Would I see significant performance improvements (e.g. compilation speeds) by moving to M3/M4?
Let's say that I do not care about energy efficiency/battery life.
https://www.cpubenchmark.net/laptop.html
It seems like the Apple Silicon chips are generally faster single-core, but a bit slower multi-core. In real life, it will probably depend on how well optimized your compiler (or whatever software) is for Apple Silicon. And native Linux on Apple hardware isn't mature yet, so if you need to virtualize it, it'll be even slower.
The advantage of the Macs are generally efficiency (performance per watt, meaning longer battery life and less heat and noise) and build quality, and well, you get a Mac (if you like macOS). If you don't care and just need something to crank out FLOPs, IMO I'd probably just build a desktop and run the workload there and remote into it from a cheap laptop of any sort.
If you do get a Thinkpad though, read up on their thermal management and try to get one that doesn't throttle so easily. The X1s I had usually had lower voltage CPUs (less performance) and would throttle even those under load, so they weren't good for long sustained CPU usage. Fine for business use or the occasional `npm install`, but if you're running multi-hour workloads, it will very likely overheat and throttle down. Not sure if their workstation models (the T models and others) are better in this regard; they are generally larger and not as squished together (read: hopefully better cooling/more fans).
They are very light though. Not sure if the T model workstations are different (heavier and more robust?)
The modern MacBooks are in a different world altogether, IMO. They make even the best ThinkPads I've owned feel like toys. They are just so so nice and leaving legacy x86 feels so liberating. Incredible power and battery life and performance.
If I had to buy a PC laptop it'd probably be a ThinkPad. But because Apple Silicon is a thing, I don't think I'll buy another PC laptop again. It's such a dramatic difference.
I don't need Linux (on the desktop) though. If that were a firm requirement, I wouldn't want to rely only on Asahi for production use.
Not to mention the architecture: Apple Silicon performs better in some tasks (e.g. running Android emulators) not because they are faster, but because they are ARM.
PS: I have a MacBook Air, but in terms of performance purely there are better options for the same price here.
An 11 year old Apple isn't in the same state. An M* processor is a gimmick for today's needs; it is a niche platform solution that's reactionary to niche needs of today; that's not a design for the future. Hence M1, M2, M2, M4, M88, etc. The Apple OS will not support decade old hardware, it's not in the business model. You have a fine computer for today that is designed to be disposable. Like buying a plastic stool instead of a wooden one. The plastic one may be lighter, in a variety of colours, your friends might go for them, and it could lie about being privacy focused, but in the end it's a plastic stool not a wooden one.
My point is just to emphasize that to the extent t your analogy works, it does not indicate one or the other is universally superior.
Old trusty hasn’t failed me yet! slaps the enormous heat pipe leading to the dust-filled fan
Apple Silicon processors are generally much faster than equivalent AMD and Intel laptop processors.
Which OS matters more to you? If Linux, then use Intel. If macOS, then use Apple Silicon. Linux runs well enough on M1 processors, but you should not buy an M1 for Linux alone. Linux does not yet run on the newest Apple Silicon processors.
Care to explain why?
And in my team, our dev environment has become increasingly local!
If you are using something like Docker on macOS that will also kill your performance even more.
Not exactly, because since M series was introduced we only have Asahi, developed by reversing engineering Apple Silicon processors instead of using drivers of macOS that were developed by Apple for the hardware that they know how works.
Apple Silicon processors have efficiency and performance cores, macOS scheduler was made to manage it. I don't know if Asahi handles them as different cores.
But I must say, Asahi team is doing a great job.
Asahi also supports the latest version of OpenGL and Vulkan was recently added, I can only imagine things will get even better.
Me too. Their work is really impressive. But I don't think it is worth getting a MacBook for running Linux nowadays.
It is indeed, I have a theory about why that is and I think it comes down to the people who work on it.
Not saying that Apple doesn't have good engineering but people like Alyssa Rosenzweig are on another level.
Also, the fact that Linux has supported ARM for years probably helps.
> But I don't think it is worth getting a MacBook for running Linux nowadays.
Why though? I probably wouldn't as well but I'm curious about what you have to say.
I will just say that one big advantage of Linux compared to other kernels is the fact that the same Linux kernel that your average Joe is running on their own machine, is the same kernel that CERN is running in their experiments.
Improvements to any area of the kernel get upstreamed and sometimes those affect (and benefit) various setups other than the intended machine, this is mainly due to the nature of open source and shared codebase, Linus himself explains it here:
https://youtu.be/MShbP3OpASA?t=1335
https://youtu.be/MShbP3OpASA?t=2280
I think this is more than what any company could ever dream of achieving, they might just lack the resources to do it.
1. I'm not sure what kernel optimizations for CERN's x86 machine have to do with ARM Apple Silicon. I get that in abstract improvements "could" trickle down from that but once again without any performance benchmarks that is the definition of speculation.
Yeah, improvements and bug fixes get upstreamed but not every algorithm is going to optimized for the architecture of Apple Silicon.
Fair enough, I am not sure whether CPU schedulers are shared across architectures, I am also not aware of any recent benchmarks other than: https://www.phoronix.com/review/apple-m2-zen4-mobile
Where I was going with my original statement/post is that I wouldn't be surprised to see Linux doing better in some benchmarks due to improvements in the kernel, there has been many recent developments such as io_uring[0], MGLRU, etc. that could make a difference compared to other systems.