Max Tech, running the M2 MacBook Air on just the efficiency cores

[pshufd]

Running only Efficiency cores is not always the way to save power. When running the fast cores you will run to idle quicker as a task can be finished in a smaller timeframe. So the CPUs can sit longer at even a lower power level as they are idling longer. I haven't tested this with M1/M2 chips but in the Intel days, a faster CPU on a MBP was often using less power than a slower one for similar tasks as it was sitting at idle longer.

I've run my M1 Pro MacBook Pro on Low Power mode since I bought it in 2021. I would assume that Low Power is designed to use less power. When I look at the core activity, the Efficiency Cores are prioritized over the Performance Cores. I suspect that I would be fine with only Efficiency Cores.

 

[jwahaus]

On his blender test, he says that it uses 79% less power, but his figures are 4.81w normal and 2.97w efficiency. That’s more like 39% less power.

Still impressive, but not quite to the extent he is suggesting.

Not sure about your math but I calculate that it uses ~62% less power: 2.97 * 1.62 = 4.81

 

[jay-A]

Not sure about your math but I calculate that it uses ~62% less power: 2.97 * 1.62 = 4.81

No - starting from 2.97 watt 4.81 is indeed 62% **more**, but what we're doing here is that we start from 4.81, and in this respect 2.97 is 38% **less**. Anyway, Max Tech should spend less time flexing every tiny muscle of their mouth and use this time to check their math instead.

 

[Vaccaria]

This is so far away from the truth it couldn't be any further. Genshin Impact can only and exclusively run low resolution at 60 FPS without throttling.

Anything higher than that - whether medium res or heaven forbid high - lags all the way down to 30-40 FPS nonstop in battles/fights/demanding overworld scenarios etc. And it takes minutes, not hours.

It's literally a topic that has been beaten to death in hundreds of posts all over the internet, and I can assure you that no one who plays Genshin regularly is gonna recommend you an M1 iPad to run it on because the game lags to the point of unplayability if you select any resolution other than low.

EDIT: Not that low looks terrible, it doesn't - but it's a very far cry from crisp and clean looking graphics, with lots of pixelation on most objects in the FOV.
That game almost guaranteed won't be playable on High res + 60 FPS until M3 or later.

I'm fully aware how GI can be a performance hog. And I never stated that it runs at 60hz, in fact:

I tried it on the 12. There is some occasional stuttering on max settings, therefore I would advise turning some settings down. Also keep in mind that energy saving mode reduces your refresh rate to 30hz.

Why would I consider to run it at 60hz when the iphone screen reduces the gameplays fluidity to 30hz anyways?

 

[Gerdi]

Isn't the headline totally misleading? Low Power mode does not disable the performance cores - in fact they still play an important role with respect to the performance of the device.

Other than this, the guys from Max Tech making themself look like idiots, if they just figured out, that lowering the voltage and frequency leads to much higher efficiency. But apparently it is "blowing their mind".

 

[Sydde]

You want a 16” MacBook Air. That’s a completely different class of computer. Maybe Apple should make one but not at the expense of the 16” MacBook Pro.

Edit: They would need to take out one performance cluster which would leave you with an M2 4/4.

This makes no sense. There is no rule that says a P-cluster has to have 4 cores. One P-core takes up about the same diespace as 4 E-cores, so if you wanted to go heavy on the E-cores, you could easily pull two Ps and have an 8E/6P configuration that would divide the Ps into 2 clusters of 3 or better still 3 of 2 and still have space left over, perhaps to double the E-cores' L3.

 

[TSE]

I know the guys from Max Tech from middle school.

They bullied me relentlessly for having a laptop with a Pentium 3 processors while they were using G4s. They would tie me up to the flagpole at recess, forcibly grab my laptop from my backpack, and run synthetic benchmarks of my Pentium III against their G4 laptops and show off how much better their laptops scored versus mine in front of all my classmates.

It made me feel pathetic and worthless.

 

[maternidad]

On his blender test, he says that it uses 79% less power, but his figures are 4.81w normal and 2.97w efficiency. That’s more like 39% less power.

Still impressive, but not quite to the extent he is suggesting.

You both mean different things. The difference is 39% of 4.81, but 4.81 may be divided by around 179% to produce 2.97.

 

[Sydde]

but 4.81 may be divided by around 179% to produce 2.97.

No, it cannot. 4.81 / 1.79 is about 2.69. Since division is reflexive, 4.81 / 2.97 is more like 162%, which is far enough from 179% as to be unmistakably not similar.

 

[locksmack]

I stand my ground. The video said % LESS, so it is indeed ~39%. Arguing otherwise means you are making the same mistake they made in the video, and it is misleading.

 

[Gerdi]

I stand my ground. The video said % LESS, so it is indeed ~39%. Arguing otherwise means you are making the same mistake they made in the video, and it is misleading.

It is not misleading but just wrong. There is no room for different interpretations, if the term "less" is used, as you correctly pointed out. So "39% less" it is and nothing else.

 

[1096bimu]

Running only Efficiency cores is not always the way to save power. When running the fast cores you will run to idle quicker as a task can be finished in a smaller timeframe. So the CPUs can sit longer at even a lower power level as they are idling longer. I haven't tested this with M1/M2 chips but in the Intel days, a faster CPU on a MBP was often using less power than a slower one for similar tasks as it was sitting at idle longer.

One would assume Apple knows which scenario uses less power and thus makes it "low power mode"

 

[maternidad]

No, it cannot. 4.81 / 1.79 is about 2.69. Since division is reflexive, 4.81 / 2.97 is more like 162%, which is far enough from 179% as to be unmistakably not similar.

I don’t understand how they obtained the value, then! Sometimes, though, we say that a number is two times smaller than the other —and it does not mean that the other value is the additive inverse of the first. I was thinking that maybe they said 79% in reference to that the ratio approached 1 : 2, but as you’ve found out, the relationship is more like 2 : 3 (1 : 1.5).

 

[tmoerel]

This is a seriously fantastic video result. You can use Low Power Mode on the M2 Air and it results in drastically lower temperatures when maxed out. It uses a tiny amount of power compared to using the performance cores.

Perhaps those guys should try to understand the Dunning-Kruger Effect!

 

[Technerd108]

The M1 Pro/Max is a great chip but it would have been even better if it were a 12 core cpu with 4 efficiency cores. No need to change the power cores. On the M2 Pro/Max it would be great to see 14 cores. 10 power and 4 efficiency cores. It would kill in multi core and have better battery life. Even an 8 power core and 6 efficiency core design would be pretty sweet.

As it is now the 10 core cpu with only 2 efficiency cores is great but I wouldn’t mind a little better battery life on the M1 max chip. I think they might do this either on M2 or M3 adding more cores while still keeping the heat envelope in check. Keeping more power cores than efficiency cores but at least adding more efficiency cores might be all they need to do to keep an advantage over the competition.

Intel uses a different strategy by using only two or four power cores and 8 efficiency cores and they seem to get a lot of performance out of them so I think a decent efficiency core combined with decent power cores can lead to some really powerful designs. Obviously there are a lot of differences between Intel x86 and Apple arm but just illustrating a point that adding a lot of efficiency cores can be a good idea.

 

[quarkysg]

The M1 Pro/Max is a great chip but it would have been even better if it were a 12 core cpu with 4 efficiency cores. No need to change the power cores. On the M2 Pro/Max it would be great to see 14 cores. 10 power and 4 efficiency cores. It would kill in multi core and have better battery life. Even an 8 power core and 6 efficiency core design would be pretty sweet.

As it is now the 10 core cpu with only 2 efficiency cores is great but I wouldn’t mind a little better battery life on the M1 max chip. I think they might do this either on M2 or M3 adding more cores while still keeping the heat envelope in check. Keeping more power cores than efficiency cores but at least adding more efficiency cores might be all they need to do to keep an advantage over the competition.

I would think more circuitry added into the M1 Pro/Max die would increase power usage rather than decrease it. Likely Apple decided than 2 E-cores would be enough to hit the power sweet spot.