iPhone 6s: Samsung And TSMC A9 SoCs Tested

Status
Not open for further replies.

Pei-chen

Distinguished
Jul 3, 2007
1,279
5
19,285
Good review and write up. I have the same exactly thought when "the news break" that some YouTuber find one phone with a TSMC A9 lasts longer than the one with Samsung A9. Putting aside the problem with sample-size-of-1, those YouTuber have no idea how binning works and assumes all TSMC/Samsung chips are exactly the same. Those of us that've been around Tom's and Anand's a long time knows even two identically binned chips would show difference in power draw and those clueless YouTuber have no idea.

Add in the fact the boards are full of chips, battery at 100% still draw power and screen at same brightness will have different power draw it is safe to say difference between TSMC and Samsung A9s are overblown.
 
"extending battery life between 3.5-10.8 percent. This is a little more than the 2-3 percent quoted by Apple, but not much, and it equates to only about 5-15 minutes of runtime under the most extreme conditions."

Well it's roughly 100-200% more than what Apple quoted - that is quite a lot different isn't it? And 5-15% under extreme conditions doesn't seem like much, but how much more is it under typical workloads? Wouldn't that equate to be as much as an hour of extra battery over the day if you get a Samsung manufactured A9?
 
G

Guest

Guest
Just what I needed to know.
iPhone 4: if u hold it a certain way, u lost most of ur signal.
iPhone 6: bends if u put it in ur jean pocket.
iPhone 6s: CPU-maker luck-of-the-draw affects battery life.
 

MobileEditor

Reputable
Nov 13, 2014
92
0
4,640
Just what I needed to know.
iPhone 4: if u hold it a certain way, u lost most of ur signal.
iPhone 6: bends if u put it in ur jean pocket.
iPhone 6s: CPU-maker luck-of-the-draw affects battery life.

Well, it's still better than the Samsung camera lottery :) All of the latest Galaxy phones come with either a Sony or Samsung made rear camera sensor that are supposed to be equivalent. In our Galaxy S6 review, however, we show that there's a notable difference between them.

- Matt Humrick, Mobile Editor, Tom's Hardware
 

MobileEditor

Reputable
Nov 13, 2014
92
0
4,640
"extending battery life between 3.5-10.8 percent. This is a little more than the 2-3 percent quoted by Apple, but not much, and it equates to only about 5-15 minutes of runtime under the most extreme conditions."

Well it's roughly 100-200% more than what Apple quoted - that is quite a lot different isn't it? And 5-15% under extreme conditions doesn't seem like much, but how much more is it under typical workloads? Wouldn't that equate to be as much as an hour of extra battery over the day if you get a Samsung manufactured A9?

No, I do not consider 5-15 minutes after 80-100% discharge to be a lot. Also, this is a worst-case scenario where the CPU or GPU runs at max frequency all the time. During normal use, the CPU/GPU will be at their idle frequencies most of the time. Because of the nonlinear relationship between voltage and frequency, the difference should be less than what we measured during normal operation, which is the scenario Apple claims only shows a 2-3% difference. If that's true, then a use case where the battery lasts 8 hours would show about a 15 minute advantage. In the Basemark OS II test where the CPU is at max frequency for the duration of the test, there's only a 20 minute difference until the battery is depleted (extrapolating from where the test ends with 20% battery remaining).

Unfortunately, our iOS test suite does not contain any tests that mimic a real-world usage scenario. I did perform a simple 1080p H.264 video playback test and saw a 2-3% advantage for the TSMC A9. This is not necessarily applicable to the CPU, however, because I believe the A9 uses a fixed-function hardware block for video decoding that would use a different transistor type optimized for lower frequency operation.

- Matt Humrick, Mobile Editor, Tom's Hardware
 

Chris Droste

Honorable
May 29, 2013
275
0
10,810
"extending battery life between 3.5-10.8 percent. This is a little more than the 2-3 percent quoted by Apple, but not much, and it equates to only about 5-15 minutes of runtime under the most extreme conditions."

Well it's roughly 100-200% more than what Apple quoted - that is quite a lot different isn't it? And 5-15% under extreme conditions doesn't seem like much, but how much more is it under typical workloads? Wouldn't that equate to be as much as an hour of extra battery over the day if you get a Samsung manufactured A9?

most smartphones average between 5.5 and 6.5hours of on screen time with app, text, email, and talk usage in the real world. even if you had a "golden sample" you're not likely to get more than 7hours in practical use, and that will vary downward in most cases. The result? you'd have to use your "special" iPhone 6S like your grandparents who don't even know what a mobile app is to reap anything truly noteworthy. If you really must have battery life, Verizon will be happy to sell you a Droid XXX Turbo/super/special/R/S MAXX phone with massive battery.
 

Sywofp

Reputable
Oct 13, 2015
2
0
4,510
One thing - why didn't you guys test using the Geekbench battery life benchmark?

Considering it is the one that started this whole debate, it seems like it would be the first test to run...

Arstechnica found similar results to yours, except for Geekbench, where the Samsung performed a lot worse.

So the important question IMO, is why that benchmark gets such a big difference, and does it reflect any real world usage scenarios - even if not experience by most people.
 

MobileEditor

Reputable
Nov 13, 2014
92
0
4,640
One thing - why didn't you guys test using the Geekbench battery life benchmark?

Considering it is the one that started this whole debate, it seems like it would be the first test to run...

Arstechnica found similar results to yours, except for Geekbench, where the Samsung performed a lot worse.

So the important question IMO, is why that benchmark gets such a big difference, and does it reflect any real world usage scenarios - even if not experience by most people.

We did not run the Geekbench battery test because it's not part of our usual test suite, and we're not familiar enough with how it works or with the accuracy of its results.

I saw the Ars Technica article. It seems a bit odd that every test except Geekbench battery shows minimal difference, while Geekbench shows a 20+% gap. This is a red flag to me that something might not be working right with this test and is another reason we did not include those results.

- Matt Humrick, Mobile Editor, Tom's Hardware
 

Sywofp

Reputable
Oct 13, 2015
2
0
4,510
I absolutely agree that it is a bit odd, which is why IMO it needs more testing!

Fair enough though that it's not a benchmark you guys are familiar with. Hopefully the Geekbench guys themselves are working on it. (And it is not like I have the skills or the resources to make any impact aside from commenting :p)

But the problem is that until someone explains the outlier result, then there isn't really an accurate conclusion to be drawn on the whole Samsung vs TSMC debate.

Not including the results lets you draw accurate conclusions based on your tests, but not the situation as a whole. Which really is what people want answers about.




 
G

Guest

Guest
Just what I needed to know.
iPhone 4: if u hold it a certain way, u lost most of ur signal.
iPhone 6: bends if u put it in ur jean pocket.
iPhone 6s: CPU-maker luck-of-the-draw affects battery life.

Well, it's still better than the Samsung camera lottery :) All of the latest Galaxy phones come with either a Sony or Samsung made rear camera sensor that are supposed to be equivalent. In our Galaxy S6 review, however, we show that there's a notable difference between them.

- Matt Humrick, Mobile Editor, Tom's Hardware

Agreed.
The thing with cell-phones-as-an-appliance model is that this kind of structure works well with Apple's philosophy, which is vertically structured to control all aspects of the hardware and software design; that's why iPhones are fascinating to use. The rest of the mobile industry is not so lucky. Someone will always find some 'thing' to point out to apple with every iteration and say the proverbial "oops...missed a spot". The reality is that the perfect mobile device does not currently exist. Apple comes close with them fancy camera optics and Retina display. The public's expectations then become unrealistic. Their minds make the rhetorical assumption that "my fridge is perfect... I put salami in it..it comes out cold...Why Can't My Cell Phone Be Like That".

-David (iPhone 4s owner)
 

wdgann

Distinguished
Sep 12, 2013
64
0
18,630
In Arstechica article, TSMC scores higher in GFXBench test, while this article shows Samsung scores higher. Which one should we believe?
 

Medianguy

Reputable
Oct 14, 2015
1
0
4,510
Did you measure the CPU frequency while doing BaseMark OS II Full ? I guess the CPU should have been throttled because it ran at 100% load for a long time. The SoCs wouldn't be throttled on Geekbench battery test because it put only 30% load on the CPU.

Edit) The ambient temperature around Samsung SoC iPhone was 82F but TSMC's was 84F on the photo. Was your imaging IR thermometer calibrated?
 

George Phillips

Reputable
Jun 17, 2015
614
0
5,360
My iPhone 6s Plus comes with the TSMC chip, and the new iPhone has so far exceeded my expectation in terms of battery life. It just runs all day without charging. At the end of the day, there is usually at least 40% left of battery life! I am glad I chose the Plus version instead of the 4.7" version.

Anyone who is wondering what they should get should seriously consider iPhone 6s Plus just for it's battery life alone.
 
I personally think overall Samsungs 14nm might have a slight edge of TSMCs 16nm but the silicon lottery will be the real deciding factor. Some Samsungs A9s will be better or worse than TSMC's A9s and vice versa. Till testing is done over a large number of phones in a controlled manner it is really hard to know.
 

Modern processors do not just run at a specific speed and voltage. They throttle their speed or scale their voltage based on how the CPU is performing. How well the CPU performs depends on random factors like contaminants in the silicon when it's being extruded, atomic-scale defects in the transistors, etc.

That's why so many other posters are calling this the silicon lottery. The differences you're measuring aren't differences between the Samsung vs TSMC processors. They're differences between the individual processors themselves. That is, if you were to test two Samsung or two TSMC iPhones, you'd still see this type of variance in the results.
 

MobileEditor

Reputable
Nov 13, 2014
92
0
4,640
Did you measure the CPU frequency while doing BaseMark OS II Full ? I guess the CPU should have been throttled because it ran at 100% load for a long time. The SoCs wouldn't be throttled on Geekbench battery test because it put only 30% load on the CPU.

Edit) The ambient temperature around Samsung SoC iPhone was 82F but TSMC's was 84F on the photo. Was your imaging IR thermometer calibrated?

Right now we only have the ability to measure CPU load % and not actual CPU frequency on iOS. We have monitored CPU frequency while running the Basemark OS II battery test on Android, and yes, we do see most CPUs throttle back frequency and even shift from the big cores to the LITTLE cores.

The temperature around the phone is influenced by the ambient air temperature and the heat conduction from the phone into the table. The longer the test runs, the warmer the table gets until it reaches equilibrium. The ambient temperature was the same for both phones in this test, since they were run at the same time. The thermal imager self calibrates and is discussed in our "How We Test Smartphones And Tablets" article.

- Matt Humrick, Mobile Editor, Tom's Hardware
 

MobileEditor

Reputable
Nov 13, 2014
92
0
4,640
That's why so many other posters are calling this the silicon lottery. The differences you're measuring aren't differences between the Samsung vs TSMC processors. They're differences between the individual processors themselves. That is, if you were to test two Samsung or two TSMC iPhones, you'd still see this type of variance in the results.

This is true, and in the article I briefly discuss the concept of processor binning. Two A9 SoCs made by Samsung, for example, running at the same frequency may require different core voltages. This will certainly add some scatter to the data points.

On top of this, though, is the difference in process technologies, which almost assuredly have different voltage-frequency curves and leakage currents. Is a sample size of one sufficient to separate the difference in power consumption due to manufacturing variance from process technology, or even determine if one version is "better" than another? No. But our data, together with good data collected by other testers, can ballpark the difference and assuage any fears that there's huge differences between the two versions.

It would be great if we could get 100 iPhones (or whatever a statistically valid sample would be) to get better data, but we do not have $65,000+ to get that many iPhones :) Maybe we could start a Kickstarter campaign.....

- Matt Humrick, Mobile Editor, Tom's Hardware
 

Hah hah. Also, by the time you completed testing that many phones, the next phone model would be out, rendering the test results moot.

Perhaps a healthier way of thinking of this is that the differences are so small, it's not worth your time worrying about them. If the phone, tablet, computer, whatever does its job to your satisfaction, that is good enough. I still think it's amazing that a device I can hold in one hand and lasts the entire day on a single charge is more powerful than the desktop computer I was using 10-15 years ago.
 

George Phillips

Reputable
Jun 17, 2015
614
0
5,360
Another very crucial factor is that not all batteries of the same model have the exactly same capacity. Some will have slightly more capacity; some others will have slightly less capacity. Unless the sampling size is many times more, this kind of tests are highly inconclusive.
 

Dan415

Reputable
Oct 14, 2015
1
0
4,510
Do you think 6s and 6s plus will have a different outcome? I've read through several test results and found out that TSMC are favorable in terms of battery life if the test model is 6s (actually I don't remember who has used 6s plus as a test model). If you can also conduct a test on the 6s model, that would be awesome!!!
 

Daniel Ladishew

Reputable
Apr 16, 2014
66
0
4,630
I think there is a question here that seems to have been lost in the battery life debate. How does each of these chips age over time? We all know Samsung makes high quality products that you can rely on for usually well beyond they warranty. I wonder if TSMC can keep up with Samsung in the durability and endurance comparisons. Even if it's only slightly different, the TSMC chip may not be as reliable over the long term given it's different manufacturing process. Because this is the core component of the phone I feel like consumers are going to want to know if there is a perceptible difference. Please develop and start a 'lifespan' cellphone test series.
 
Status
Not open for further replies.