Most benchmark tests support Apple's claim that differences in battery life among A9 chips is 2%-3%
In the real world, an iPhone 6s with the TSMC chip was getting more battery life, although not as much as the benchmark tests showed. And Apple issued a statement that said that the actual variation could be 2% to 3%, which would be within the range tolerated by the manufacturer.
New benchmark tests conducted today by Ars Technica do show that the TSMC built A9 offers a slight advantage in battery life, just as Apple has been saying. In a test to measure battery life while browsing on Wi-Fi, an iPhone 6s stuffed with Samsung's chip provided only 2.2% fewer minutes of battery power. On GFXBench, the difference was 4.3%, again in favor of TSMC. Testing the battery using WebGL, the phone with the Samsung chip outscored one with the TSMC A9, but only by 1%. The one unusual result took place on Geekbench. Testing the battery on that benchmark site, the TSMC powered iPhone 6s outscored the iPhone 6s powered by Sammy's A9 by 28%.
According to Ars Technica, as long as Samsung's chip is idling, or not struggling with a heavy CPU or GPU load, the battery life will be within the 2% to 3% variance that Apple mentioned. But once the processing units are hit with some heavy duty tasks, the Samsung chip will consume more battery life than the TSMC A9, by more than the 2% to 3% that Apple says is okay. That explains the huge difference with the Geekbench test. Keep in mind that every effort was made to have the settings on both phones match. But some components inside the phones have multiple sources other than the SoC. Each test was run twice, and the results were averaged.
We've already explained how you can determine whether your iPhone 6s is using an A9 built by TSMC, or employing an A9 produced by Samsung.
source: ArsTechnica via AppleInsider