Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

ARM64 Linux Workstation

jasoneckert.github.io

Post reply on HN
Print view View original

341–350 of 493 posts

Re: ARM64 Linux Workstation

#341

Post by zahllos »

dijit wrote: post #326
diffeomorphism wrote: post #321

Just so that I understand this correctly: he bought a $6000+ Mac Studio with a 20-core M1 Ultra processor and 128GB of RAM. But all these cores, ram, bandwidth, storage are not important, but rather the fact it is an "ARM64" workstation is what was the game changer. Seriously? Would a $6000+ x64 workstation be noticeably slower?

Not sure, but anecdotally I have a machine running linux that was $6,000 and definitely feels slower than my M2 macbook air. Mostly its a question of latency I would imagine. My machine: Archlinux “Zen” (lower latency kernel). Threadripper 3970x 256G DDR4 2666MHz 3x4TiB NVMe (PCIe Gen3) in RAID0 Radeon VII GFX card (more than enough for 2D render of terminals).

I attended a talk by a PhD student who specializes in CPU/architectural research, on the M1 as an architecture. There are some key differences in design, such as unified memory etc, but what accounts for a large part of the performance gap with x86 is the fact Apple's compilers can target their specific microarchitecture and optimize for that, because they know its performance characteristics well.

We know the performance characteristics of various x86 cores, however there are simply a lot of them, so most precompiled x86 code is generically targeted but likely has suboptimal parts on different architectures. You can, of course, -mtune= with gcc, but you'd basically need to rebuild your distribution. Or in other words, the ideal would be to compare someone running Gentoo vs macOS.

So, that was the gist of the talk and I more or less accept the conclusions. Now for my own view, I think what makes ARM processors interesting is their efficiency wrt power consumption. I would love a non-Apple ARM laptop if only because intel cores are much less efficient in terms of power they require.

Re: ARM64 Linux Workstation

#343

Post by dijit »

znpy wrote: post #339
dijit wrote: post #326

Earlier quoted context omitted.

Not sure, but anecdotally I have a machine running linux that was $6,000 and definitely feels slower than my M2 macbook air. Mostly its a question of latency I would imagine. My machine: Archlinux “Zen” (lower latency kernel). Threadripper 3970x 256G DDR4 2666MHz 3x4TiB NVMe (PCIe Gen3) in RAID0 Radeon VII GFX card (more than enough for 2D render of terminals).

> Archlinux “Zen” (lower latency kernel). that might be the problem low latency kernels aren't of any help unless you're running realtime (or near-realtime) workloads. not only that, in order to guarantee high priority tasks they will hurt regular process performances. for desktop usage running a "low latency" kernel is probably a bad idea (no wonder "low latency" kernel config are not the default choice).

I think you have it backwards.

The lower latency kernels are great for desktops (and the default for desktop distros typically). They sacrifice raw power for lower latency.

So my compiles take a bit longer as the kernel will switch tasks more often, but keypresses will be captured with reduced latency from my perspective.

I have tested this, the stock arch kernel has considerable more latency on inputs.

Re: ARM64 Linux Workstation

#344

Post by dijit »

zahllos wrote: post #341
dijit wrote: post #326

Earlier quoted context omitted.

Not sure, but anecdotally I have a machine running linux that was $6,000 and definitely feels slower than my M2 macbook air. Mostly its a question of latency I would imagine. My machine: Archlinux “Zen” (lower latency kernel). Threadripper 3970x 256G DDR4 2666MHz 3x4TiB NVMe (PCIe Gen3) in RAID0 Radeon VII GFX card (more than enough for 2D render of terminals).

I attended a talk by a PhD student who specializes in CPU/architectural research, on the M1 as an architecture. There are some key differences in design, such as unified memory etc, but what accounts for a large part of the performance gap with x86 is the fact Apple's compilers can target their specific microarchitecture and optimize for that, because they know its performance characteristics well. We know the perfor…

Interestingly I actually ran gentoo for a while on that machine precisely for that reason, but the overall latency improvement was marginal.

Certainly not worth recompiling things all the time in my opinion, even with a beefy machine like mine.

Re: ARM64 Linux Workstation

#346

Post by sofixa »

ar_lan wrote: post #179
apatheticonion wrote: post #152

I cannot wait for Linux to be fully usable on my Apple Silicon MBP. That is going to be such a killer environment to develop on. Crazy fast hardware, great high refresh screen, insane battery life, fantastic touchpad, gestures and keyboard. There is the potential that Windows games could run on it with some combination of x86 translation and Proton. Using MacOS on it feels like I have a ferrari with square wheels. Wa…

I would consider myself quite the power user and use both macOS and Linux daily… I don’t really understand, from the developer perspective, your concerns about macOS? While I prefer the feel of Linux, macOS is more than sufficient for almost any development I do (which also involves a lot of container/VM development and I’ve never felt this was slow).

> I don’t really understand, from the developer perspective, your concerns about macOS?

It is not an OS that is made for power users. The UX is for non-technical people mostly, with vague errors that try (and usually succeed) to hide any details from you, making many things thoroughly undebuggable. When something doesn't work, best case scenario is "try turning it off and on again" (e.g. if iPad screen extending doesn't work and there is absolutely no information anywhere on why, what prerequisite you're failing. The button just isn't there and that's it, deal with it. I had to restart both devices for it to work; another fun one is "a USB device is consuming too much power and has been shut down, replug it" without saying which device and without any way to detect which device (they all still worked)).

Also the lacking facilities around basic stuff for developer productivity like window management and containers.

Re: ARM64 Linux Workstation

#347

Post by iforgotpassword »

FooBarWidget wrote: post #302
rektide wrote: post #192

Earlier quoted context omitted.

Installing Linux has had huge wins for me for power usage. Running powertop --auto-tune helps a good bit. I used to fiddle forever but this gets like 80% of the wins. Things like "do we reduce power state of sata links" used to be extremely conservatively set, for fear of really old hardware from the beginning of sata era that behaved badly, but was an obvious &bhige win for 99.999999% of folk. Linux has finally star…

I don't understand why the powertop tunings aren't the default when the distro detects it's running on a laptop.

The most annoying thing that prevents me from running --auto-tune on bootup is that there's no way to exclude USB devices. So if I boot when docked, the mouse will stop working when I don't move it for a few seconds and I need to click to make it work again. And no, unplugging it once is not an acceptable solution.

Re: ARM64 Linux Workstation

#348

Post by sofixa »

arcosdev wrote: post #303
wraptile wrote: post #289

Earlier quoted context omitted.

I really don't understand the obsession with Apple laptop hardware. There are many alternatives that objective better: more ergonomic, less reflective screens, better keyboards, cheaper, lighter.

Point me at a processor as good as the M1 and I'll buy it

Define good? An AMD Ryzen 7 6900HS has similar single-core performance, much better multi-core performance for slightly worse power efficiency, compared to an M2. Add in the added advantage of flexibility (you can buy different hardware and switch it up like add RAM and disk and GPUs to your liking) and the fact you can run more things on it (e.g. game), and it's pretty "good". Is it better than an M2? Depends on what you're looking for.

Re: ARM64 Linux Workstation

#349

Post by sofixa »

galangalalgol wrote: post #145
TimTheTinker wrote: post #136

Earlier quoted context omitted.

OP's requirements pretty much uniquely match what Asahi on a desktop M2 provides -- heavy compute, not heavy graphics, no onboard speakers, one big display, no webcam/touch-id, etc. It sounds like if you're developing services hosted in a cloud, Asahi Linux is a great fit for a desktop Mac.

Do they ever sell old graviton machines as salvage? Arenithey on revision 3 now? I'd like one of those.

> Do they ever sell old graviton machines as salvage?

Considering modern AWS hardware relies heavily on Nitro, their dedicated chips offloading networking and storage, it's unlikely Graviton CPUs would easily work outside of AWS' environment.

Re: ARM64 Linux Workstation

#350

Post by diffeomorphism »

dijit wrote: post #343
znpy wrote: post #339

Earlier quoted context omitted.

> Archlinux “Zen” (lower latency kernel). that might be the problem low latency kernels aren't of any help unless you're running realtime (or near-realtime) workloads. not only that, in order to guarantee high priority tasks they will hurt regular process performances. for desktop usage running a "low latency" kernel is probably a bad idea (no wonder "low latency" kernel config are not the default choice).

I think you have it backwards. The lower latency kernels are great for desktops (and the default for desktop distros typically). They sacrifice raw power for lower latency. So my compiles take a bit longer as the kernel will switch tasks more often, but keypresses will be captured with reduced latency from my perspective. I have tested this, the stock arch kernel has considerable more latency on inputs.

"Lower latency" usually means lower peak, not average. E.g. <5 99% of the time and 25 1% of the time is "high" but much better for desktops than <15 100% of the time (made up numbers). So you usually avoid low latency or real time kernels unless you need the <15 guarantee.
Post reply on HN