Tests confirm marginal difference between Samsung, TSMC-made A9 chips in Apple's iPhone 6s
New battery test results, published on Monday, appear to support Apple's claim of little difference in the longevity of an iPhone 6s with a Samsung-made A9 chip versus one manufactured by TSMC.
TSMC's chip performed marginally better in Wi-Fi browsing and GFXBench tests, but actually ranked a few points lower in WebGL scores, noted Ars Technica. For the site's comparisons, two SIM-less AT&T phones were used, each set to the same display brightness.
The TSMC model did do substantially better in Geekbench 3, leaping ahead of Samsung by 28 percent. Previously, Apple complained that Geekbench was "unrealistic" because it forces a phone to operate at peak CPU usage until its battery dies.
It's only under extremely CPU-intensive tasks that Samsung's A9 chip can become a detriment, Ars concluded.
Responding last week to various reports, Apple insisted that there should be no more than a 2 to 3 percent battery life gap between any two comparable 6s models. Some anecdotes hinted that the difference could be as much as two hours.
TSMC's version of the A9 is manufactured using a 16-nanometer FinFET process, whereas Samsung depends on a 14-nanometer technique. Smaller process sizes typically result in more efficient chips.
TSMC's chip performed marginally better in Wi-Fi browsing and GFXBench tests, but actually ranked a few points lower in WebGL scores, noted Ars Technica. For the site's comparisons, two SIM-less AT&T phones were used, each set to the same display brightness.
The TSMC model did do substantially better in Geekbench 3, leaping ahead of Samsung by 28 percent. Previously, Apple complained that Geekbench was "unrealistic" because it forces a phone to operate at peak CPU usage until its battery dies.
It's only under extremely CPU-intensive tasks that Samsung's A9 chip can become a detriment, Ars concluded.
Responding last week to various reports, Apple insisted that there should be no more than a 2 to 3 percent battery life gap between any two comparable 6s models. Some anecdotes hinted that the difference could be as much as two hours.
TSMC's version of the A9 is manufactured using a 16-nanometer FinFET process, whereas Samsung depends on a 14-nanometer technique. Smaller process sizes typically result in more efficient chips.
Comments
Dont be silly. There are small variations in all products.
I bet if they tested 2 TSMC phones they would test differently
28 percent is not small.
Oh the horror...ChipBatteryComboGate /s
Thats not scientific proof. They need to test more than just a single phone. More like several thousand phones in a controlled environment.
Just too many variables to take any of those results to the bank.
The difference is significant if CPU is under load, and that was repeated and verified by at least 4 groups.
Being a frequent mobile gamer, I would pick TSMC version any time of the day.
For non-gamers (with what Apple identifies as a "normal use") indeed it would not make a difference.
Any significant differences can be "fixed" in the 9.1 update.
Hello, Volkswagen?
Why isn't everyone trashing Samsung like they do in every other article? The Samsung haters couldn't trash Samsung's LCD and SSD performance which were clearly superior to Apple's other suppliers, and it must have caused them a lot of frustration. But now with the A9 benchmarks, the haters can finally let it all out.
If Apple is extremely honest to the users, it should explain why the Geekbench 3 test show such big difference between the two chips.
Personally, I would consider any, and all, software algorithms used to reduce energy consumption in a mobile device to be proprietary information.
How does TSMC managed to last longer despite being bigger?
You cannot compare absolute sizes unless the process is the same. 16nm FinFET may be more efficient than 14nm Samsung-process. If they both used the same process then the 14nm would probably be more efficient.
Thats not scientific proof. They need to test more than just a single phone. More like several thousand phones in a controlled environment.
Just too many variables to take any of those results to the bank.
The previous test also shows the same difference. Not every measurement is statistical. You know that. When scientist measured the speed of light, no one said he needs to make several thousand measurements.
If Apple is extremely honest to the users, it should explain why the Geekbench 3 test show such big difference between the two chips.
It would appear you and others are desperately seeking to invent the next #gate. Give it a rest.
Personally, I would consider any, and all, software algorithms used to reduce energy consumption in a mobile device to be proprietary information.
It is probably not software. Why would software run differently for different chips?
Was the battery defective?
Was the chip itself defective?
Is the geek bench test accurate?
Where the phones ran the exact same way?
Does the geek bench 'test' favor certain chip sets?
Just way too many variables to make any conclusions.
And yes the speed of light HAS been tested THOUSANDS of times.
The test is not run on different chip sets. The phones all run with Apple A9 chip sets.
Being a frequent mobile gamer, I would pick TSMC version any time of the day.
Well you can’t and you never will. Apple is not going to allow people to specify which A9 they want. That’s not in the cards.
Obviously you don't know much at all about physics and what is acceptable to claim something as fact.
The previous test also shows the same difference. Not every measurement is statistical. You know that. When scientist measured the speed of light, no one said he needs to make several thousand measurements.
Previous attempts at measuring light speed HAVE resulted in a variety of results btw. As would be expected by any measurement of anything.
http://math.ucr.edu/home/baez/physics/Relativity/SpeedOfLight/measure_c.html
So replicates would be helpful....
Obviously you don't know much at all about physics and what is acceptable to claim something as fact.
You don't know physics. You don't know what I said.
Previous attempts at measuring light speed HAVE resulted in a variety of results btw. As would be expected by any measurement of anything.
http://math.ucr.edu/home/baez/physics/Relativity/SpeedOfLight/measure_c.html
So replicates would be helpful....
Your understanding is better than the other poster. Experimental results all have deviations. Scientists don't throw out their measurements because they are all different. What can be accepted as long as the difference does not vary unreasonably.