Question Apple A14 Bionic - Disparity between iPad Air and iPhone

[bobmans]

Earlier today I was looking at the iPhone 12 GeekBench scores and I noticed that the iPhone's A14 compute (Metal) score is significantly lower than the iPad's A14 score compute score.
The Single/Multi core scores basically equal but the disparity in the compute (Metal) score is huge. The A14 in the iPad scores even higher than the A12Z in the pro and more than 30% higher than the iPhone's A14.
I don't think a disparity this huge can be explained by the iPad having better thermal management because for some tests the compute score is equal and for some it's more than double that of the iPhone. The A12 iPad also had an almost equal compute score compared with the A12 iPhone's despite having better thermal management.

https://imgur.com/WZnowM3

Direct score comparison:

iPad13,1 vs iPhone13,2 - Geekbench

Anyone has any thoughts on how this is possible? According to GeekBench both chipsets seem to be clocked at 2.99Ghz so I doubt the iPhone's A14 is underclocked.

 

[IntelUser2000]

Geekbench's compute is basically testing the GPU.

With CPU you can sort of cheat by using burst modes since most usage scenarios are bursty.

With GPUs you can't since the only thing that really matters is long term performance. Therefore it'll run into thermal/power limitations which means a smaller device will perform worse.

 

[Doug S]

It might not be an actual thermal limitation (i.e. where it first gets hot enough that it is forced to clock down) instead it might be programmed to simply draw more power for the GPU on the iPad. That would make sense, given that it has access to a much larger battery as well as better ability to dissipate heat.

There's not much point to racing the GPU along at a rate they know will heat up the phone and have to throttle it. Some of that is unavoidable of course, due to different environments, different ways of holding it that transfer more or less of one's body heat to it, etc. so it needs to have the ability to throttle but you hope the "starting speed" can be maintained over time in the right setting (lower ambient temperature, not getting a lot of heat from your hands, not using cellular, etc.)

Letting it start at a speed that would soon force throttling even on an iPhone being used outside in 30F by someone wearing gloves would be pointless, whereas the much larger iPad with its much larger mass and surface area might be able to run the GPU at that speed at room temperature and held with bare hands for extended periods no problem.

 

[moinmoin]

In addition to heat and battery GPU performance may also be tweaked to the amount of pixels they have to feed on each device. iPad Air 4 has the most of the devices containing A14.