Re: Apple requires too much money and sacrifice of control

On 2025-03-23 05:01:09 +0000, Your Name said:

On 2025-03-23 03:15:14 +0000, vallor said:

On Sat, 22 Mar 2025 19:17:14 -0700, Alan <nuh-uh@nope.com> wrote in
<vrnqva$18oag$4@dont-email.me>:
 

On 2025-03-22 14:52, Joel wrote:

Alan <nuh-uh@nope.com> wrote:
 

[A souped up Mac mini] would match what I have with Linux [in
support of needed apps and use].

What actual tasks do you use the computer to complete?

 Keeping everything I run loaded at once.

 What do you run, and what resources does each use?

  Take a look at my Cinnamon taskbar:  https://i.imgur.com/yPGDm6a.png

 So nothing a Mac Mini with a base configuration couldn't easily handle.

 I've owned two Mac Minis -- it's notebook hardware that runs slow as
molasses.
 We have a Mac Studio now, which was _way_ overpriced for what we got.
Probably the most intensive application I've tried on it is Fooocus
(which uses pyTorch), and I'd roughly estimate it is 1/4 the speed
of my Linux workstation.
 (You might blame pyTorch for that, as perhaps it doesn't use
the GPU/NPU --

 Almost certainly.
 If it's an older version of PyTorch, then possibly you are running the Intel Mac version on the Mac Studio via Rosetta x86 emulation, which would be slower than running a newer Apple Silicon version.
 Enable the GPU on Apple Silicon Macs means making changes to the PyTorch settings, which you may already have done:
<https://wiki.cci.arts.ac.uk/books/how-to-guides/page/enable-gpu-support-with-pytorch-macos>  <https://medium.com/@mustafamujahid01/pytorch-for-mac-m1-m2-with-gpu-acceleration-2023-jupyter-and-vs-code-setup-for-pytorch-included-100c0d0acfe2> 

Apparently there are also speed differences depending on how you installed the MacOS version of PyTorch (conda vs PyPI):
<https://github.com/pytorch/pytorch/issues/2163>