Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

uller6

macrumors 65816
Original poster
May 14, 2010
1,075
1,805
I picked up a 16GB/1T M1 MBA yesterday, and this thing is incredible. I can't believe the performance per watt of the M1. Given that I got the high end MBA with 8 GPU cores, and the lower end MBA only has 7 GPU cores, I started thinking about how Apple does performance binning of these chips and wanted to start a discussion.

Apple doesn't explicitly promise CPU wattage, exact performance, or frequency of these CPUs, so I suspect there's plenty of room for performance variation and some M1s will actually perform better than spec.

My take is that the MBAs actually have the highest binned (lowest leakage current) chips in them, to enable lowest power operation in the fanless MBA. Those that are still capable of very low power, but have a GPU defect are sold as the low end MBA with 7 GPU cores.

Then, the M1 MBP has the "second level" of chips, which have 8 fully working cores but use slightly more power than the MBA M1 chips to hit the performance requirements.

Last on the totem pole are the M1 Mac mini chips, which demonstrate slightly higher performance but require more power to do so.

Or, do I have it completely wrong, and Apple is instead binning chips for ultimate speed, as long as they hit the required power envelope for each application?
 
I think it’s more likely that Apple uses power binning to differentiate the 7 core and 8 core GPU on the MBA. The GPU core itself is very small, so relying on defects to bin is probably not effective. The idea is to take M1 chips that are actually less power-efficient and disable one GPU core to keep the overall power in check.

I also believe that the performance differences between different models simply boil down to the differences in cooling capacity. This is further demonstrated by people who mod their MBA to achieve performance identical to Pro/Mini. Apple probably doesn’t need to do much binning with these chips to begin with, because all of them run with very conservative clock limits. Most of them could probably boost past 3.5 ghz and still have better power consumption than x86 chips. Besides, it has been speculated that because of its mobile background, M1 power management is much more sophisticated that Intel or AMD do, with tighter control over individual components. This would make traditional frequency binning less important.
 
My take is that the MBAs actually have the highest binned (lowest leakage current) chips in them, to enable lowest power operation in the fanless MBA. Those that are still capable of very low power, but have a GPU defect are sold as the low end MBA with 7 GPU cores.
This would imply that if Apple has a good manufacturing run of GPUs without defects, they wouldn't be able to sell any low end Airs.

I don't think the Air and Pro have different versions of the same chip since each can be had with 7 or 8 cores.
 
Whoops, you're correct. For some reason I thought you could get a 7 core when I was looking at prices.

Still, is there really a noticeable performance difference between the 8-core chips in all three computers?
I’ve seen at least one report on these forums of an under performing M1 in an Air. Tested on Cinebench and getting about 10% less than the reported average. But I think that was likely a manufacturing defect with thermal paste or something else in the cooling system; not directly to do with the CPU.

The M1 Macs with fans get better performance under load with the mini performing best on average. Again, this is the cooling system more than CPU variations.
 
Last edited:
I don't believe those who say that the M1s are binned to get 7 or 8 GPU cores. There is so much more to those chips than the GPU cores. They contain 16 billion transistors. Are we really supposed to believe that everything works but one of the GPU cores?
 
I think it’s more likely that Apple uses power binning to differentiate the 7 core and 8 core GPU on the MBA. The GPU core itself is very small, so relying on defects to bin is probably not effective. The idea is to take M1 chips that are actually less power-efficient and disable one GPU core to keep the overall power in check.

I also believe that the performance differences between different models simply boil down to the differences in cooling capacity. This is further demonstrated by people who mod their MBA to achieve performance identical to Pro/Mini. Apple probably doesn’t need to do much binning with these chips to begin with, because all of them run with very conservative clock limits. Most of them could probably boost past 3.5 ghz and still have better power consumption than x86 chips. Besides, it has been speculated that because of its mobile background, M1 power management is much more sophisticated that Intel or AMD do, with tighter control over individual components. This would make traditional frequency binning less important.
This is an interesting idea, and one I hadn't considered. Disable 1 GPU core on the lowest end models for a power savings of ~10% if the rest of the M1 components use 10% more power, thus keeping the entire package within the power budget.
 
  • Like
Reactions: jdb8167
I don't believe those who say that the M1s are binned to get 7 or 8 GPU cores. There is so much more to those chips than the GPU cores. They contain 16 billion transistors. Are we really supposed to believe that everything works but one of the GPU cores?

Exactly why I think switching off one core to save power makes more sense.
 
  • Like
Reactions: chabig
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.