How’s My New (Used) Linux Box Doing?
In a word: terrific.
I picked up a 2-year-old, Mac-mini-sized PC with a Ryzen 7 (8 cores / 16 threads), 64 GB RAM, and a 2 TB NVMe drive—about the same price as a Mac mini. I’ve since moved all of my hosted VPS activities onto this local Linux box.
I continue to be very pleased with Linux Mint. For remote desktop, I’m using NoMachine NX, which is significantly better than VNC. I’m usually running Linux on much older hardware, so it’s genuinely enjoyable to have a machine that feels this responsive under Linux.
The primary workload is Docker containers, and unlike Windows, Docker is Linux-native—so all 64 GB of RAM is available. On Windows, you have to pre-allocate memory specifically for Docker, which always felt a bit clunky.
This machine was running Windows 11 originally, but it had started rebooting every night—not tied to Windows Update, just… something else. It reached the point where I was preparing to send it back for repair (again—different issue than the first time).
Linux has now been running for days without a hiccup. Whatever the problem was, it appears to have been a Windows-plus-hardware issue that disappeared with Linux on the same hardware. That’s a nice outcome.
I’ve been doing my AI development on this box instead of the Mac mini. Since the LLMs aren’t running locally anyway, there’s very little difference between the two in practice—Claude Code, Gemini CLI, Antigravity, VS Code, all feel the same. The Mac mini is still a wonderful machine, but 64 GB of RAM on the Linux box is far more comfortable than 16 GB on the Mac.
I originally thought the Mac’s neural engine would be useful for AI work, but none of the local model tooling I’m using actually takes advantage of it. They do use the GPU, and I’m sure local models would run much faster on the M4 Mac—but that’s not my current workflow.
One notable shift: I’m now SSH’ing into this Linux box regularly, which I hadn’t done much before. Since it’s replacing my hosted VPS, interacting with it from the command line just feels natural.
A funny detail: I use Warp Terminal on my MacBook Air to SSH into the Linux machine. Warp is also installed on the Linux box. When I SSH from the Mac, Warp’s AI features are still running locally on the Mac—not on the Linux machine. I discovered this when I kicked off a long-running task and then closed the laptop… which promptly suspended the process.
Lesson learned. For long-running jobs, I now remote into the Linux desktop and run Warp there.
I briefly considered Omarchy, but NoMachine doesn’t play well with that window manager, and I’m simply far more fluent with Linux Mint.
Bottom line:
Claude Code, Gemini CLI, OpenCode, VS Code, Antigravity, Docker—everything works beautifully on Linux.
