Apple A9X the new mobile SoC king

[Rakehellion]

I disagree. Apple admitted that the results were genuine. They then went on to say the test isn't representative of real usage.

But the difference might be significant for someone who isn't the 'typical' user.

The difference is only significant for a Geekbench user.

 

[erunion]

Normal usage, 2-3% difference, but if you test geekbench continously, the difference is quite big.

Right, but its conceivable that for some people, who's "normal use" includes more demanding CPU usage, the difference will be higher than 2-3%. We'll have to wait and see.

edit: The question is how many people fall outside "normal use".

 

[erunion]

The difference is only significant for a Geekbench user.

And any other app that peaks CPU performance

 

[Rakehellion]

And any other app that peaks CPU performance

For several hours continuously.

 

[erunion]

For several hours continuously.

Continuously has nothing to do with it. The test showed, and Apple has just admitted, that the Samsung SOC draws more power than TSMC at peak performance. Apple's point was that the SOC is at peak usage for only a small percentage of the day so total battery life is barelly effected. Plus the SOC is only one of many components drawing power anyways.

Even if Apple is right and its only an academic difference in real world usage between the processors, academic differences is exactly what this forum discusses! ( And maybe they are right, I'm playing devils advocate)

But rushing to sweep this under the rug presents as an attempt at damage control.

 

[Eug]

Of course it's damage control. They're doing so because so many people are trying to make a mountain out of a molehill.

 

[ShintaiDK]

Of course Apple has to downtone it. It doesn't matter if its 2-3% or 20-30% as the tests show. Its to avoid people returning Samsung SOC based phones to get the "best one" with a TSMC SOC.

 

[khon]

Whether is it 2-3% or 20-30%, the bottom line is that the tsmc SOC seems to be better, and noone wants to buy a premium product like an iPhone, only to find out that they got the inferior version.

If it was just normal chip to chip variation that would be one thing, but this is a systematic difference, and Apple is now admitting it.

 

[davygee]

Whether is it 2-3% or 20-30%, the bottom line is that the tsmc SOC seems to be better, and noone wants to buy a premium product like an iPhone, only to find out that they got the inferior version.

If it was just normal chip to chip variation that would be one thing, but this is a systematic difference, and Apple is now admitting it.

I agree, this is an issue. I just don't know why Apple didn't get Samsung to provide the chips for the 6S and TSMC to provide for the 6S Plus, or the other way about. Considering they are different chips. So in short they are not the both the A9 chip, they may be effectively the same chip, but it seems obvious that they are different sizes and drain different power, so to me they are the A9a and A9b.

Personally, I don't have an iPhone 6S or 6SPlus, but I would be very unhappy if I had the Samsung version.

It doesn't matter if there is just a 2-3% difference or not...there is now a public knowledge that you can get 2 different chips inside the phones and one "it seems" is inferior to the other.

I also am not impressed by the claims "in normal use". I'm sorry, but everybody uses their phones differently. I know how much it drains the system by using VLC or playing Vainglory and going by reports, heavy CPU-based tasks could show huge difference in heat and battery consumption. I use my phone for watching videos using VLC and play power intensive games, so should I expect up to 2 hours less battery with a Samsung-based 6s compared to the TSMC version?

Anyway, I won't be upgrading until May next year, and I would like to think that Appl have sorted out this mess and just go with TSMC chips in their phones by then.

 

[ShintaiDK]

The performance/watt between TSMC and Samsung is mindboggling tho. When considering the rest of the phone. We talk something like up to 50% difference on the SOC.

I can fully understand if people with Samsung SOCs may feel like second class customers.

 

[Nothingness]

Are people really drawing conclusion about a process based on a sample size of one or two?

 

[Eug]

I agree, this is an issue. I just don't know why Apple didn't get Samsung to provide the chips for the 6S and TSMC to provide for the 6S Plus, or the other way about. Considering they are different chips. So in short they are not the both the A9 chip, they may be effectively the same chip, but it seems obvious that they are different sizes and drain different power, so to me they are the A9a and A9b.

Personally, I don't have an iPhone 6S or 6SPlus, but I would be very unhappy if I had the Samsung version.

It doesn't matter if there is just a 2-3% difference or not...there is now a public knowledge that you can get 2 different chips inside the phones and one "it seems" is inferior to the other.

I also am not impressed by the claims "in normal use". I'm sorry, but everybody uses their phones differently. I know how much it drains the system by using VLC or playing Vainglory and going by reports, heavy CPU-based tasks could show huge difference in heat and battery consumption. I use my phone for watching videos using VLC and play power intensive games, so should I expect up to 2 hours less battery with a Samsung-based 6s compared to the TSMC version?

Anyway, I won't be upgrading until May next year, and I would like to think that Appl have sorted out this mess and just go with TSMC chips in their phones by then.

If you're using vlc on an iPhone on a regular basis and you're concerned about battery life, then you're Doin' It Wrong.

vlc is way worse than other options available when battery life is a concern. Other apps like nPlayer and Infuse Pro overall make much better use of hardware acceleration, which means that battery life is far more dependent upon stuff like screen power, since the CPU isn't actually working very hard when you're using the right application.

Meanwhile, a few of my friends have the Samsung and they're perfectly fine with that. I haven't bothered to check for my wife's 6s yet, but I might do so later just out of curiosity, assuming that the App Store app works for it. I'm not going to install a non-app-store app for this. But there's no way in Hades I'd return the phone over this.

BTW, as I mentioned before, there were both 45 nm iPad 2 and 32 nm iPad 2 units in stores. The 32 nm version had considerably longer battery life, and that was a much bigger real-world battery life difference than with the iPhone 6s. There was no chip-gate back then about it.

 

[Arachnotronic]

Are people really drawing conclusion about a process based on a sample size of one or two?

The process delta between TSMC 16FF+ and the Samsung 14nm used to build the A9 is very real, unfortunately for those of us who got stuck with the Samsung-built A9s

 

[NTMBK]

The process delta between TSMC 16FF+ and the Samsung 14nm used to build the A9 is very real, unfortunately for those of us who got stuck with the Samsung-built A9s

I'd be happy to take your defective phone off your hands :biggrin:

 

[kimmel]

You're holding the phone wrong.

The difference is only significant for a Geekbench user.

If geekbench is a good benchmark then the difference is significant to anyone. Are you suggesting that the workloads are artificial and not applicable?

 

[Nothingness]

The process delta between TSMC 16FF+ and the Samsung 14nm used to build the A9 is very real, unfortunately for those of us who got stuck with the Samsung-built A9s

Oh I don't doubt there's a difference. But if the Samsung process was that bad wouldn't there be issues with GN4/5 and S6 phones? Or perhaps it isn't that Samsung process is bad, just that TSMC process is incredibly good

 

[Thanatosis]

I've been using my samsung 6s for two weeks now and I've had no issues with heat or battery life. It's at least as good as advertised.


I think the differences shown in the geekbench battery test are not representative of real-world usage or performance. The test probably runs the CPU/GPU as fast as possible while dimming the display to as low as it can go and even doing that for several hours the difference isn't even 25%.

If I hadn't checked mine with the app, I would have been 100% sure I had TSMC. It performs great, but I guess not quite as good as TSMC chip. Oh well.

 

[kimmel]

I've been using my samsung 6s for two weeks now and I've had no issues with heat or battery life. It's at least as good as advertised.

And I've handed my iphone 6s (no idea what chip is in it) to my wife and she commented on how warm/hot it was and I wasn't really doing anything other than browsing the web, recording some videos and updating some apps. Battery life is merely ok but it's always so hard to tell with new phones compared to old ones since inevitably you are playing with the new shiny toy more than the old phone you are replacing.

 

[Arachnotronic]

I'd be happy to take your defective phone off your hands :biggrin:

Hah! Despite the Samsung chip, the phone is awesome

 

[witeken]

Idontcare / semiconductor question

Can someone help me, I just realized something I hadn't realized before...

The thing that is normally associated with performance, is Lgate or the length of the gate.

Well, what I realized is that I always thought it was the length of the channel, but the channel is of course not called gate.

So here's my question: am I indeed wrong? Are Lgate and Lchannel different? Then what is Lgate actually?

 

[Rakehellion]

If geekbench is a good benchmark then the difference is significant to anyone. Are you suggesting that the workloads are artificial and not applicable?

No, maxing out the CPU is not a good test for mobile battery life. The most common test is wifi web browsing, which is still contrived but more realistic.

 

[Thala]

Are Lgate and Lchannel different?

Gate and channel are different things, but their length should be the same.

 

[Thala]

If geekbench is a good benchmark then the difference is significant to anyone. Are you suggesting that the workloads are artificial and not applicable?

The benchmark certainly is good, because it amplifies the impact of CPU power on the observation. Geekbench achieves this by choosing an extreme workload, which, while good for comparison, is not reflecting typical use-cases.

 

[stingerman]

Whether is it 2-3% or 20-30%, the bottom line is that the tsmc SOC seems to be better, and noone wants to buy a premium product like an iPhone, only to find out that they got the inferior version.

If it was just normal chip to chip variation that would be one thing, but this is a systematic difference, and Apple is now admitting it.

IF TSMC is better than Samsung THEN
APPLE.Advantage
ELSE
APPLE.Advantage
ENDIF

 

[kimmel]

No, maxing out the CPU is not a good test for mobile battery life. The most common test is wifi web browsing, which is still contrived but more realistic.

It's a simple test for exposing differences between silicon. It may not be what you or Apple are looking for in a test, but it is informative nonetheless.

If we get a larger sample size and the data holds up then sure there is a superior and inferior version. You may not care, but not caring doesn't make the data go away.