[Updated] The Apple A9 Silicon Lottery

[Mew]

Things just get more and more interesting. The A9 chip in the 6s and 6s+ are fabbed by both TSMC (16nm FinFET) and Samsung (14nm FinFET). 

 

 

Picking up from our conversation this morning on Apple’s A9 SoC, the crew over at Chipworks has been working hard over the weekend to decap A9 and their first results are in. As it turns out, Chipworks has found chips fabbed by both TSMC and Samsung, which means that A9 is in fact a dual sourced part.

 

In taking apart iPhones and decapping SoCs, Chipworks has turned up two SoCs. The first, APL0898, is a 96mm2 A9 that’s fabbed by Samsung. The second SoC, APL1022, is a 104.5mm2 SoC fabbed by TSMC. And while Chipworks isn’t naming the specific manufacturing processes used, based on Apple’s comments on a “new transistor architecture,” we’re certainly looking at a form of FinFET. In which case the two chips are made on versions of Samsung’s 14nm FinFET and TSMC’s 16nm FinFET processes respectively.

 

Right off the bat, I could see this affecting battery life with the Samsung A9 chip having slightly lower power consumption (as well as possibly heat) than the TSMC A9 chip. I also wonder what's the extent to what the 2 different dies would create- 2 slightly different mainboards? Would one phone be denoted as iPhone 8,1 and the other iPhone 8,2? Or would it be like the panel lottery that existed in their rMBP (lg vs samsung panels I believe) where the only way to see which had what was to look in the code?

 

Source: http://www.anandtech.com/show/9665/apples-a9-soc-is-dual-sourced-from-samsung-tsmc

 

10/7 GIANT UPDATE:

 

It looks like despite the smaller Samsung A9, the TSMC A9 uses much less power and results in significantly better battery life. 

 

(Picture from Macrumors from a reddit poster of the TSMC (left) vs Samsung (right)

 

 

Perhaps the most dramatic result comes from a reddit poster who compared Samsung and TSMC versions of the iPhone 6s Plus using the battery life test included in Geekbench 3, finding the TSMC version lasting nearly two hours longer than the model with Samsung A9 chip. 

 

The article also says that the TSMC scores slightly higher on Antutu and runs slightly cooler but I think the results are negligible. 

 

As of right now, the sample size is really small, with not enough to draw any real conclusions. But if the TSMC chip is able to use significantly less power than the Samsung chip, that might create some problems.....

 

Source: http://www.macrumors.com/2015/10/07/tsmc-samsung-a9-battery-tests/

[The Cool n00B]

I'm confused if the 6S has one and the 6S+ has the other, or is it that you could have either regardless of the phone?

[soaringchicken]

interesting...

[KemoKa]

All it would really be is a slight difference in footprint of the SoC. The die would be different in size, not the interposer or the package.

In any case, it's ARM, ARM's SoCs are notorious for being very power-sippy. This would amount to a negligible difference in power consumption at the absolute worst.

[VagabondWraith]

So is this like Samsung vs Hynix GDDR5?

[TroubleKlef]

wonder what I have. This is definitely going to both me now

[Mew]

Updated post with new findings.

 

[JAKEBAB]

Seriously, dafuq! Is this even legal?

[captain cactus]

One benchmark doesn't tell the story. Unless you can see a big difference when you use the phones like any normal person and do a side-by-side comparison over the space of one month, only THEN can you call one chip better than the other.

[LAwLz]

One benchmark doesn't tell the story. Unless you can see a big difference when you use the phones like any normal person and do a side-by-side comparison over the space of one month, only THEN can you call one chip better than the other.

That makes no sense. You should never disregard scientific tests under controlled circumstances and then trust what people "feel" and "think" after a month of usage. The usage over a month might vary so much that you might actually say the phone with better battery life was worse, just because you happened to do a bit more with it during that time.

When you do tests like that you also introduce the risk of placebo effects.

You can't say something is better because it "felt" like 30 minutes was longer than 35. 35 is objectively longer than 30 and it doesn't matter what people think.

(The numbers 30 and 35 were just grabbed out of thin air to make a point)

 

What we should do is the exact opposite of what you are saying. We should disregard subjective conclusions made under tests with an incredible amount of variables, and we should instead focus on the objective results.

[Imabigmac]

I'm confused if the 6S has one and the 6S+ has the other, or is it that you could have either regardless of the phone?

This actually wouldn't be a bad tactic, as the 6+ has a larger battery and would have more space to use something to help with the heat, but the downside being a slower performer and you need all the power for the 1080p screen. 

 

 

That makes no sense. You should never disregard scientific tests under controlled circumstances and then trust what people "feel" and "think" after a month of usage. The usage over a month might vary so much that you might actually say the phone with better battery life was worse, just because you happened to do a bit more with it during that time.

When you do tests like that you also introduce the risk of placebo effects.

 

 

Well they could have more concerned with the reality vs. Scientific data. Using your data if one phone lasts 30 and the other lasts another 5.35 minutes based on one sample however how repeatable is this with other devices, others user results may vary - This where @lots of unexplainable lag had a point because it is just one sample and not definitive proof just a case. 

 

Then in the end what is 5 minutes longer life when you are comparing battery life in hours anyway. However the results show a much greater spread so this statement is basically irrelevant aside from demonstrating a point that regardless of the accuracy/results of a experiment sometimes the data doesn't mean much to a consumer. But Im sure you know this much. 

[DildorTheDecent]

Seriously, dafuq! Is this even legal?

GPUs have like 3 different types of GDDR5. From "best" to "worst" it is Samsung, SK Hynix, Elpida. 

 

Samsung usually being a better overclocker and out performing SK Hynix and Elpida at the same clock speed. Samsung however is only really found on the higher end GPUs becasue of its traits. 

 

Maybe the higher end iPhones get the better chips...

[Mew]

GPUs have like 3 different types of GDDR5. From "best" to "worst" it is Samsung, SK Hynix, Elpida.

Samsung usually being a better overclocker and out performing SK Hynix and Elpida at the same clock speed. Samsung however is only really found on the higher end GPUs becasue of its traits.

Maybe the higher end iPhones get the better chips...

Well, according to a previous article, there's a 60/40 split in favor of tsmc chips. So it's really a lottery as to what you'll get. The 6s+ is roughly 50/50 iirc.

[syoumans54]

Doesn't seem like a valid test to me... One phone is using cellular data while the other has no sim at all. You need to keep all the variables the same if you want to have a fair comparison.

[Suika]

Doesn't seem like a valid test to me... One phone is using cellular data while the other has no sim at all. You need to keep all the variables the same if you want to have a fair comparison.

Nice catch, and excellent point, but if anything it wouldn't help the Samsung SoC and show an even larger difference between the devices.

[79wjd]

but if anything it wouldn't help the Samsung SoC and show an even larger difference between the devices.

nvm How do you know which one is which from those pictures?

 

/Also, I'm mildly curious what's in my 6s+, is there a way to check that? 

[syoumans54]

Nice catch, and excellent point, but if anything it wouldn't help the Samsung SoC and show an even larger difference between the devices.

Possibly, and I'm not for sure but it looks like location services is turned on as well (I'm an android user), so it's possible that the cellular data is helping it from using the gps as much to find its location. Without knowing everything that's running on the device at that time it's hard to say what would help or hinder the battery life.

[ItsAFeature]

It's an interesting update but you'd obviously have to see a larger sample size before we can draw conclusions. It also seems to me that the battery life is really low on both. Were they running something during the test? Or screens forced to stay on?

[JAKEBAB]

GPUs have like 3 different types of GDDR5. From "best" to "worst" it is Samsung, SK Hynix, Elpida. 

 

Samsung usually being a better overclocker and out performing SK Hynix and Elpida at the same clock speed. Samsung however is only really found on the higher end GPUs becasue of its traits. 

 

Maybe the higher end iPhones get the better chips...

I knew someone was going to bring overclocking into it but here's where that point fails, overclocking is out of spec and not directly from the manufacture, when you buy a device or anything for that matter you expect all the products to be equal if not it's called false advertising, manufactures can't just say yea here is this device it does x,y and z but the device you buy with the same name only does x and y.

[bcredeur97]

wonder what I have. This is definitely going to both me now

it's been bothering me since I bought my phone....

 

My battery life seems short(might just be the new phone "use it all the time" thing). But my phone doesn't run hot (significantly cooler than my iPhone 5 actually, doesnt dim the screen because of heat when I use the GPS in 90F+ weather like my iPhone 5 did)

 

 

I really, really want to know though It's driving me crazy.

[Mew]

nvm How do you know which one is which from those pictures?

/Also, I'm mildly curious what's in my 6s+, is there a way to check that?

In the article there's a link that takes you to an unofficial app that macrumors doesn't recommend you using to see which chip you have.

http://appshopper.com/utilities/lirum-device-info-lite-system-monitor

[VagabondWraith]

Macrumors and other websites across the net is blowing up because in tests shown, the TSMC chips are outperforming their Samsung counterparts. So much so, TSMC chips in tests are showing 20% longer battery life and TSMC runs cooler 36 vs 40 on the Samsung. I know these are subjective and there's many variables involved. But its a little disturbing for people who have Samsung having tests consistently showing TSMC being superior in each and every test.

 

Here's some threads/articles to check out. Some people with Samsung CPU's are going off the deep end with this. Threatening Apple with a lawsuit (good luck with that!) or returning their iPhone's in hopes of getting a TSMC branded CPU.

 

http://forums.macrumors.com/threads/which-chip-does-your-6s-6s-have.1922967/unread

http://forums.macrumors.com/threads/should-there-be-a-samsung-soc-battery-gate-settlement-for-the-iphone-6s.1926575/

http://forums.macrumors.com/threads/iphone-6s-6s-plus-battery-test-results-only.1926607/

http://www.macrumors.com/2015/10/07/tsmc-samsung-a9-battery-tests/

http://www.phonearena.com/news/Samsung-made-A9-vs-TSMC-A9-benchmarking-two-iPhone-6s-versions-yields-surprising-results_id74448

 

Here's a good one from Reddit:

https://www.reddit.com/r/iphone/comments/3nn00t/live_test_results_iphone_6s_samsung_14nm_vs_tsmc/

 

Quite the controversy so far! There's many many more articles but this is the gist of them. Personally, I think its too early to say. I think more testing needs to be done but certainly early tests seem to favor TSMC.

[vetali]

Threatening Apple with a lawsuit (good luck with that!)

Nah Apple will release a patch to fix mask the issue.

 

Hopefully by the time I upgrade there will be more TSMC phones out there. 

[VagabondWraith]

Nah Apple will release a patch to fix mask the issue.

Hopefully by the time I upgrade there will be more TSMC phones out there.

Luckily I got the TSMC version. Apple I believe went to Samsung because TSMC couldn't produce enough chips to meet demand.

I wonder though because I believe TSMC has much more experience in producing chips so they were able to produce better yields from their 16nm FF+ compared to Samsungs 14nm node. You might think a smaller node may yield slightly better performance, but I may be unjustified in that.

Besides, a vast majority of people don't frequent tech forums so it's probably a non issue.

I think too many people are freaking out. I've seen plenty of people report great performance with Samsung and others report subpar performance with TSMC.

[TechStyle]

How can one find out if they have a TSMC or a Samsung chip without tearing apart their phone?