Apple MacBook Air 13 M3 review - A lot faster and with Wi-Fi 6E

[Neenyah]

and overall about "working 10hrs is not self respectful" - imo it's more like an opinion of someone who never worked and probably typing this 'high-end philosophy' on a cheap lenovo, while $200-300k/yr engineers are working their asses off and sleeping at work lol.

To be fair he said:

Absolutely nobody with a brain or an ounce of self respect is working nonstop for over 10-11 hours.

And there really is no sane person on the planet who won't take a break of at least 20-30 minutes if they really have to work 10-11-12 hours shift.

[DontFearTheFuture]

A laptop with the supposedly amazing Meteor Lake processor has 200 minutes less runtime despite a 20Wh larger battery. It's wild how far behind intel STILL is.

You can't compare a CPU with a Base TDP of 10.5 watts (M3 base) to a CPU with a base TDP of 28 watts (Intel 155H)
Here's what I mean:

I'm going to compare an Apple M3 chip to ...........another Apple M3 chip.

Let's compare both these M3 chips on a "Sustained" Run of Cinebench R23:

Apple M3 Base 8 Core (10.5 watt sustained) CPU:  8,237 points / 10.5 watts = 784 points / watt

Apple M3 Pro 12 Core (27 watt sustained) CPU: 15,106 points / 27 watts: = 559 points / watt

784 points / 556 points = 1.41. 

So, the M3 Base has 41% more performance per watt than a M3 Pro.  Therefore, I conclude that an M3 chip is better than an M3 chip.  (what?)

Hmmm, I wonder what happens when I compare 2 chips with the same TPD.  Well, let's see:

Let's compare the 27 Watt TPD M3 Pro to a 28 Watt TDP Intel 155H (both running at their base TDPs):

Intel 155H at 28 watts sustained: 13,000 Points.  13,000 Points / 28 watts = 464 points / watt.

As I mentioned above, the Apple M3 Pro 12-core (27 watt sustained) CPU: 15,106 / 27 watts: = 559 points / watt

559 / 464 = 1.2.

So when comparing these 2 chips, M3 Pro chip is roughly 20% more performant per watt under load than a 155H when using the same amount of watts. That's not that much, considering an optimization of a 155H (a second gen. 155H if you will) may increase efficiency by 10% alone.

Now, let's run that same Intel 155H at 45 watts instead of 28 watts:  14,500 points / 45 watts = 322 points / watt.

464/322 = 1.44.  Therefore, a 155H running at 28 Watts is 44% more performant per Watt than when running at 45 Watts.

Therefore, a 155H is better than a 155H (what?)

Well, what this shows (and proves), is the more energy that goes through a processor, the less efficient it is. 

That's why the only way to correctly compare chips is when the TDPs are the same.

Folks, Intel is not as far off as people think.

[lmao]

Cinebench R23

r23 and earlier ones can't be used to compare arm and x86, only 2024 got official apple silicon support

all your calculations are way off too because benchmark numbers weren't achieved at some 'TDP' wattage. they all have their specific power consumption numbers that is different from tdp by unknown amount, somewhere between 'idle' and 'full load'.

what this shows (and proves), is the more energy that goes through a processor, the less efficient it is

there's a correlation but there's no causality

[hubris]

I will never understand that "better battery" argument

Longevity. 4-6 hrs might be enough in short term but after years of use when battery health goes down, you looking at only 2-3 hrs by then. Sure you could replace the battery but that costs additional money. Would be nice to have such good battery life to begin with that even after the cells deteriorate you're still getting at least 4 hrs even after 5+ years.

[lmao]

good news btw, base 256Gb models are coming with 2 nand ssd chips again, so no more halved ssd speeds for base M3s

[DontFearTheFuture]

Cinebench R23

r23 and earlier ones can't be used to compare arm and x86, only 2024 got official apple silicon support

all your calculations are way off too because benchmark numbers weren't achieved at some 'TDP' wattage. they all have their specific power consumption numbers that is different from tdp by unknown amount, somewhere between 'idle' and 'full load'.

what this shows (and proves), is the more energy that goes through a processor, the less efficient it is

there's a correlation but there's no causality

Actually, Cinebench R23 does run Natively on Apple M Processors.  Has been since the very first M1 Processor.  Anything Before R23 is not run Natively.

Regarding TDP:  Notice I state "SUSTAINED wattage". (AKA PL1 or Powerlimit 1, which is set by the manufacture of every laptop)

Therefore, every score I gave in my post was based on SUSTAINED performance in R23 running in a loop; and after either 10 to 15 minutes, the score of the last run, along with how many watts were be used during that run, are posted.

Literally, EVERY Tech Reviewer uses this method to conclude sustained performance.

Cinebench R24 is newer and is a longer test; and long enough of a test to where most laptops will have reached thermal limits in 1 run.

Notebookcheck does this same test I explained above using R15.  (though, R15 does not run Natively on Apple Silicon (maybe this was the Cinebench your thinking of))

In short, I'm correct in everything that I said.

PS, MacOS is known to be a lighter OS than Windows.  There is a really good chance that Apple gains a bit of efficiency due to such.

[lmao]

Cinebench R23 does run Natively on Apple M Processors

'running natively' doesn't mean 'optimized'
it's right there on Cinebench homepage, support was added only in cinebench 2024

Therefore, every score I gave in my post was based on SUSTAINED performance in R23 running in a loop

no, you've grabbed some random value called 'TDP', that is unrelated to power consumption and is actually there for thermal design - and started crunching unrelated numbers without any clue what real power consumption during that benchmark was. you need to divide score by ACTUAL power consumption during test.

Literally, EVERY Tech Reviewer uses this method

probably that's why people ignore many reviewers

In short, I'm correct in everything that I said.

i'm getting strong nikob vibes

[DontFearTheFuture]

Cinebench R23 does run Natively on Apple M Processors

'running natively' doesn't mean 'optimized'
it's right there on Cinebench homepage, support was added only in cinebench 2024

Therefore, every score I gave in my post was based on SUSTAINED performance in R23 running in a loop

no, you've grabbed some random value called 'TDP', that is unrelated to power consumption and is actually there for thermal design - and started crunching unrelated numbers without any clue what real power consumption during that benchmark was. you need to divide score by ACTUAL power consumption during test.

Literally, EVERY Tech Reviewer uses this method

probably that's why people ignore many reviewers

In short, I'm correct in everything that I said.

i'm getting strong nikob vibes

This is right from the article:

"During the multi-core tests, the processor consumed ~21 watts for a short while, after which it steadily decreased and balanced out at 10.5 watts during 12 minutes of continuous load (R23 Multi: 8,237 points, so ~19 % less"

The same way notebook check got it's numbers is the same way I got my numbers. A Sustained score of 8,237 when consuming 10.5 watts.

If you don't understand this.... we'll, just tell notebookcheck they know nothing about computers.

[lmao]

This is right from the article:

yep i was wrong on this one

[lmao]

This is right from the article:

i mean i'm admitting wrong only to the fact that you've actually did use real consumption numbers. i was thinking you are picking them from the top off your head, because i got used to M3 having PL1 of 20W - forgot it's Air, not Pro.

the rest still stands - you used wrong benchmark, your conclusion has no causality.

[DontFearTheFuture]

This is right from the article:

i mean i'm admitting wrong only to the fact that you've actually did use real consumption numbers. i was thinking you are picking them from the top off your head, because i got used to M3 having PL1 of 20W - forgot it's Air, not Pro.

the rest still stands - you used wrong benchmark, your conclusion has no causality.

So, notebookcheck used the wrong benchmark too, right?

PS, what software is optimized for Meterlake?  None of the Cinebenches are.

Is Windows even fully optimized yet?

[lmao]

notebookcheck used the wrong benchmark too, right?

are you surprised, they've even used R15 and R20 for M2 and M1 and are still using non-native games as 'gaming benchmarks'

Meterlake

meteorlake is x86 architecture, cinebench is based on intel's own opensource rendering library, so don't worry about that

[DontFearTheFuture]

notebookcheck used the wrong benchmark too, right?

are you surprised, they've even used R15 and R20 for M2 and M1 and are still using non-native games as 'gaming benchmarks'

Meterlake

meteorlake is x86 architecture, cinebench is based on intel's own opensource rendering library, so don't worry about that

And yet Nootbookcheck stillgives  MacBooks the highest scores in their reviews; and rightfully so.  They deserve those scores.  They are amazing machines.

That's because they know how to enterpret the numbers, see the overall picture.

[lmao]

That's because they know how to enterpret the numbers, see the overall picture.

it's because apple silicon is powerful enough to push even through outdated tests.

i have my own ideas why they were/are picky about tests, and are always making it look like r23 is some kind of a primary benchmark (which it's not by a mile, it's cpu-only rendering, no one does that irl), and don't use the same becnhmark routine for every laptop, e.g. intel laptops tend to not have cb 2024 results in reviews:
www.notebookcheck.net/Asus-Zenbook-14-OLED-review-The-1-2-kg-subnotebook-with-120-Hz-OLED-and-Core-Ultra-7.805236.0.html#toc-5

[Reza]

Hi,
Does it use temporal dithering?