iPhone 6s A9's and battery life

[Oyeve]

Apple already solved that problem via its hiring practices.

hehehe.

 

[kpkp]

I linked to an article that sources Apple as saying that the real-world performance difference comes down to 2-3%

That 2-3% from Apple can't be read as a fact but as PR, they have a conflict of interest and as such they have no say, otherwise why even review phones, let's just watch keynotes and ads and decide form that.

There are 2-3% tolerances even in battery capacity. That number is just pointless.

 

[lopri]

2 hours difference out of how many hours? Surely it is not 2~3%? I suppose this is why Apple had avoided mixing chips until the 6s.

 

[poofyhairguy]

hehehe.

You laugh but if Anand was still around by Monday we would have the EXACT percentage of difference in battery life, plus some throttling difference that only he would have the fortitude to discover.

Without him or Brian on the scene the mobile market is like a wild west town that doesn't have a sheriff.

 

[Trader05]

You laugh but if Anand was still around by Monday we would have the EXACT percentage of difference in battery life, plus some throttling difference that only he would have the fortitude to discover.

Without him or Brian on the scene the mobile market is like a wild west town that doesn't have a sheriff.

This.

 

[lopri]

Knowing by first-hand experience, I respectfully disagree.

 

[Mopetar]

That 2-3% from Apple can't be read as a fact but as PR, they have a conflict of interest and as such they have no say, otherwise why even review phones, let's just watch keynotes and ads and decide form that.

There are 2-3% tolerances even in battery capacity. That number is just pointless.

Or it's the result of average performance deviation from the norm once all factors have been considered. They're the only one with enough samples to have an accurate picture. Otherwise you have to consider everything that could possible influence the results, like battery capacity, difference in screens across multiple manufacturers, etc. Until you can isolate and control for all of those variables, you can't reasonably test SoC efficiency.

And if you're not going to trust them on this, I'd definitely suggest not using their finger-print scanner as they might be lying about that or other security features. Hell you can't reasonably trust any company about anything then as it's all just PR.

Do you have a good reason that you don't believe Apple? Hell, the person who posted the initial 2 hour difference could have been lying as well. How do you know that's the truth?

2 hours difference out of how many hours? Surely it is not 2~3%? I suppose this is why Apple had avoided mixing chips until the 6s.

My guess is that they haven't needed to until now as one company could supply enough of their demand for components, at least in terms of chips. But given that they're selling more and more devices and that they tend to make their SoCs some of the largest on the market, it's not a surprise that they need to source from multiple vendors at some point. That both Samsung and TSMC are working with new processes also points to lower than expected yields so Apple may have needed to diversify to ensure they could meet shipping targets.

 

[kpkp]

And if you're not going to trust them on this, I'd definitely suggest not using their finger-print scanner as they might be lying about that or other security features. Hell you can't reasonably trust any company about anything then as it's all just PR.

Do you have a good reason that you don't believe Apple? Hell, the person who posted the initial 2 hour difference could have been lying as well. How do you know that's the truth?

Messing up fingerprint security would hurt Apple, so it's in their best interest of doing it right, but with all the NSA tech and resources caution is welcome.
Selling iPhones with both SoCs, without returns, is in Apple interest, so they have no incentive of telling the truth if the truth might lead to returns and unsatisfaction. It's as simple as that, follow the money.

All this started based on complains that under load certain SoC is significantly less efficient. Apple "rebuttal" does not address that and because of that it looks even more like damage control PR.

 

[poofyhairguy]

But given that they're selling more and more devices and that they tend to make their SoCs some of the largest on the market, it's not a surprise that they need to source from multiple vendors at some point.

Sure, the iPhone got so big that it's impossible to source from one vendor. With that said if Steve Jobs was still alive the Samsung one would have had a 5-10% bigger battery so no one would have found out the secret.

 

[Mopetar]

All this started based on complains that under load certain SoC is significantly less efficient. Apple "rebuttal" does not address that and because of that it looks even more like damage control PR.

I'm willing to bet if you took 50 iPhones all from either TSMC or Samsung, you'd be able to find at least one pair that have a similarly high performance gap as did the tested devices that started this whole mess.

If you look at overclocking threads you'll find the same thing. Some chips perform a lot better than others, and they're from the same manufacturer.

Also, does it really matter if TSMC can run GeekBench3 for an extra ~10-20% longer if no one does that? Sure Apple gets to pick their own definition for "Average User" to mean whatever looks best for them, but if in typical performance cases you only see a 3% difference, does it matter if someone can demonstrate an extreme benchmark that makes one come out better than the other?

The issue I'd be more concerned with are the reports of some iPhones overheating or the home button becoming overly warm to the touch. That has more of a tangible impact to end users than a small difference in battery life.

 

[Mopetar]

Sure, the iPhone got so big that it's impossible to source from one vendor. With that said if Steve Jobs was still alive the Samsung one would have had a 5-10% bigger battery so no one would have found out the secret.

Someone would eventually figure it out and that would be an even bigger mess.

 

[Midwayman]

Sure, the iPhone got so big that it's impossible to source from one vendor. With that said if Steve Jobs was still alive the Samsung one would have had a 5-10% bigger battery so no one would have found out the secret.

iFixit, or other teardown would have found that out pretty quickly I'm sure. The only difference is users would have less reason to be annoyed. It could possibly end up with "Why do I have a smaller battery if a larger one could fit."

 

[kpkp]

I'm willing to bet if you took 50 iPhones all from either TSMC or Samsung, you'd be able to find at least one pair that have a similarly high performance gap as did the tested devices that started this whole mess.

If you look at overclocking threads you'll find the same thing. Some chips perform a lot better than others, and they're from the same manufacturer.

Also, does it really matter if TSMC can run GeekBench3 for an extra ~10-20% longer if no one does that? Sure Apple gets to pick their own definition for "Average User" to mean whatever looks best for them, but if in typical performance cases you only see a 3% difference, does it matter if someone can demonstrate an extreme benchmark that makes one come out better than the other?

The issue I'd be more concerned with are the reports of some iPhones overheating or the home button becoming overly warm to the touch. That has more of a tangible impact to end users than a small difference in battery life.

It's up to Apples quality control to eliminate those 2 iPhones not the costumers, so you are basically betting on Apples bad QC.

That's binning, that has always happened but there were no major complains from iPhone owners before, I suspect because the differences were never this big. I am totally open that this could go either way, a bad SoC rev, Samsung process is worse, internet complaining on few bad samples, bad benchmarking practices... But believing Apple PR and dismiss everything, nope.

I think loading the CPU to 30% isn't exactly an extreme use case, I suspect some games do that, so to that "gamer" it would matter.
Again you are using those 3% as some use the Bible.

Anyone is free to care about what they want, I find the SoC controversy way more interesting, but you can open a new topic about the home button.

 

[Artdeco]

Hmmm, my phones overheated the other day. Think I'm going to ask for an exchange

 

[gar655]

Et tu Samsung, Et tu...

 

[chin311]

I hope my 6S coming isn't a samsung :O

 

[Mopetar]

It's up to Apples quality control to eliminate those 2 iPhones not the costumers, so you are basically betting on Apples bad QC.

You also need to consider that the GeekBench test wasn't a strict test of the CPU performance. Because we measured battery life, the results would also depend on the power efficiency of the screen as well as the capacity of the battery (in addition to the efficiency of the wireless baseband if that were turned on as well).

Imagine you have a scenario where one phone has a CPU that has a screen, CPU, and battery that are each 7% less efficient than the other phone. Taken together that translates into a ~20% perceived decrease in battery, but the CPU is only responsible for one-third of the total result, and that's assuming that the CPU is eating as much battery as the screen, which in phones with an LCD panel tends to use up more power than anything else.

That's why it's even more pointless to base this around a limited number of devices, especially when we're not not able to directly test how much juice the CPU is drawing.

Some of the AT crew have a write up that points this, and other factors out rather well.

 

[Midwayman]

I hope my 6S coming isn't a samsung :O

That's what the "S" on yours stands for.

 

[lopri]

Maybe it is Samsung's evil master plan to discourage consumers from buying new iPhones.

 

[lopri]

Does anyone know what Geekbench battery test actually test?

 

[Ns1]

Does anyone know what Geekbench battery test actually test?

According to the AT link 2 posts up...

The GeekBench test runs parts of the GeekBench CPU benchmark in a loop, making sure to do a fixed amount of work per time interval while idling the rest of the time, and using the score result as a modifier for the runtime score. This makes the GeekBench battery life benchmark primarily a SoC/CPU/Memory benchmark, and that in turn has repercussions for interpreting the data.

 

[lopri]

I did not know there was an AT article. Thank you, Ns1.

 

[Ns1]

I did not know there was an AT article. Thank you, Ns1.

Mopetar found it -I actually did not see it on the main website until the 2nd time I tried looking. Put new articles in a small sidebar to the right, instead of on the main table column - web design brilliance.

 

[openwheel]

You laugh but if Anand was still around by Monday we would have the EXACT percentage of difference in battery life, plus some throttling difference that only he would have the fortitude to discover.

Without him or Brian on the scene the mobile market is like a wild west town that doesn't have a sheriff.

Word

 

[lopri]

It could be a firmware issue as well. A few Youtube videos leave impression that Samsung chips might be overvolted, which generates more heat and in turn reduces energy efficiency. If so, whether that is to ensure consistency among varying qualities of silicons or Samsung-made A9s on average require more voltages is going to be difficult to determine without painstaking investigation of a decent number of samples. Either way not the best showing by Apple. From the looks of it there is a display lottery as well - displays with greenish/yellowish white (jade-like tone), and displays with bluish white. Brightness seems to have gone up which is good but with a side effect of visible backlight grids around the edges of screen.

Probably not a big deal for the vast majority of folks, though. Apple's statement has no real revelation other than "Nothing to see here, folks. Move along" which is understandable.