Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
My real world experience with TSMC vs Samsung
- Thread starter Markiie
- Start date
- Sort by reaction score
Care to learn more about the difference between the two CPU dies from Samsung and TSMC? Here is the link, I am sorry that it's in traditional Chinese which is from a Taiwanese site (not China). You probably don't know there are two Chinese language in written, but that doesn't matter.
http://tu0925399900.pixnet.net/blog/post/204868429-為何台積電16奈米a9處理器會勝過samsung-14奈米
The title says "why TSMC 16nm is superior to the SAMSUNG 14nm". I am not sure if you want google to translate them for you or not since you are anti-google from your avatar. But I am sure those diagram can help.
Majority of your arguments is saying that there is not enough sample. Well, as you know APPLE has the sample that you think is large enough, but they won't announce the true difference between the CPU. They want to stir it all so you won't get real logical answer.
You missed the point. And many of the forum members missed the point. You think the CPU difference under normal people's regular usage should resulted in being less than the claimed 20-30%. Common sense of people can agree with you. Statistical significancy in between the TWO CPUS is undeniable. And the answer is not from statistics, it's from the foundry difference in between the two companies if you can see from the diagrams of the site I linked. And the 14nm by SAMSUNG is copied or replicated, or so called brought by previous employee of TSMC, but the leakage of the die is something other than what's easily copied. Therefore, SAMSUNG A9 has less power efficiency in this case
Why the first few blog would like to test the CPU difference once they got new iPhone 6s, and 6s plus? Because that is what they are interested in. And its also a test that can be single out rather than mixing all other factors, radio chip, antenna, battery, apps. If there were different radio chip, and someone is interested in knowing the difference, they can do it. It doesn't need that much sample size as you claimed, hundreds, or thousands, because all other factors can cancel out or omitted. Have you got math problem in middle school by giving you only 5 variable to find an average? Anantech did so many first hand test between only two CPU, GPUs, who as asked him "please not fair, need more sample".
Your knowledge should tells you when you mix those variables, it cancel out most by each others. For example y = x1+x2+x3..... In this case, people only want to know x1, (cpu for example) and they know there are two CPU makers,and are interested in finding the difference x1 in contributing y, the battery life, that's all.
So, the buttery life Y, can be affected as much as 2 hrs by the cpu using the geek bench app. Isn't that crystally clear.
If I am a heavy gamer, that's the most factor that my regular usage of the phone being affected. You claim yourself waiting for more daily normal usage is actually trying to blur the picture while people already focusing on. So don't argue with saying it needs more regular, normal usage comparison. That's not something meaningful to APPLE, and by that you are helping APPLE to cover that big difference in between the two CPUs.
Last edited:
Half your post is utterly nonsense....
You missed the point, entirely.
Those are not two chips. Are the same chip , manufactured by two different foundries, using two different processes. But still the same chip designed by the same engineering team (Apple, not TSMC or Samsung).
So far there are no demonstrated performance difference between the twos, as expected (being the same chip). There are some differences in power consumption, being the thermal behavior of the SoCs different. We don't know how much different, but Apple does.
The 2 hrs difference so far was demonstrated only by a test on ONE DEVICE, a number statistically ridiculous
First of all a round of applause from me how you the put the issue its terrific. Usually I don't read long post but yours is fun and very informative to read. It could be possible apple wanted to avoid lot of exchanges that's why they may have put these numbers. Now a days consumers like us can do battery test like these how come company like apple come up with their results? They may have the most advanced diagnostic devices. Apple needs to cut Samsung off from now on, it's Samsung who has created this mess. Just like they fired Scott forestall for map issue they need to exclude Samsung for their hardware parts.
lol I like your burning reply
Last edited by a moderator:
LOL, by your mean, those are not two chips? but are the same chip, fine. Then you said manufactured by two different foundries. Yes, that's the point. So they are not the same. People found they are not having the same power efficiency. Dude.
"Statistically ridicules". How many sample do you think Anandtech every time can test a new generation Intel, AMD or Nvidia chips for you? one, maybe two, that's it. So people still buy their site's story? This time Anantech choose standby APPLE's side, at least for what they have posted. That's ridicules, too.
My point is. You can only know how much the CPU power consumptions between the TWO chips from the TWO foundry by single out other factors. The APPLE way to stir the whole thing into regular daily usage ignore the significancy between the TWO different foundries' chip's thermo and power efficiency. Consumers who found their iPhone with SAMSUNG chip is less power efficient, they have right to exchange for one with TSMC chip because they are paying the same ridiculous high price for an iPhone 6s or 6s Plus. And why not? And I think lots of them are doing so. APPLE has to accept it by all means.
In some ways Apple has created unreasonably high expectations for themselves. They're the ones that say perfection, precision, etc. Their marketing department is wonderful at their job. But yeah, if you're charging $1000 for a smartphone, the customer should have every right to be as picky as they want within the 14 days. You got to take the bad with the good. So I say if you want to exchange for a TSMC, go for it, while you are within your 14 days. The high price of the phone accounts for expected return rates...and even their returns will sell back through the used market and Apple will still make a profit on them, just not quite as high. And I do not for a moment believe that Apple would lower their prices if say, by some miracle there were no returns because buyers suddenly banded together and said "let's stop returning so the iphone 7 will be $100 less". No, Apple would just keep the price as high as possible.
If you're a video editor, or a gamer, or someone who uses the camera alot. That's processor intensive use and there will be a big, big difference. You definitely want a TSMC phone unless you simply don't care about 1.66 hour battery difference. And that's with heavy use. With light use I'd suggest that the differences are far more than 2-3%.
The way I see it, the TSMC is not any worse than the Samsung, and most likely better, though by how much is debateable. So TSMC it is.
Last edited by a moderator:
I'm the less credible user to defend Samsung here, believe me, but we have to PROVE that Samsung 's chip is less efficient yet.
What I saw on the net aren't a proof. Not even close to that.
Speaking about apples and oranges...
Performance comparison between different devices, you can take a few samples. This is what anandtech and others usually do.
But when you are questioning the whole production you need a statistically credible base, and only Apple could have it.
I'm saying that since the beginning: TSMC is maybe the best in the field. Since a while...
There's already some evidence that Samsung doesn't have the efficiency of TSMC. The increased heat is a big one. What energy that doesn't do 'work' on a chip is converted into heat.
Why do I get the feeling unless it comes out of Apple's marketing team you wouldn't believe any testing? Anandtech usually only gets one device to test. Not 100's. 3Dmark is a popular synthetic benchmark that gpu owners like to use to compare their rigs against each other. Yes, it's not a game but it's been proven again and again to come pretty close to predicting the difference in performance between gpus. No one is going to have the money to test 100's of iPhones to compare the power usage so we're left with the community to find it out. Another example is nvidias fiasco with the GTX 970. Majority of users won't feel the difference but us who gamed on 1600P/4K on SLI could feel the stuttering as Nvidias crippled the last 512MB of vram. Their marketing dept claimed the difference was negligible, but real world use at high resolutions showed constant stuttering until I replaced my 970 SLI setup with 980's.
Anandtech usually does a pretty good job of approximating use of a smartphones in their battery life testing. Hopefully they at least get the chance to compare a TSMC/Samsung 6S during their benchmarking.
Last edited:
Apple marketing team doesn't count on the matter.
Who knows is Apple engineering team, collecting data from millions of phones....
When you are comparing performance, a single unit of each model could be an acceptable approximation ... But here we are comparing two units of the same model, to highlight differences. Little variations could exist and you need a number statistically big enough to say something definitive about that.
I'm quite sure TSMC chip are more efficient, but how much ? Is it noticeable in real life ?
i agree in real world we don't use phone so heavily maybe someone but not me.
Almost nobody is going to max cpu and gpu from full battery to dead in the way this "test" was designed. If you read the newest news thread you can see many more bullet points for said testing done by a reputable source (though they still seem to have only compared two units). The TSMC chip is still better, if we are dead set on picking a winner, but the results aren't a flat as overwhelmingly in favor of the TSMC chip as this thread would have you believe. In fact, the conclusion was Apple didn't lie.
All three of our phones are TSMC chips by the way. I have absolutely no reason to be rooting for the Samsung here
Lol.
On a side note I'm always astonished on how people here could believe Apple is lying.
They are a big multinational company, they are under the magnifier every day, they have shareholders to report, they have customers to keep loyal.... A company like that could avoid comments, could make press conferences to appear in a good light, but they can't just lie. It would be a suicide
A polished PR release that claims real world use, while the option of a much bigger battery drain is very real, is basically lying, or hiding the truth.
Except respected independent experts at Ars, having done their own testing, agree with them. Are Ars lying too?
Except that I don't buy the real world use excuse. One test still drains more than 30% on the Sammy chip. This test may amount to some real world situation such as games, shooting video, and I'm sure many other functions people might do. A 1000$ phone suppose to be made the same for all consumers, period.
Geekbench remains the only test of any kind that shows such a margin. I haven't seen any evidence that it corresponds with any kind of real world use, whereas most of the other genuine attempts at "real world use" (eg gaming and video recording etc) come up with results much closer to the other benchmarks. At this point I think the onus is on Geekbench to explain why its results are so different.
Lol, you're too much of a believer of the corporate religion. Corporations will always lie, or at least distort the truth, when it means more profits or less of a loss in the end. Look at Volkswagen (diesel cheating) and GM (too cheap ignition switches), both thought by cutting corners they could save costs and increase profits. Even the almighty Apple continues to regularly dispute that their logic boards with DGPUs have a reliability problem, only after class action lawsuits are threatened do they begin to extend warranties to try to address the issue. BTW, this has already happened in 3 separate instances!
Last edited:
except two competent independent websites confirmed the "PR release" ....
But we should trust an amateurish youtube video and some OCD bashers on forums, right ?
A single test.
And in a single test the Samsung version is slightly better.
Are you going to run Geekbench all the day with your iPhone or are you going to actually use it ?
The test surely ISNT a representation of real world situation, since you don't usually deplete your battery in less than 3 hrs in real life.
I'm still waiting for a credible test demonstrating such a difference in real world usage.
VW is paying an high stake for that lie.
Do you think Apple's executives are stupid ?
thanks. i really appreciate the efforts to calm down various anxieties here.
i will be on the 4K shooter and HD footage handling user spectrum,
so my reasons to feel it is unfair are more founded.
i won't check my chip (i hope) when i get the phone.
the only thing that will really ease my mind is seeing a test when shooting 4K and stuff.
if that will show 5-10% difference, i'll feel much more relaxed.
if this test will show 20-30%, i will feel very pissed.
and i don't live in US, so the many "easy" replacement options are non existent for me.
I know you are worried and as you said you might have trouble swapping it if you get the Samsung chip, well as luck would have it I got one with it and posted my impressions with it in the other thread and thought I would share them in the situation it might help
https://forums.macrumors.com/thread...hone-for-a-tsmc.1926937/page-20#post-22074795
as far as I'm concerned either way you are getting a great phone, after all this is a phone in the same form factor that is nearly twice as fast as last years and even with the Samsung chip gets the same battery life. that's pretty impressive!. I hope it works out for you anyway and you end up being happy with your phone whichever chip you get
It's been replicated by several different people and tech sites. There is a difference and loading the cores at 30% will give you 2 hours less battery life. No one has been able to get a samsung phone to run same length as TSMC in that test. In doesn't matter what does it in the phone, it will happen on samsung vs TSMC phone.
I don't thing at this point anyone else needs to run the geekbench test, it's very well established that the Samsung phones do considerably worse on that test, it's not a question of the sample size or statistical error, it's an eminently repeatable result.
What is still very unclear, as I and others have said before, is what actual bearing that result has on anything other than geekbench.
None of the other tests or situations people are trying to emulate indicates anything like that kind of difference between the chips. With a few exceptions (and history tells us there are always people with terrible iPhone battery life out there, whatever the iteration) the majority of Samsung users actually using their phones seem happy. The "real world" data, including the results of other benchmarks, is all a lot closer to Apple's 2-3% than Geekbench's 20-30%. So the question is, is geekbench revealing a real problem that will effect anyone, or is it simply revealing a flaw in the test that Geekbench has devised, as a measure of real world battery life?
What is still very unclear, as I and others have said before, is what actual bearing that result has on anything other than geekbench.
None of the other tests or situations people are trying to emulate indicates anything like that kind of difference between the chips. With a few exceptions (and history tells us there are always people with terrible iPhone battery life out there, whatever the iteration) the majority of Samsung users actually using their phones seem happy. The "real world" data, including the results of other benchmarks, is all a lot closer to Apple's 2-3% than Geekbench's 20-30%. So the question is, is geekbench revealing a real problem that will effect anyone, or is it simply revealing a flaw in the test that Geekbench has devised, as a measure of real world battery life?