It Can't Reason?

We swept Qwen2.5-Coder from 0.5B to 14B on one GPU. Graded pass-rate climbs 1 to 6 of 10 — but you need about 7B before a model can actually chain reasoning steps.

Bigger Draft Model = Faster? A Speculative Decoding Sweep

We swept 0.5B → 3B draft models against a fixed 14B target. The 0.5B won at 1.37× — despite the lowest acceptance rate of the three. Here's why bigger drafts lose.

The VRAM Cliff: 15× Slower the Moment Layers Spill to CPU

We swept -ngl from 0 to 99 on a 14B model: 2.89 → 43 tok/s as it moves onto the GPU. Partial offload is a cliff, not a slope — and the last 8 layers matter most.