• 0 Posts
  • 17 Comments
Joined 1 year ago
cake
Cake day: June 12th, 2023

help-circle















  • there has never been a competitive, high performance laptop like the current MacBook Air build on x86

    That bit is easy to explain. Apple (again) is on the latest node, so they do currently have the highest performance per watt SoC out there.
    So it seems unsurprising that it’s hard to compete with the latest. But the N5 is starting to get out. AMDs 7840U should be comparable, but of course is out roughly a year later. And that’s going to be true for a while, because Apples markup allows them enough profits on the latest node and Apples vertical integration means they can be quicker to release a new device with their own new SoC, whereas for the competition they have to wait until AMD releases their products, and then build their product (the Laptop) around that.
    I feel like MacOS could also be more efficient than Windows, especially in daily use but I may be wrong on that feeling, Apple is not making it easy to tell.

    And of course there have been plenty of passively cooled x86 devices, but they’ve not been “good enough”

    And finally, none of this is meant to knock Apples Achievements with ARM.
    The native extensions for x86 translation they put in are pretty genius.
    Being able to compete with AMD/Intel/Nvidia on their first out is really impressive as well.
    M1 M2 etc. are great products, they’re just not magic, and unfortunately intentionally very limited (no Vulkan, no DirectX etc.).


  • With Apple’s M-processors reigning supreme in the laptop space with insane values for performance-to-powerdraw (and in turn heat radiation and cooling requirements) the days of x86-by-default laptops are probably numbered and more manufacturers may want to switch to ARM, to avoid unfavorable comparisons to MacBooks.

    I think this is a misconception.
    M-processors are not amazing because they are ARM, or because they are Apple.
    They are pretty much where everyone else is al well, just one node shrink ahead, because Apple is the first in Line, because they can pay for it.

    for example, Apple M1 GPU vs Steam Deck GPU, Apple has a ~60% GPU lead (in performance measured in TFLOPs fp32). On the CPU side it’s ~70% (in a fairly bad comparison, as there are notable differences between the analog used here). But the thing so many people ignore is that the M1 is on TSMC N5, whereas the Steam Deck GPU is built on the N7 node, (and there was the N6 node in between those two!)
    The A12 is Apples N7 SoC, and draws up to ~6W, and the GPU has roughly 1/3rd of AMD Steam Deck compute, pretty in line with power draw.
    Watt for Watt, Node for Node pure performance seems just good to me, not really surpassing anything else by a lot.