Linux is not “ready to run” on Apple Silicon, but give it time
arstechnica.com
arstechnica.com
This article is addressing misleading comments that made it sound like everything for apple silicon was merged into upstream kernels. It's not. You're still going to need to use Asahi kernels and userspace bits for a while.
I mean, the fact that we have a widely available alternative makes the picture slightly less bleak, but still. How many politically-inclined persons rave endlessly about freedom and sovereignty while making no effort on such an important front? The Linux usage numbers are so far from where they should be.
Essentially the entirety of modern humanity is built on black boxes. I'd like to find one person who knows exactly what and where every single component on their motherboard is and why it is there, the precise transistor layout of every chip from the GPU and CPU down to the tiniest microcontrollers, and every single instruction that is running on their computers at any time.
Sure, you can run Linux, but it still is running on a closed-source CPU. For all we know (unlikely as it is), Intel-AMD CPUs could have a backdoor saying 'hey, send the instruction and data cache to the network controller and to this IP address'. These CPUs still have proprietary microcode.
Today, I run Asahi Linux on a M1 Ultra Mac Studio using Sway (https://triosdevelopers.com/jason.eckert/stuff/AsahiSwayM1Ul...), and 100% of the features I use are there in Asahi kernels and userspace (sound/Bluetooth/audio/HDMI/USB/10G Ethernet/WiFi/GPU driver).
In short, it's the fastest RISC Linux workstation you'll find, and has been rock solid in every way. In fact, it's so fast, it messes up the timing of K3s (you just have to remember to put resource limits on each pod to prevent the CrashLoopBackOffs).
Not sure why people are blaming various software projects for the breakage that was introduced by apple.
I would give you that Apple is not third party friendly, but how many hardware vendors are making hard modifications so third party software runs on their hardware?
Most hardware manufacturers suck in that regard too. That in itself doesn't make Apple any better.
It's more like making roads that don't have the normal line markings on them and occasionally have gaps in the parts of the road apple cars don't normally drive on, etc.
I would like to know of a single software vendor, that is not Microsoft, who may have influenced the design choices of a hardware vendor.
Because it's not like Linus himself randomly walks into Intel offices to tell them how to get things done.
Probably lots of game mega-companies (influencing GPUs). Arm had special JavaScript instructions for a while.
Linus did use to work for transmeta, at least.
Haven't ever heard of this.
> Arm had special JavaScript instructions for a while
Which is a design choice. ECMAscript is a standard, and the instruction seems to be a way of making certain conversions faster. From the release document [0]:
"Armv8.3-A adds instructions that convert a double-precision floating-point number to a signed 32-bit integer with round towards zero. Where the integer result is outside the range of a signed 32-bit integer (DP float supports integer precision up to 53 bits), the value stored as the result is the integer conversion modulo 232, taking the same sign as the input float."
It is not unusual for hardware vendors to include instructions that would make certain common operations faster. That does not mean that software vendors, and in this context, compiler engineers, dictate CPU design choices.
[0] https://community.arm.com/arm-community-blogs/b/architecture...
Who would buy a Mac and put Linux on it?
What am I not understanding?
https://lore.kernel.org/lkml/CAHk-=wgrz5BBk=rCz7W28Fj_o02s0X...
2. Per watt they have some of the best performance
3. Overall some of the best battery life
4. Only real option if you want to run ARM
5. Most important - just because. That’s what geeks and hackers do.
> I haven't done it on the new M laptops
From these two comments, I am guessing you are running on pre-touchbar (2016) Macbook Pro's, because the touchbar models before the M1/M2 were absolutely horrendous for running Linux. You can see a table of features that do or do not work still on those models[0], notably wifi -- perhaps the most critical feature for a laptop to be useful as a mobile device.
The pre-2016/pre-touchbar Macs are such nice pieces of hardware despite their age, but it gets me wondering why not invest in another hardware platform that supports Linux well, is modern but has high quality hardware?
Some examples I can think of are the XPS and the Thinkpad Carbon.
Yes, but Apple has a great combination of high performance (laptop-wise), long battery life and high portability. Before I had the Zenbook series where you had the high end with the i-5 chips that died in 2-3 hours, or the low end that lived 5-6 hours but could barely play a video in 1080p (IIRC, from memory).
With the Air for example I have it all; very long battery life, pretty high performance (more than enough for what I use it), super portable, and a great keyboard+trackpad. The only thing I'm missing is Linux :)
Frankly i'm not even sure i want to buy Apple. Not a big fan these days. But those Apple cores just seem so damn battery efficient.
The number one thing i want from a laptop is to be able to use it on my lap, disconnected from a wall for as long as possible. With reasonable performance.. ie turning a CPU down to 3% doesn't count hah.
I'm kinda split these days on if i buy Apple or buy a Linux and repair friendly laptop (Framework laptop, maybe one of the System 76, etc). A fully working Linux (NixOS for me) installation on an M1/M2 laptop would make this an even more difficult decision.
macbook air (m1 or m2) fits the bill. Long battery life, no noise, rarely any heat.
While I like macOS as a "desktop" OS, I run also Linux on my Macs. So far in a VM, which obviously is very convenient. But I am certainly looking forward to installing Asahi.
Apple charges an absurd amount for tiny little features that real computers have in spades. Like RAM going for $200+ increments per 8GB, or $200+ increments per 512GB of SSD storage. That's almost offensive pricing if you know actual hardware costs. Why? Because Apple "Silicon" computers are not modular.
For a top end Apple M2 you're spending $2500 just to get feature parity with a $1000 PC and even then you're maxing out at a pathetic 2TB of storage and only 24GB of ram. If you built a $2500 PC it would absolutely, unequivocally, blow the highest end M2 out of the water in both performance (especially after thermal throttling on the M2) and ability to do computer task (ie, having more space than a Dell computer from 2006 which Apple M1/M2 products cannot).
The criticism of the onboard flash is more valid - it's costly and if you don't get the dual bank setup, paying yes, a big premium like I did for a 512Gb version, performance isn't great. https://www.theverge.com/23220299/apple-macbook-air-m2-slow-... The lack of an upgrade path is, again, not great, but ask yourself honestly, aside from some ThinkPads which really were nicely upgreadable in the past, how many laptops are upgradeable and what can you upgrade?
Your bias ("Apple charges an absurd amount," for "tiny little features that real computers have in spades") is showing and it isn't pretty. Show us the modular laptop with comparable performance or shut up; you're stuck in the nineties.
Go on - find me a PC laptop that performs better than an M2 and let's look at prices, weight, and battery life. Oh, and modularity.
https://www.cpu-monkey.com/en/cpu_ranking-best_mobile_proces...
Or even show us your PC build that is comparable to the Mac Studio and let's talk about how big, noisy, and power-hungry it is; these are things that people who want to use real computers in real offices and studios care about, more than PC hot-rodders who, let's face it, are mostly gamers bragging to each other about their frame rates rather than professionals using computers as tools.
For a workstation there is no comparison in terms of either performance or price. (Mac does have the high-bandwidth unified memory going for it which can make a difference in some situations.)
A 7950X is 50% faster than an M1 Ultra and it only costs 550 or so.
It also has 1/8th the memory bandwidth, a MUCH slower iGPU, and takes much more power.
Sure you can add a nice GPU, which helps for GPU workloads, but if it's from the current gen it's at least another $800 and 300 watts.
The M1 Ultra in comparison has the fastest iGPU out, competitive (depending on workload) with some of the mid range GPUs, but generally good enough for things except for gaming. The apple studio is small, quiet, sips power, and has a single large fan that under most workloads is silent.
A 3070 goes for under $500 these days and is a very capable GPU, certainly significantly more powerful than a PS5. It's apparently around 50% faster than the M1 Ultra's GPU at OpenCL and gives you access to the CUDA ecosystem.
The Mac hardware is a great piece of engineering, but for many workstation use cases the latest x86 beats it handily in both performance and price.
The RTX3070 is physically large, does have a nice 448GB/sec or bandwidth, but only for 8GB ram, and of course only for things that can run on a GPU. The 7950X is limited for 83GB/sec (at least without overclocking), which is 5GB/sec per core. The M1 max is 400GB/sec and the M1 Ultra is 800GB/sec, granted the CPUs can only use approximately 50% of that, but as a bonus the GPU can access all of ram, not just 8GB.
So even a MiniITX + 3070 is going to be physically large, power intensive, and hard to get silent, at least until you start getting pretty exotic and expensive with water cooling.
PyTorch and related are supporting Apple's CPU, GPU, and AMX I believe. Benchmarks vary of course. One interesting development is I've just started seeing support for Apple's neural engine and on the models benchmarks it was something along the lines of 7x as fast as the M1's GPU. Something to keep an eye on.
But if gaming framerate per $ is your goal I'd definitely get the 7950x, it's going to love the RTX3070, is cache friendly, and won't care that the 7950x has a 80GB/sec memory interface.
> it's going to take massive investment by intel or AMD to catch up in the price/power/performance category
in the context of desktop workstations is simply untrue. The post compared a Mac Studio to a Threadripper that retails for over $3000, but my point was that it's not a fair comparison: a high-end consumer part that retails for less than 20% of the price also beats the Mac Studio at many use cases, such as code compilation.
There's this weird thing I see where a lot of people think, probably due to Apple's marketing, that Apple Silicon hardware is not just "strictly better" but a gargantuan leap beyond x86. That just isn't true, now that AMD is on a comparable process node. Apple Silicon is better at some things and worse at others, which is great -- competition is good! -- and completely normal.
I'm no mac fanatic, but arguably no good competitors came out for the 2012-era macbook pro (the retina thingy, like I said, not a zealot) in terms of hardware quality. Yeah the CPUs eventually got faster, and Apple drank their koolaid with the bad keyboards and ribbon.
The laptop industry never really responded to that, and I don't really see the ARM competition from AMD/Intel... just excuses about "well they were on the next generation of process".
Like, what process jump previous to this for x86 produced such dramatic gains in efficiency and power? Oh right, certainly none in the last 20 years.
I do hope an ARM M-killer is being worked on. But the absolute silence means it is far away. If it was even a year away there would be rumblings in the downstream Windows/Microsoft market, hardware OEMs, compilers, etc.
Anyone heard anything?
Also, holy crap, $3000 for a processor? That you bought for personal use?
My company got it for me, standard issue here (working on a very compute instensive backend that we run locally)
There's probably a laptop out there that's as nice as a piece of art than a MacBook but the internals are probably crap in comparison.
Some people might buy a mac and put linux on it to help out asahi linux with testing/troubleshooting etc, or because they like the hardware more than what other brands offer. Others in this comment section are running Asahi Linux as their daily driver apparently. There really isn't an ARM machine on the market like an M1/M2.
> you could buy a faster Intel or AND machine
Genuine question, what intel or AMD chip is better than the M1 or M2? Are you referring to desktops, laptops, or both?
I get that you can buy/build a very good machine for less money than brand new Mac, but is the performance really as good as the m1 or m2?
You can say plenty of things about apple but they do make great hardware more often than not.
Generally the competition from AMD, Intel, and Nvidia generally has much less memory bandwidth (2 to 4 times less), much larger, and full of pitfalls. Oh you bought the wrong GPU that's catching on fire? Or the GPU with the fans that die or whine? Or a system with a iGPU severely hamstrung by a 128 bit wide memory interface? Or the motherboards with active cooling muffin fans that die/squeak/whine? Oh you bought brand X motherboard which was good last year but terrible this year? Even the PC case designs are terrible. Why do GPUs form a cube with minimal airflow on nearly every standard desktop chassis in existence?
AMD/Intel desktops remind me of non-tesla EVs where often the suspension, drive train, battery, in car entertainment, traction control and similar departments don't talk, compete for resources, and minimize shared designs. PC airflow is engineered for each device separately. Even simple things like front to back airflow can not be assumed. Some GPUs don't exhaust air, just stir it. Others intake air (fighting front to back airflow). Others exhaust air ... often sucking in hot air from the CPU side. Sadly the hottest component (generally the GPU) has the worst airflow, while the CPU at 1/3rd the power often has 2x the airflow in a normal ATX case. Why do motherboard chipsets depend on tiny high rpm fans that make more noise/vibration than airflow? Why do high performance M.2 cards often require heat sinks ... that violate the size specifications for the M2? Why are dimms often perpendicular with the airflow? Why does even a small PC case require 5 or more fans instead of just 1?
Except for gaming a Mac Mini with a M2 Pro or a Mac Studio M1 Max seems plenty fast, likely to last a long time, silent, and plenty fast for a wide variety of purposes while being substantially quieter, smaller, and lower power than it's PC competition. Personally I'm waiting for apple to bring out a Mac Studio with the M2 max. I'd love to see something arm or x86-64 that would make a nice quiet, fast, small, power efficient desktop with good memory bandwidth to support a decent iGPU. Sadly I'm not expecting anything. It's possible AMD will sell a small/cheaper version of the MI300 or Nvidia will sell a smaller/cheaper version of the Grace+Hopper, but that seems like years away.
That's the problem: the ATX standard. It's almost 30 years old at this point. It was never designed for modern GPUs or good thermal performance, it was just a step up from the creaky old IBM PC-AT standard that came before. The only way to get a desktop PC designed for good airflow and proper thermal management is to abandon the ATX standard and do something custom and proprietary, like the small-form-factor or workstation corporate desktops from Dell. But that means you need to buy the entire system at once (just like buying a laptop), and not put it together from parts like in the 1990s. And it's impossible to have a new open standard because no huge players are stepping up to make such a standard and then force everyone else to use it, and it's impossible for the industry as a whole to sit down and make a standard together that doesn't suck.
Sadly Apple can build low power small systems with 2x, 4x, and 8x the memory bandwidth, but nobody else seems interested.
They didn't? Admittedly I don't know that much about it, but the Wikipedia page is very thorough. I don't remember IBM making the PC-AT a standard either; it became a de-facto standard because everyone just copied it, and the ISA bus was open. I don't see what's stopping anyone else from copying Intel's NUC form factor.
The difference between the 80s/90s and now is that there isn't a single 800lb gorilla for everyone to copy, and for some reason other companies just don't care to take the stuff Intel has done and copy that. According to the Wikipedia article, Gigabyte made their own version of NUC that was similar but not the same, so of course parts aren't going to be compatible; when companies are doing this instead of just copying a single big competitor, then there's no de-facto standard, just a bunch of competing and incompatible standards.
Correct, they did not. They did with ATX and BTX (but BTX was dropped before it caught on broadly, in a focus shift.)
> I don’t remember IBM making the PC-AT a standard either
Yeah, but you are the one who brought IBM up, so what?
> It became a de-facto standard because everyone just copied it,
Yeah, its not the 1980s, and there isn’t one manufacturer that is both as dominant as IBM and as unconcerned with protecting itself from cloning as as IBM was until the PS/2.
> The difference between the 80s/90s and now is that there isn’t a single 800lb gorilla for everyone to copy, and for some reason other companies just don’t care to take the stuff Intel has done and copy that.
There wasn’t an 800lb gorilla to copy in the 1990s, either, because IBM went proprietary with the PS/2. The =difference between the 1990s/2000s and today is that Intel (or someone else similarly situated; Nvidia could plausibly do it) isn’t actively promoting form factor standards, the way it did with ATX (from 1995) and BTX (from 2004, but stopped being actively promoted in a few years because Intel temporarily shifted business focus to low power designs, and BTX was centered around improved airflow for high power applications.)
Feels like it would be good antitrust brownie points for something unlikely to upset their desktop OS market share for the general public.
(Sub-0.1% chance given their track record, I know.)