Controlled Testing Supports Apple's Claim of Nearly Equal iPhone 6s Battery Life With TSMC and Samsung Chips

[firewood]

Chips from any single manufacturer will vary from lot to lot.

What happens if you try to trade "up", and it turns out that you traded a Samsung A9 chip that ran on the fast and cool end for Samsung processors, and instead got one of the power hungrier TSMC A9s (but still within Apple's specs)?

Your typical battery life would get worse instead of better!

 

[Benjamin Frost]

Ars Technica's testing is worse than the real life tests that have been reported here and elsewhere. Testing in a controlled environment is worse than real life because it is so limited. No one uses their phone in such a manner.

There have been many authoritative reports of substantially inferior battery life by Samsung 6s owners. These have much more validity than one laboratory test of two phones.

ChipGate is alive and well.

 

[rtr1985]

A win's a win. Roll tide!

 

[ulyssesric]

Apple just said "up to ??? hours" for some tasks in general in spec which means Apple doesnt promise anything about battery at all. How to meet a spec when there isnt one?

There is a standard spec and test model for all silicon chips called 'process corner'. The test procedure involves variation of supply voltage and environment temperature. When you outsourcing silicon chips, you must design the specs and setup threshold for different corner. If the test results pass designated target, that shipment is acceptable.

Of course silicon chips varies from chip to chip. There are superior and inferior individuals even within the same shipment. So chances are that you'll have a chip that runs well in high temperature corner, while the other one from the same shipment will perform poorly. To overcome this situation, designers will tend to use over-spec design. This is why you may sometimes 'overclock' your computer, while others would fail, for the same model.

The truth is: Apple choose Samsung as provider in the beginning. The reason why they choose Samsung is quite simple: TSMC's 16nm production line is not ready in early 2014. I knew that because I was participated in TSMC's plan. Apple had designed the spec, and Samsung had met with the standard. But Samsung failed to meet the quantity due to low yield, so Apple transferred some of the order to TSMC. However, it turns out TSMC chip is outperformed Samsung chip in some circumstance.

In other words, the performance behavior of iPhone 6s with Samsung chip is considered as 'standard' model. iPhone 6s with TSMC is 'superior' than original design.

This is exactly what happened here.

 

[BobZ521]

Even though most people are probably not going to notice a significant difference, I look at how much less time I'm going to get from my phone for the same money and how much more it will cost me over time. If Samsung lasts even 3% less on a charge per day by Apples own testing, this will add up to additional charges per year which adds to the cost to own. The additional charge cycles that will now be required for the same amount of usage will also deplete the overall life of the battery. The real test seems like it may be in a year or two of usage which by then know one will probably care.

 

[NoahK17]

So basically you get 14 EXTRA minutes of Safari browsing when using the non-Samsung chip. Perfect!!

 

[JackANSI]

Yeahhh...well I for one, I don't want to send my phone to some random poster on the internet. Will I really get one back in return?

Name your escrow service, I'll pay the fee.

No matter what, there won't be a single taker. Because it is a real issue. Just give it up.

 

[j2ee]

There is a standard spec and test model for all silicon chips called 'process corner'. The test procedure involves variation of supply voltage and environment temperature. When you outsourcing silicon chips, you must design the specs and setup threshold for different corner. If the test results pass designated target, that shipment is acceptable.

Of course silicon chips varies from chip to chip. There are superior and inferior individuals even within the same shipment. So chances are that you'll have a chip that runs well in high temperature corner, while the other one from the same shipment will perform poorly. To overcome this situation, designers will tend to use over-spec design. This is why you may sometimes 'overclock' your computer, while others would fail, for the same model.

The truth is: Apple choose Samsung as provider in the beginning. The reason why they choose Samsung is quite simple: TSMC's 16nm production line is not ready in early 2014. I knew that because I was participated in TSMC's plan. Apple had designed the spec, and Samsung had met with the standard. But Samsung failed to meet the quantity due to low yield, so Apple transferred some of the order to TSMC. However, it turns out TSMC chip is outperformed Samsung chip in some circumstance.

In other words, the performance behavior of iPhone 6s with Samsung chip is considered as 'standard' model. iPhone 6s with TSMC is 'superior' than original design.

This is exactly what happened here.

We can simplify say samsung chip is the suck version

 

[ElZeus]

I thought it was the standby time that was much better with the TSMC version. Why didn't they test that?

 

[dampfnudel]

You guys are blowing this way out of proportion. If you ever notice a difference, it'll be trivial. It's important for Apple to source chips from multiple companies, and it's too expensive sometimes for both companies to use the same manufacturing process. Just deal with it and move on.

As far as I can tell, this is the first time Apple sourced an iPhone CPU from two companies and that makes it a big deal. In a way, this could be viewed as an experiment by Apple. Either that or Apple needed to do this to avoid serious shortages. In other words, this could just be a temporary fix until TSMC can produce enough to satisfy Apple's A9 needs and the Samsung A9 will be given the boot at that time. That boot could come sooner now if the pressure continues to build.

 

[hasanahmad]

in term of battery performance..my new ip6s+ with TSMC chip outperformed my ip6s+ with Sammy chip. don't ask me why but from my personal experience I can see there is a huge difference in battery duration... the Sammy chips drained so much power to the point that I had to carry a wall charger at all time

Fake


One post

Drive by troll


Please ban

 

[hasanahmad]

Everyone who is saying there is a vast difference are actually too dumb to ignore that all the tests except geek bench test the phone in normal use which is web browsing, gaming, graphics intensive apps, and scripting apps. Geek bench is not a battery benchmark, it ups the chip to 60-70% usage which will never happen in real world usage even during games and movies because its benchmark is design for short bursts of cpu performance not 4 hours of cpu maxed out. It's a unrealistic use

The only thing closer to the geekbench benchmark is you editing a 4K file which is an hour long and which takes 4 hours to process with high cpu usage; and yet still both will process the whole edited file


Anyone who thinks anything else is a power usage has either no idea how usage works or are really dumb or are simply trolling. Opening Instagram and maps and an hour of candy crush and Twitter and email at once is not the definition of a power useage

 

[dampfnudel]

16nm and 14nm are the same generation in terms of silicon manufacture procedure. The previous generation is 20nm/22nm and the next is 10nm. 'Non mature' is not the reason for higher power usage. 'Non mature' will only lead to low yield, which will affect gross margin of the enterprise. And that's also the reason why Apple transferred part of the order to TSMC.

I've no evidence, but I'd guess that the problem of Samsung chip is probably inferrer design of voltage step control.

I read somewhere that the A10 will also be 16nm, not 10nm. At least the A10 from TSMC will be 16nm. I'm not sure if Samsung will also get an A10 order from Apple, but I have a feeling now they won't.

 

[Carotdo]

Fake


One post

Drive by troll


Please ban

you can say whatever you think about me..why dont you check macrumors on FB the current article about Controlled testing .. Look for user name KEITH ANH LE... You can see my return Iphone 6S+ to applestore proof

 

[jlseattle]

Zzzzzz. Just take your phone back to Apple if the battery isn't performing to where it is advertised.

 

[Carlanga]

"kinds of CPU-intensive work that the Samsung chip seems to struggle with just aren't that common on smartphones."

Such as????

Anyways, I still believe the bigger chip is throttling speed to reduce heat.

 

[mkimid]

Everyone who is saying there is a vast difference are actually too dumb to ignore that all the tests except geek bench test the phone in normal use which is web browsing, gaming, graphics intensive apps, and scripting apps. Geek bench is not a battery benchmark, it ups the chip to 60-70% usage which will never happen in real world usage even during games and movies because its benchmark is design for short bursts of cpu performance not 4 hours of cpu maxed out. It's a unrealistic use

The only thing closer to the geekbench benchmark is you editing a 4K file which is an hour long and which takes 4 hours to process with high cpu usage; and yet still both will process the whole edited file


Anyone who thinks anything else is a power usage has either no idea how usage works or are really dumb or are simply trolling. Opening Instagram and maps and an hour of candy crush and Twitter and email at once is not the definition of a power useage

I understand as geek bench use CPU only, and it is just 30% of duty to avoid a throttling, and, high CPU usage can be happening with any games (for AI), and video recording and photo editing or Java/html5 base web pages;
If you compare the runtime via % under battery info in setting, you may can find more deep information

 

[aberrero]

I read somewhere that the A10 will also be 16nm, not 10nm. At least the A10 from TSMC will be 16nm. I'm not sure if Samsung will also get an A10 order from Apple, but I have a feeling now they won't.

Yes, there is no way that 10nm will be ready by next year.

However, the 14/16nm processes should be more mature by then, and the A10 can become a bigger and faster SOC, but it will probably only be slightly faster.

 

[mkimid]

"kinds of CPU-intensive work that the Samsung chip seems to struggle with just aren't that common on smartphones."

Such as????

Anyways, I still believe the bigger chip is throttling speed to reduce heat.

TSMC version is cooler than Sammy

 

[mkimid]

So basically you get 14 EXTRA minutes of Safari browsing when using the non-Samsung chip. Perfect!!

That is depend on how much Java script has used or Html5 any has involved or GIF animation has involved, at least TSMC version is longer than Sammy

 

[hasanahmad]

I understand as geek bench use CPU only, and it is just 30% of duty to avoid a throttling, and, high CPU usage can be happening with any games (for AI), and video recording and photo editing or Java/html5 base web pages;
If you compare the runtime via % under battery info in setting, you may can find more deep information

Yes but 99% of apps don't have cpu useage at peak or consistent for more than a few seconds unless you have a file which is being processed for 4 hours after editing, in games the gpu takes the brunt while cpu often idles while in browsing and in apps the cpu idles on most times Inside the app once the information is loaded,


The geekbench benchmark only would be realistic if you did 4K video editing and then processing it all day everyday on your phone and even then the difference at that unrealistic scenario of real world useage is 20 percent while in realistic scenario it's not more than 1-2%

 

[SlipperySlop]

We can simplify say samsung chip is the suck version

The only thing that sucks is the absolute lunacy displayed here. Even if there is a difference, it won't be of significance to 99% of users. For the 1% of you here, obsessing and driving everyone crazy over your concerns, the difference will be 10 minutes of battery time as you run 14 high cpu games at once.

 

[Aloft085]

says the person with a TSMC chip LOL

I don't know what chip I'll have yet, and I don't really care.

 

[DakotaGuy]

Because it is a real issue.

I'm a combat veteran who served in Iraq in 2004. I can assure you this isn't a real issue. At least not one to stress and worry about. Now I can discuss a real issue, but I don't think you would want to hear the details.

 

[macbookfan]

Quick Apple PR payment to ArsTechnica and we get this.

P.S - How are people saying this isn't a problem. The one test that constantly pegs the phone, Geekbench, shows around a 30% difference between the two chips. There are times when I can use my phone at full load. If I'm commuting, sitting through something boring and play a demanding game etc.

It's not a problem because Apple sells a phone with minimum advertised specs. They both meet these specs. They never promised you the better chip and you don't get to decide.