Tag: Hardware

Qwen 3.6 27B is finally a smart model we can use for coding on Macbook or Nvidia RTX - with llama.cpp and OpenCode.

I built a 10-inch mini rack from aluminium extrusions and I had a lot of fun doing it. I want to share my build in this post.

We've all heard people say that Qwen is near-Opus level, but I have receipts and am here to be transparent with you.

Since the PiKVM came out in 2017, there's been an explosion of IP KVMs. I've tested almost every one. But what are they good for?
You can use Remote Desktop, Screen Sharing, or VNC to remote control a computer from anywhere on a LAN. And if you don't have a private VPN, you could use RealVNC, Raspberry Pi Connect, or wire up Tailscale or Pangolin for fully remote access. Those solutions are great, and so is SSH if you don't need a full desktop.

Or running Gemma 4 on a 2016 Xeon with no GPU, 25 flags, 128 GB of DDR3, and a 25B-parameter MoE.

Do you want some inspiration for your ultimate home office desk setup? How about smart lighting, smart blinds and more? Look no further than my 2026 guide!

This is follow up to my previous blog post Deploying VCF 9 in a Minilab – I have made some changes in my lab and upgraded to VCF 9.1. My original HW BOM: Two GMKtec K8 Plus mini PCs. One with…

How to set up Intel Arc SR-IOV on Proxmox VE with three ASUS NUC 15 Pro units for Plex hardware transcoding, including notes on kernel compatibility and the required Plex beta release.

Happy Friday! 🥳 A couple of weeks back, Wenchao (creator of the Realtek Network Driver for ESX Fling) reached out to me to share an exciting development he had been working on. Unlike traditional e…

To track CPU Support in VCF releases, we have a dedicated KB article that shows which CPU are either Deprecated or Discontinued. See CPU Support Deprecation and Discontinuation In VCF Releases R…

I had been troubleshooting a stubborn CPU utilization issue with a workload that, over time would also overrun the CPU on my physical ESX host. The assumption was that the workload was causing the …

For years, the best way to get 10 gigabit networking on laptops was to buy an expensive, large, and hot 10 GbE Thunderbolt adapter. With new RTL8159-based 10G USB 3.2 adapters coming onto the market, the bulky adapters might be a thing of the past. Just look at the size of the thing in comparison to my Thunderbolt adapters:
2.5G and even 5G USB adapters have been out for a while, but sometimes you need more bandwidth.

How the maintainer of the popular For You feed serves it from their living room!

How to move fast when building for the physical world

I have been running the full VMware Cloud Foundation (VCF) 9.0 stack using the Minisforum MS-A2 in a three-node configuration for almost a year now. While the MS-A2 is not hardware that you would f…

I've had my Framework 13 for a year now, but am I still enjoying it?

Why VMware VKS Is a Stronger Enterprise Choice Than KubeVirt | vmtechie.blog KubeVirt is a capable open-source project and a legitimate choice in the right context. But when the workload is enterpr…

Cloudflare’s Gen 13 servers double our compute throughput by rethinking the balance between cache and cores. Moving to high-core-count AMD EPYC ™ Turin CPUs, we traded large L3 cache for raw compute density. By running our new Rust-based FL2 stack, we completely mitigated the latency penalty to unlock twice the performance.

A guide on building a simple Linux distribution from scratch. Detailed guide on building the kernel and the init process. Finally, a little distribution is built with u-root that is capable of connecting to the Internet.

Disaggregating Prefill and Decode: Faster First Tokens, Faster Streams

Découvrez comment remplacer votre box Internet SFR, Free, Bouygues ou Orange par du matériel UniFi. Guide complet opérateur par opérateur.

The Intel 285K CPU in my high-end 2025 Linux PC died again! 😡 Notably, this was the replacement CPU for the original 285K that died in March, and after reading through the reviews of Intel CPUs on my electronics store of choice, many of which (!) mention CPU replacements, I am getting the impression that Intel’s current CPUs just are not stable 😞. Therefore, I am giving up on Intel for the coming years and have bought an AMD Ryzen 9950X3D CPU instead.

Sourcing data directly from diskISfaster than caching in memory. I brought receipts.
Because hardware got wider but not faster, the old methods don't get you there. You need new tools to use what is

Hardware got wider, not faster. More cores, more bandwidth, huge vector units — but clocks, IPC, and latency flatlined. Old rules like “memory is faster than disk” are breaking. To go fast today, you

Datacenter-Scale Heat Management

After Claude Pro changed to weekly limits, I explored self-hosting Qwen3-Coder-480B with 400k context windows. Here's what I learned about costs, alternatives, and why Claude Code still dominates the landscape.

Miniature rack builds, for portable or compact Homelabs.

The cloud you own. Hardware, with the software baked in, for running infrastructure at scale.

We finished pulling seven cloud apps, including HEY, out of AWS and onto our own hardware last summer. But it took until the end of that year for all the long-term contract commitments to end, so 2024 has been the first clean year of savings, and we've been pleasantly surprised that they've been even better than originally estimated. F...

macOS system monitor in your menu bar. Contribute to exelban/stats development by creating an account on GitHub.

I spent a month repeatedly building my website in Docker, and now have horrors to share.

Since launching the MS-01 in 2024, Minisforum has steadily gained popularity for its unique design that sets it apart from established players in the small form factor (SFF) market. Following the s…

Having the latest compatible software for one’s NVIDIA vGPU investment should be a priority. New features are added all the time, or there may be security enhancements. In this post I’ll demonstrate a new PowerShell module I’ve built that employs the NVIDIA License System (NLS) API to be able to list and download vGPU software like drivers or the NVIDIA Delegated License Service (DLS).

When you read my blog articles and stuff – you may get the idea that everything I do – just happens to be right and that I succeed at every attempt. This article is here to remind you t…

Burstable VMs run on a fraction of CPU and burst to a higher level of CPU usage to support occasional usage spikes. To implement them, we leveraged Control Groups v2 (cgroups v2), a Linux kernel feature that helps manage resource usage. We thought our open-source implementation of burstable VMs might be interesting enough to write about. We also learned a lot about Linux cgroups in the process!

I'm fortunate enough to live in a place where 10Gbps fiber (FTTH) is not only available but also cheap. Here's how I'm taking advantage of this.

Take an interactive journey through the history of IO devices, and learn how IO device latency affects performance.

When it comes to infrastructure engineering, building a data center is probably closer to building a house than to deploying a Terraform stack.