Hacker News new | past | comments | ask | show | jobs | submit login
Ask HN: Performance of Apple Sillicon vs. ThinkPads (Or Other Linux Laptops)
13 points by endorphine 5 months ago | hide | past | favorite | 28 comments
How's the performance, in terms of pure speed (CPU, memory bandwidth) when comparing macOS running in the latest Apple Silicon series vs. running Linux on a comparable (if there's such a thing) laptop like a Carbon X1 or similar?

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.




You can check out some benchmarks: https://www.notebookcheck.net/Apple-M4-SoC-analysis-AMD-Inte...

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).


Thinkpad thermals seem to be an issue. Apple Silicon laptops seem to have great thermals, and the Air doesn't even have a fan.


Yeah, I've loved the ThinkPads I've had, but universally they felt slow, hot, and poorly built (especially the Yogas). Loud too, under load. And I always buy the on site service with them because I've had to use that a few times as various parts broke. They're Lenovo's premium brand but ultimately still just commodity PC parts enclosed in plastic/carbon.

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.


It's hard to compare. You need a clearer criterium for "comparable". If they are comparable in terms of speed, so they are the same. If they are comparable in terms of price, so you need more info about what's the price that you are considering and what are the laptops with similar prices where you live (here in Brazil, Apple products are far more expensive than the others).

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.


My 11 year old ThinkPad still runs Linux fine. It is up to date and as secure as that makes it. Firefox, Thunderbird, VSCodium, Zettlr are snappy, background processes from python or postgres are stable. I have confidence a similar ThinkPad bought today will be fine in a decade.

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.


An 11 year old Mac can also run Linux fine.


Or Windows.


Plastic stools have their share of superior qualities. They tolerate liquids better, are generally lighter and cheaper, and may even be more durable. The manufacturing controls are likely better - no weak branches that get turned into a support.

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.


Right?! What a Fad! Arm SOCs only run billions of devices today. What do they know.

Old trusty hasn’t failed me yet! slaps the enormous heat pipe leading to the dust-filled fan


It doesn't really matter. They're both fast and you're aren't really going to notice or care for most workloads. I go between an M3 Max 30-core GPU system and an i5-12600K and 99% of the time I don't really discern a notable performance difference (although some of that 1% is "whoah!").

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.


> Linux runs well enough on M1 processors, but you should not buy an M1 for Linux alone.

Care to explain why?


Because you can buy a more well-supported PC laptop for less.


Fair enough, I think the "well-supported" thing is going to be less true over time, as Asahi keeps improving.


In > two decades I've never set eyes upon someone using Apple hardware for development unless it's as a dumb terminal. I work in embedded, hardware and applications, not mobile, web or scientific. I am sure such users are out there.


Heh, I used to work for a solar PV system manufacturer. Half the office was embedded systems and hardware engineers, who all ran System76 or Framework laptops. All the web devs had Macs. Management all used ThinkPads. You could tell what somebody did there just by looking at their laptops. And well, I guess the oscilloscopes were a bit of a giveaway too :)


In my much shorter 5 years, I’ve seen devs exclusively using Apple hardware.

And in my team, our dev environment has become increasingly local!


If you arent doing stuff that touches low level OS or high performance compute, it doesnt really matter what you use. Ive switched to Samsung Dex with a lapdock, and can easily do python and node dev. For high performance, I simply run a cloudflared tunnel to my home box.


In my decade its been majority MacOS. But Ive worked on the other side of this fence.


I have the exact opposite experience. Anyone not using a Mac gets raised eyebrows.


Linux will most likely beat macOS in performance on the same hardware due to it having better drivers/schedulers/filesystems and lower overhead.

If you are using something like Docker on macOS that will also kill your performance even more.


> Linux will most likely beat macOS in performance on the same hardware

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.


I've seen some Asahi benchmarks and Linux beats macOS in some of them while it loses in some others.

Asahi also supports the latest version of OpenGL and Vulkan was recently added, I can only imagine things will get even better.


> 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.


> Me too. Their work is really impressive.

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.


Can someone say baseless speculation?


Far from baseless speculation, there are many things that Linux does just _right_ compared to anything else, unfortunately my time is limited to list all of them here.

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.


I wanna preface by saying I'm a huge linux fan, but a couple things:

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.


> I get that in abstract improvements "could" trickle down from that but once again without any performance benchmarks that is the definition of speculation.

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.

[0] https://www.youtube.com/watch?v=9-IWMbJXoLM&t=1004s




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: