..XD so help me out with the maths, here, will you. The tdp of the m3 (not the m3 max or pro) supposedly is 21W.. on the cpu part. Given that it is anything like the m2, there is a boost stage that will draw up to 30W or so. And a gpu part that isn't included in the chart, but clocks in at another 20W. These kits will not reach anywhere near the desktop kits. And if it's anything like the m2, to get the advertised combined graphics and cpu performance out of this kit, it will exceed it's tdp by a lot. Your test doesn't explain how this works, how the clocks are running, nor how the single-core targets are reached.
Presumably - given that any of this is even remotely indicative of actual performance - Apple would be extremely happy to promote the advances from the m2 to the m3 kit. And yet .. no numbers. Why is that? Not suspicious at all, obviously..
In the same way, where this kit shines is - in theory - on the idle draw, and on the very lowest load burns. In theory you would be able to sit in your IDE and program, or stream registry-decoded video or other things, for towards 5-7W on average (which nets you the possible battery life of the 20h, thanks to the gigantic battery). Multithreaded tasks on low intensity is basically where it's at in ARM-land, given that the kits are optimised for that.
But you don't actually test for this, and the actual circumstances of "idle" draw may very well be different on a kit that has a different core layout. The suspicion obviously is that because the base model comes with more cores, that the idle is higher - since this is how it worked on the m2 and m1 versions(and indeed any ARM kit, including other relevant ARM-comparisons with embedded instruction sets).
Meanwhile, your (relative) comparisons seem to then pick out an average actual burn on the Intel chips (even though the burn will be higher when reaching the peak performance, so it's still underreported for the numbers you're using - the efficiency target, however, could be made higher by using realistic numbers of the actual draw in battery modes, since these runs would still reach somewhat high single-core speeds for shorter durations). And then you're pitting that against what looks like an ideally clocked performance(equivalent of the "avalanche") core on the m3, giving some expected ARM boost core load at 8W -- on that one core over time. If you did similar maths for an intel kit, would the boost core be achieved at 9W, ignoring the rest of the ambient cost, and all practical scenarios where the burn is higher?
In any case, whatever numbers are used -- by a relative standard this lets Apple beat the single-core performance of not just the Intel kits at 45W(very obviously, since at that target they are only allowed the shortest of boosts, if any at all), but of any kit - and so allow a combined synthetic score that hits above most kits in general.
Or put in a different way: the math you're operating with here doesn't tell us what the expected burns will be like in 3d, what the cpu burns will be on compilation loads, or what the nominal and idle burns actually will be. And in reality, the math doesn't even tell us what the theoretical peak performance of this kit really is.
What you're posting here genuinely tells us nothing.
Outside of that you're bending backwards to misrepresent both Intel and Apple kits(that both are higher in the usage contexts you're suggesting), so that none of the numbers actually reflect what you're getting in practice, I guess.
Deutsch
English
Español
Français
Italiano
Nederlands
Polski
Português
Русский
Türkçe
Svenska
Chinese
Magyar