This could be easily explained by the fact that the M1 on battery run at full speed on Linux because Linux doesn't implement the proper power management. When the two are running on AC the difference is only around 15% for my use case (compiling C projects)
15% is still very huge. That is 1 or 2 generations of processors. That basically means by running Linux instead of OSX you have the performance of a to be M2+ chip.
You may find running the compilation workload in Linux in a VM on MacOS runs faster than on MacOS directly, while benefitting from the MacOS battery management.
This test is really too short to prove anything (less than half a second). Needs 10x 100x data, or thousands of runs to figure out if it's just the machines configuration/caching making the difference.
I am just guessing, but it's probably (surprisingly) because of e.g slow directory traversal API in Go: https://github.com/golang/go/issues/42027 . In my experience the biggest slowdown is on a mac for whatever reason. And yes, probably APFS itself is slower than ext4 which is honestly expected.
> macOS is known to be slow. If someone has Asahi running and wants some karma, try running lmbench and post the results.
Clarification required: slow by lmbench or what measure?
In my (personal, anecdotal) experience macOS is more responsive – with fewer UI hiccups – than Ubuntu on the same hardware (== faster!) but takes a lot longer to boot up (== slower!).
Anecdotal data here, MacOS is the slowest in basic operation, Browsing the web, managing files and more.
Since I use all three operating systems I don't really care. The thing is... I expected much more from the MacBook, because the whole internet said it's faster, when in my experience it's way slower. Two Macbook Air M1 vs Envy X360 2021 5700U 2021, vs Ryzen 2500u ThinkPad E585.
Pure anecdotal, but the internet has made more claims and backfired in full effect.
YMMV I think - I use macOS, Windows and Linux and find macOS to be the most responsive on comparable hardware. Nothing wrong with the others, but they just feel clunky and inconsistent. On the other hand whenever I start Apple Music on macOS (but not iOS) it immediately beachballs, so nothing is perfect. Also macOS doesn't work as well as it used to with magnetic hard drives - I think Linux and Windows may be better.
As you observed, you can't expect that great results on the internet (or on benchmarks) will automatically translate to great results in your own environment unless you're running the same workload. At the end of the day, I think the only meaningful benchmark is using the machine for your own work (or other use case!)
Regarding the M1 Macbook Air, its claim to fame is long battery life for the weight. Did you find your other laptops do better in terms of battery life?
>As you observed, you can't expect that great results on the internet (or on benchmarks) will automatically translate to great results in your own environment unless you're running the same workload. At the end of the day, I think the only meaningful benchmark is using the machine for your own work (or other use case!)
Regarding the M1 Macbook Air, its claim to fame is long battery life for the weight. Did you find your other laptops do better in terms of battery life?
I completely agree with you, I have no hate against MacOS or anything from Apple.
>Regarding the M1 Macbook Air, its claim to fame is long battery life for the weight. Did you find your other laptops do better in terms of battery life?
iCloud was using a lot of battery life and it really took a hit, because I was working on my Envy using Excel, Chrome, Edge, Firefox and Photoshop.
I think, I was just unfortunate to experience it. I think i'm not willing to buy a Macbook. The keyboard interaction is Out of the box kinda bad. On paper it's still one of the best laptops out there (even if my experiences where kinda... bad)
Is this supposed to be surprising? Try streaming video transcoding or 3D rendering and I'm pretty sure you get the opposite result. Some workloads perform better in one than the other. My guess would be that filesystem and IO latency is a major factor in these results.
Go has been optimized for Linux moreso than macOS.
Also, as noted in the thread:
> The time (210ms) is too short for a meaningful benchmark because initialization and the like are overshadowing the work done. You need a test with 100 times more pages.
