[kwlug-disc] m1 RISC silicon: impression

[Mikalai Birukou]

I'll add that crown for power efficiency per work unit probably also 
goes to apple's m1, away from amd. And this is a server-side consideration.

Link with power comparison: 
https://nanoreview.net/en/cpu-compare/apple-m1-max-vs-amd-ryzen-threadripper-3960x


On 2021-11-06 7:53 p.m., Jason Eckert wrote:
> I've been using a Mac Mini M1 (16GB RAM) for over a month now to do 
> curriculum development for a new program that includes iOS development 
> (so students will need to have an M1 Mac as we don't know how long 
> before Apple ditches support for Intel).
>
> The platform is more impressive than I expected. I've been using it as 
> my main workstation to see what the student experience will be 
> going through each course in our new program.
>
> Long story short, during a typical day I run a lot of things and have 
> been keeping them open to see if I could impact the performance. I've 
> been keeping a few web browsers with plenty of tabs open, as well as 
> Slack, VScode, Office, Adobe CC and GitHub Desktop and it doesn't even 
> hit the system. Everything is unexplainably snappy, and both %user and 
> %sys remain low.
>
> ARM-compiled versions of Photoshop and Illustrator run faster than on 
> my i9 MBP, and Android apps also compile in under a quarter of the time.
> The VMs we make (ARM versions of Fedora and Ubuntu) run lightning fast 
> in UTM/Qemu (we don't use XAMPP since students should know how to 
> stage a proper Linux server environment for what they need). Brew 
> times are also much faster than on Intel.
>
> I did explore Rosetta (running Intel apps), and the experience is 
> hit-and-miss. Some run fine (but not any faster than Intel), while 
> others run noticeably slower and have a big hit on %sys and memory 
> usage. One example of this is the Microsoft Teams app (which doesn't 
> have an ARM version for macOS yet). Luckily, Microsoft Edge has a 
> native ARM version for macOS and you can run Microsoft Teams as an 
> Edge app flawlessly with all the features of the standalone app (even 
> during a video call, it barely impacts %user, %sys and memory). The 
> only things I run in Rosetta currently are Logitech Options (for my 
> mouse) and Adobe Bridge. Everything else I installed already has a 
> native ARM version for macOS.
>
> I've had to eat all of the Apple-ARM jokes I've made this past year. Boo.
>
> On Sat, Nov 6, 2021 at 6:40 PM Doug Moen <doug at moens.org 
> <mailto:doug at moens.org>> wrote:
>
>     Yes. I'm considering buying a new macbook once Asahi Linux has a
>     working GPU driver. The new thicker, heavier Macbooks with more
>     ports and better repairability have tweaked my interest.
>
>     On Sat, Nov 6, 2021, at 6:27 PM, Mikalai Birukou via kwlug-disc wrote:
>     > My developer got a new apple with m1. We ran version of our soft
>     with
>     > scrypt KPDF algorithm in javascript, totally non-optimized. We
>     ran the
>     > x64 version first, and it went like regular linux x64 on x64
>     processor.
>     > Then we tried m1 version on this m1 chip, and we didn't really
>     notice
>     > that part of the program with the progress bar. Stark
>     difference. My
>     > hope that whole RISC industry will pick lessons from m1.
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://kwlug.org/pipermail/kwlug-disc_kwlug.org/attachments/20211107/2e0a5690/attachment.htm>