Improving LLM prefill on the NVIDIA GeForce RTX 4060 mobile

The Acer Nitro V15 has a discrete nvidia GeForxe RTX 4060 mobile. Turns out, the default firmware leaves some performance untapped [GPU].1. follow the project instructions [GPU] to download, build and install.2. specifically, the 4060m requires a kernel parameter in grub: "gpu_unlock.targets=0000:01:00.0 gpu_unlock.ad107_experimental=1"You'll have to find the exact address with lspci. Yes, it needs the … Continue reading Improving LLM prefill on the NVIDIA GeForce RTX 4060 mobile →

Beating ChatGPT 5 on a laptop PC

TL;DR: Qwen3.5 31B at 15 t/s generation (500 t/s prefill) on 8GB VRAM as a daily driver, not a proof of concept. Context & motivation My crusade to democratise affordable AI is in full swing, whether that means patching inference engines[COL] or pushing cheap hardware to its limits[XLM]. I’m glad to see developments in software … Continue reading Beating ChatGPT 5 on a laptop PC →

Headless boot HP z620

Just a quick note: the HP z620 supports booting without a graphics card, but it's hidden. Steps to enable: enter the BIOS -> replicated setup -> save to a FAT32 formatted USB boot some computer, attach the USB, edit the file: there's a "Headless Mode" section with two options, "Enabled" and "Disabled" and a silly … Continue reading Headless boot HP z620 →

Auditing VRAM: Ensuring Every Gigabyte is Real

Counterfeit storage devices, such as USB flash drives and SSDs with modified controller firmware that report false capacities, rely on address wrapping: writes beyond physical memory silently overwrite earlier blocks modulo the actual capacity. With increased interest in home AI labs, one could imagine similar practices to start appearing in the second-hand GPU market. While … Continue reading Auditing VRAM: Ensuring Every Gigabyte is Real →

VScode copilot: No lowest priority node found

I've found the vscode copilot error "No lowest priority node found" to be about a context overflow somewhere in the chain "inference engine" -> "LLM" -> copilot. Some/all of the workarounds you can try when that happens: Reduce/disable agent tools Increase the input window in chatLanguageModels.json, eg. { "id": "gemma-4-12b-it-qat-q4_0", "name": "gemma-4-12b-it-qat-q4_0", "url": "http://localhost:11434", "toolCalling": … Continue reading VScode copilot: No lowest priority node found →