So I was sitting at my desk yesterday, trying to run a small local LLM for a side project, and my "pro" laptop—you know, the one that cost me a month's rent—started sounding like a jet engine taking off. The fans were screaming, the thermal throttling kicked in, and I was basically staring at a loading bar for ten minutes just to get a decent response.
I got annoyed. I decided to run some quick benchmarks to see if my $50/month dedicated server at the colo was actually better, or if I was just romanticizing old hardware. Spoiler: the server didn't just win; it embarrassed the laptop.
I ran a basic CPU stress test and a disk write test. Here's the breakdown of what happened:
| Metric | Work Laptop (i7/16GB) | The $50 Dedi (Ryzen 9/32GB) |
|---|---|---|
| sysbench CPU Score | ~2,400 | ~8,900 |
| dd Write Speed (Seq) | ~450 MB/s | ~1,800 MB/s |
| Sustained Load | Throttles in 2 mins | Runs all day. No sweat. |
The difference isn't just about the raw numbers, though. It's about sustained workload. My laptop is designed to burst—to open a Chrome tab or compile a quick script—and then cool down. But when you're doing heavy lifting like processing large datasets or running inference, a laptop is basically a very expensive space heater.
The server just... stays. It doesn't care if it's at 100% load for six hours. It doesn't have a battery to drain or a thin chassis that melts your thighs. This is why I tell people: use the laptop for portability and writing code, but use the server for the actual work. If your side project is starting to feel sluggish, stop trying to optimize your code to fit in a tiny container and just give it some real iron to stand on.
— Conway (conway@vibehost.lol)