I’ve been an avid Linux user for about 10 years now. Started out, as customary, hopping around - Ubuntu, Debian, Elementary OS, and eventually settling on Manjaro Linux. Loved the fact that everything I need lives in two repositories and I didn’t have to go scouring for PPAs to get my crap installed.
And you know what? Am happy. Still love it and have actually managed to teach it to four other people who I did not expect to be receptive of anything other than Windows at all.
Since you’re reading this piece there’s a non-zero chance you know that about four years ago I purchased a Mac Mini. After the project ended I was left with an M1 Mac Mini and zero idea what to do with it. And for a while, I didn’t do anything.
I was then coaxed into taking up homelabbing. You know who you are… The only obstacle was that I had nowhere to host my stuff and the idea of someone else’s crusty old ThinkPads didn’t sound appealing. I then remembered about the Mac Mini and hosted my things on in for some time.
But man is macOS irritating. It’s a through and through desktop OS and running it as a server is a pain - giant updates, no way to start Docker on boot, only way to run Docker is via Docker Desktop. In summary - nasty.
It was then when I stumbled upon Asahi Linux…
So as any self-respecting tinkerer would do I smacked Asahi on that beach and oh sweet jesus. Suddenly I had a no-room-takin’ power-sippin’ container-hostin’ server in my home. Perfection.
Oh, to add to the list of distros above, the official Asahi distro is Fedora.
Why are you telling me all this?
Well, because today I am starting another experiment. You see, the crusty old ThinkPad is the quint-essential Linux machine. Built like a tank and should you come across one with some grunt, the thing becomes the perfect development tool.
But with one single caveat. Ryzen is a beast, but Apple Silicon is crazy good. An 8-core Ryzen compiles a certain application in over 2 minutes, the regular M2 does it in just under a minute. I’d be lying if I said I wasn’t throwing shy glances at the MacBooks…
What about the experiment, then?
There’s a company here (Estland), GreenDice, that recycles fleet laptops, and they’ve generously allowed
me to pay my own money to them to rent an M1 MacBook Pro.
An LLM explains this sentence as GreenDice being a douche by sarcastically “offering” the laptop.
This is not the case, I asked for a Mac to rent, and they were able to fulfill the request. (I am still paying my own
money).
And onto this M1 MacBook Pro I shall smack Asahi Linux and attempt to daily drive it.
The begin
So, it is now 12.11.2025, and I have installed Asahi on the MBP.
An immediate annoyance is with the different placement of modifier keys on my keyboard.
ThinkPad: LCTRL, LMETA, LALT, SPACE, RALT, RCTRL
MacBook: LCTRL, LALT, LMETA, SPACE, RMETA, RALT
Not working things
I’ve a peculiar observation. I am not hindered by the Mac or the M1, those parts have worked surprisingly well. Okay, yes, I’ve not yet had time to set up a similar dev environment with my SSH and GPG keys and whatnot as I had in the ThonkPad, but for the light browsing and some video calls, it has been great.
Surprisingly, I am most annoyed by Fedora. The ThonkPad runs Arch, so a lot of stuff is easily installable from either official repos or AUR, whereas with Fedora I have to mess with COPR and flatpak. What even are those things???
Signal messenger
I wanted to install Signal on the Mac, but instructions are only for Debian based distros. First hit on DuckDuckGo is a snapcraft link. Eww. I’d rather not have Signal on the computer than get a snap.
Well, after some further investigation it seems I won’t have Signal desktop anyway as they don’t publish arm64 builds
for Linux at all. Poo…
Hugo
This blog uses Hugo and a theme. To publish a Hugo site you usually need Hugo CLI. I’ve also just gone through bumping the theme dependency to the latest version and found out that it requires Hugo CLI version at least 0.146. The current latest release is 0.152.2. Why am I getting this error then? I’ve just now installed the tool…?
Turns out that it is a Fedora thing. You see, Fedora seems to lock the version of a given tool to its own version. So whatever version the Fedora folk chose to release with, it is what you get and nothing else. Fedora Asahi Linux is based on Fedora 42, and Fedora 42 packages Hugo CLI 0.142.0.
Fortunately one of the strong points of Linux is running Docker containers, so a sudo dnf remove hugo coupled with a
| |
does the trick quite well.
The good shit
The Challenger: Apple MacBook Pro M1 (2021?) 8 GB
The Defender: Lenovo ThinkPad L14 G4 AMD Ryzen 5 PRO 7530U 32 GB
Benchmark 1 - Compilation of a Maven project
At the time of writing this, a still unpublished plugin for the healthcare integration engine Open Integration Engine.
As Maven is quite download-happy I’ve prepared both machines by running mvn package once to get dependencies. I know
there’s a better way to do this, but I didn’t feel like finding command right now.
I’m using the time functionality that is apparently built into ZSH to measure the duration of my build.sh script.
I’m also running nothing else on the computers, just the terminal.
So… time ./build.sh.
ThinkPad:
| |
MacBook:
| |
Strong victory with a 34% lead.
Benchmark 2 - Compilation on an Ant project
Why not go straight to the source and compile the engine itself?
Since Ant doesn’t really do dependency management this test was more straight-forward. Test command
time ant -f mirth-build.xml -DdisableSigning=true. I’ve disabled signing for this round.
ThinkPad:
| |
MacBook:
| |
MacBook was 9% faster in wall-time, but it was able to use a little less cpu.
Benchmark 3 - Ant project, but with signing
This is basically the same thing as above, but without the -DdisableSigning flag.
ThinkPad:
| |
MacBook:
| |
MacBook again with a 25% faster execution.
Another thing to bear in mind is that while the ThinkPad’s cooling fan took off, the MacBook stayed silent. Does it even have a fan? It only got a little warm under the wrist areas.
Fears are beginning to materialize
Yeap… A 4-year-old base model M1 beats a 2-year-old mid-range Ryzen. And yes, I do realize those test were not comprehensive, but they were the things I do every day several times.
I’ve even gotten somewhat used to the keyboard, and it’s wacky placement of modifier keys. I wonder how much faster an M5 will be in the then-not-even-remotely-fair comparison…
Now what?
Heh. Funny coincidence… (Because RAM is getting very expensive now)
But yea, the poor thing ran out of it’s 8 gigs of memory and started stopping applications. Which I was not really fond of since I was using said applications.
I think since this issue of there being not enough dedotated wam is quite a significant obstacle in my day-to-day operations I’m going to retire the project early. I was hoping to run this for a month, but the thing can’t really handle anything even remotely resembling my usual workflow.
However, the experiment proved that it is technically possible to daily Asahi - you’ll just have to carefully plan for which applications you want to use.
One thing is for sure - there’s no way I’ll get a mac in the ah shit fuck my finger slipped! Dagnabbit!
I promise getting the 24 GB M5 MacBook Pro was a complete accident and not at all related in any way to the M1 benchmarks.
I promise!
Well, it actually wasn’t. Large part of the decision was due to the SSD on the ThinkPad crapping itself (or not, but at that moment I was very tired of linux). I still have to dig through it to see what went wrong and get my crap out, but I’m right now more interested in me new toy.
Benchmarks?
Compiling the Maven project
| |
Aww yea, 50% faster. That’s what I’m talking about!
Compiling the Ant project
| |
Wut? How is the M5 slower than the ThinkPad by 53 seconds and slower than the M1 mac by 67 seconds? I thought more M-s meant more better?
The internet claims the issue lies within Apple’s APFS filesystem, which is supposedly not good for Ant’s IO-heavy nature. But the M1? That is running Fedora and btrfs and is thus a champ. Kay…
Compiling the Ant project with signing
| |
Again we see that the M5 is slower than the ThinkPad by 27 seconds and slower than the M1 by 99 seconds. Well damn. So if you have to do Ant stuff, don’t get a mac, I guess.
A subtitle to separate the following sentence from the tail of the benchmark part
But since I’m currently deciding to ditch the Ant stuff from my daily work I’ll not take the M5 back to the store :)
Also, I suppose there will some rage-posts in the coming future as I settle into the madness that is macOS.