Anonymous, 19 Apr 2016Commonsense....If Exynos is indeed a better chip, why did Samsung even waste time over SD. U actually don't know much about anything do u?... :\ Samsung didn't just "waste their time" with the snapdragon they actually wouldn't have even used it if they didn't have a problem with putting the exynos version on sale in places like the US and others that got the snapdragon version. Exynos is made BY SAMSUNG so of course they'd want to have used THEIR OWN chipset. Don't post nonsense u know nothing about
Well, the information I've dug up shows the Mali-T880MP12 (Exynos 8890) running at 650 MHz and 265 GFLOPS and the Adreno 530 MSM8996 (Snapdragon 820、820A) running at 510-624 MHz and 407.4-498.5 GFLOPS. I think Adreno wins.
Commonsense....If Exynos is indeed a better chip, why did Samsung even waste time over SD.
Anonymous, 18 Apr 2016Just because ur getting the snapdragon version doesnt mean u have to act so cool about it. Eve... moreNo, it is not. Like another user said the SD chip outperforms it in single core usage, graphics, and in-app performance. It seems that you don't want to look at the benchmarks that confirm this.
EvanTheTerrible, 04 Apr 2016Snapdragon variant has better Single core performance, way better gpu performance, and a sony ... moreJust because ur getting the snapdragon version doesnt mean u have to act so cool about it. Everyone got really pissed out side america because they thought they will be getting the worse one. Actually the exynos version is better so accept.
Snapdragon variant has better Single core performance, way better gpu performance, and a sony camera sensor. The whole second it takes faster to open up an application for the first time is not nearly as important as "in application performance" in todays high graphics using applications. Therefore I believe the Exynos variant loses out in this round. Yes the multi core performance is a little better, but all around I think the Snapdragon setup is a better setup. Not to mention that it has the slighty faster clocked lpddr4 ram @ 1896 vs. 1794 on the Exynos. If anyone has any argument here I would like to hear it. Thanks ;)
it 's just speed test yet another. not too far but still Snapdragon 820 is better than Exynos 8890
DarkMaster, 26 Mar 2016here is your S6 ;) https://www.youtube.com/watch?v=jE3oiEVliWQ it's lagging piece of cr#p beca... moreI don't need to see it in YouTube, as I experience it everyday in my hands, and I am very pleased. Sorry to see you so bitter for just a phone.
What the heck is it with all the strange little Apple fans comment dropping on a processor comparison for Samsung... seriously???
For an audience that hates Android with such passion... you sure do spend a lot of time reading and talking about it...!!
Have you nothing better to do whist waiting for your battery to charge... :p
Look; it is evident you enjoy paying ott prices for run of the mill hardware with an Apple stamp on it for the convenience of not having to get your hands dirty; doesn't mean you are amazing by any standard - rather pretentious trophy wielding plums with little to no knowledge about what you are actually holding in your hands....
Why not just try an Android for a little while?
I've been dual booting Windows 10 on my Samsung S5 for a couple of months now and must say I'm impressed by the power and functionality offered even by older hardware, so I must profess; I am extremely excited about getting hold of and demoing one of these slinky little rippers!!
Wouldn't mind getting my hands on this new multi-core to see what she truly blows!
Anyway, apologies for the rant!
Now go on - clear off you freakin' Granny Apple tyrants!!!!!!
Anonymous, 27 Mar 2016Haha Qualcomm fan boys are so fun. They probably haven't finished primary school. I always won... moreIt sees not have to look, you have not finished any class, even at primary school. Snapdragon is KING BOSS to all processors.
Anonymous, 28 Mar 2016Well, comparing Apple core design to the rest is literally ... Apples to Oranges. Apple built ... moreYes Apple know exactly what they want and need since they develop both hardware and software, this gives them an edge for sure.
But the basics of application design in regards to multi threading being hard or even impossible depending on the problem you have to solve do not change completely because its running on a different OS.
Again, if developers or the OS in some magic way could make everything take full use of multi cores then we would see the Android phones running circle around A8/9 on the common plattform applications, which certainly isn't the case.
You saying it yourself, does it happen often enough? I know i'm repeating myself but you always gain from high single threaded performance, its not a case of "often enough", its all the time.
Reading forums like Beyond3D and alike (not your standard phone/tech-blog forum that is) is very informative. There are alot of different types of persons with knowledge in this area sharing opinions. People working closer to the hardware than your usual application developers.
Im talking about nVidias own custom core (Denver): http://www.anandtech.com/show/8701/the-google-nexus-9-review/2
And as I said, its a much easier to go the high core count way than the developing a bigger, more powerful core. Thats why the "budget" SoC usually have these CPU designs.
But all companies designing their own custom cores Qualcomm, nVidia and Apple and even Samsung to a certain degree has gone for bigger cores with higher single threaded performance compared to ARMs reference designs (and ARM Cortex A72 is not so bad in this regard either).
It feels like we are going in circles, don't think this will lead to anything :) I guess we have to agree to disagree.
It's been nice to have a discussion on a quite reasonable level, even if you keep nagging about me being a Qualcomm fanboy which i'm certainly not! :)
Anonymous, 28 Mar 2016No, we (I) do not only do "unit" tests on code parts. If your developers do not know where thi... moreWell, comparing Apple core design to the rest is literally ... Apples to Oranges. Apple built their CPUs around the OS/app philosophy of theirs. Whilst that's respectable and works well for them it does not transfer well over the Android isle. Basically that's my point.
Now I have to say it again because it is/was somehow passed by. What you can see in Google Apps (and according to our tests to big 3rd part apps) is not "balancing the load of a single core". If anything Anandtech makes that very clear, exactly because it contrasts that with Samsung's App (which is indeed -mostly- 1-threaded).
What you see is a true multi-threading applications. That is to say that those apps can indeed divide one task into small subtasks. Do they do it perfectly? Not even close (apart from play store updates, where damn-near perfect parallelization happens), but it *does* do it well enough to fill 3 cores. That is not just use them, but utilize them 100% *if/when the load exists*.
Does that happen often? Maybe not often enough, but when it does happen a true quad core does have an edge to 2-core and 2+2 cores, the rest of the times they perform similarly.
Like I said it does not happen to most Android apps, not even to many, but it happens to enough, and it happens to the popular ones. So yeah there is a measurable gain by using 4 cores (instead of 2+2 cores) and even a slight gain when using 4+4 cores. Again, I'm talking performance here, not just power consumption.
I'm afraid I'm not frequenting forums, so I wouldn't know what is told there. I'm mostly referencing things I know, and things we've measured so that to create our testing frameworks, so I would say that my opinion on this matter is -at least- as well informed.
BTW newer tegras are 8 cores (4+4). I don't know when and if they'd revert into low core-count. Most of the rest are using high core count too (Chinese and Korean makers) and it's the reference design by ARM too.
Anonymous, 28 Mar 2016You only test use cases and individual components. That's not real testing, of course it has t... moreNo, we (I) do not only do "unit" tests on code parts. If your developers do not know where things tent to break then I would say they don't know their job very well.
And we do also have dedicated testers, they know testing and some code of course (writing alot of automated tests) but they do not know more than our developers.
I'm not sure you understand what i'm saying. Yes, of course background processes (by definition) can make use of extra cores. No doubt about it.
But as Anandtech says multiple times in that article there is noting one single big core could not handle by its own.
If you read Anandtech or the Beyond 3D forum for example (were alot of very knowledgeable people discuss different mobile SoCs, including ppl from Anandtech crew) Apples A8/A9 is regarded as having the best CPU design of all ARM CPUs currently.
And both benchmarks and "real" tests show that too. And for the record I too think Antutu is a crappy benchmark.
Most of the apps you mention that uses many threads are so light on load that it does not matter anyway.
But the heavy load apps, like games, the time difference is quite huge in some cases as you can see in the clips in my previous post.
Again Apples A9 is has the fastest CPU in majority of tests today so there is no question that have two big cores is a design that works very well.
But its also a much harder (and by that more expensive) path to take, designing a large core with high IPC than "just" clustering alot of smaller cores.
Ask yourself why Samsung even bothered designing their own big M1 core which also has alot better single threaded performance/IPC than the ARM cores used in previous Exynos (almost the same as Snapdragon 820).
If multi core is all that matters they could just as well has stuck with using referens ARM cores, why bother with big cores?
nVidia also went for the big core design with its own custom Denver CPU, just as Apple.
Anonymous, 28 Mar 2016What Anandtech says about you earlier example (page 4):
"The big cores also see a fair amo... moreYou only test use cases and individual components. That's not real testing, of course it has to be done, but the real and hard thing to be done is how a software reacts to its environment. And that's f***ing science and let me tell you, high quality testing is far harder than developing, because you actually have to know developing almost as well as developers, but also have to have a good understanding of where things tend to break. Like memory leaks, processes that override timers and a myriad other things that a developer don't have to think (why do you think testing pays so well?). So don't tell me that I don't know how software uses resources, I *have to*, so that to do my job.
Now if you read the review I gave you, in ALL app launches more than 3 *full* cores are utilized in each and every app launch. Granted it's burst utilization (it only happens for some milliseconds) but if you stuck enough of them (as you often do during a typical workload) it's already quite a difference than just having 2 powerful cores (like both Qualcomm and Apple do).
What's more, *all* Google Apps are highly parralelised, not just play store. The chrome test I gave you spent most of its time to 3+ *full* cores, sometimes it would fall as low as 2, but never less.
From internal tests we're doing we know that Android system (background processes) are well parallelized too. Same is true for the Facebook App and most of the top ten play store apps.
You seem to take it too lightly, but if 70% of the people *only* use those apps that's a pretty significant finding.
Yes a new app would mostly use one thread, yes games do that a lot too, *but* well established apps eventually get heavily parallelised. The idea that just building good single-thread performance is enough, is long dead. Actually Android (and its apps) parallelizes stuff better than any other OS I have seen in similar scenarios, by buying a 2+2 core CPU like Qualcomm's you actually misuse that very strength of Android.
I can't even start talking about Antutu. Those guys are completely off the deep end. Now that Android is more parallelized than ever, it is now that they decided to release a version that cares more about single-thread performance. They're a joke, and probably paid (I would not be surprised at all).
Let's hope Qualcomm and Apple's way of doing things (2 bulky cores) won't catch on, it's the way of the past...
Exynos actually shows in real time tests how far ahead is despite the "lower" single-core scores. That wouldn't be if your mantra was correct.
Exynos 8890 (GPU Mali-T880 MP12) = 265.2 GFlops
Snapdragon 820 (GPU Adreno 530) = 407.4~498.5 GFlops
What u think? =]]
Anonymous, 27 Mar 2016Tomato, tomatoh ... both Apple and Qualcomm processors I put on the same pot, all they care ab... moreWhat Anandtech says about you earlier example (page 4):
"The big cores also see a fair amount of load. Similarly to the S-Browser we have 1 very large thread that puts a consistent load on 1 CPU. But curiously enough we also see some significant activity on up to 2 other big cores. Again, in terms of burst loads we see up to 3 big CPUs being used concurrently."
Now look at how much each core is been used in this example, they are not running 100% (1 in the core scale). Again, not many cases where one single big core is not enough to handle the load.
To be clear, of course there are applications where you can make use of many cores, even for performance and not only from an efficiently point of view (like the "rare" example as Anandtech put it on the page 11 test).
But most of the time you do not and then the single threaded performance is what matters since EVERY application gain from having high single threaded performance, no matter how they are designed, multi threaded or not.
I don't care one bit about Qualcomm, im talking general application design.
I think its interesting that you say that you as a tester have done your research and know "how things are". Developers writing the actual code don't you say? :D
Trust me, besides from actually writing code I too do alot of testing of both my own and co-workers code ;)
If multi cores where all so important as you say Apples two cores in A9 would not stand a chance against a quad/octa-core SoC. But guess what, not only is it keeping up and out perform other SoC up in many benchmarks but it does outside of benchmark tests too:
Especially look how well it does in those game start up, were we can guess from Anandtech test not many cores are actually being used.
To be fair, the storage solution is quite a bit faster on iPhone as well so its not all CPU of course but again, if many cores where so important as you say the android phones
should be able to run circles around the iPhone with "only" two cores...but they do not.
DFranch, 25 Mar 2016SnapDragon processor supports all the radios used in North America. I think the Exynos doesn'... moreThe International S7s also don't have Band 4, which is used by AT&T and the Snapdragon has.
https://www.youtube.com/watch?v=vR2zq5I3i4g thats how the speed test should be done
Anonymous, 27 Mar 2016Haha Qualcomm fan boys are so fun. They probably haven't finished primary school. I always won... morewhen you want to say something to fans of some brand, when you start talking how they stupid for example of something like that then say "most of them",not all,becouse every samsung fan boy,not exynos,samsung galaxy fanboy i know,dosent know anything about phone except front camera,and that on the front of device he is using is writen samsung,and to people like that samsung is the best,but they dont know why,and they probably never heard about exynos,its probably to complicated to pronounce that word... im sort of qualcomm fan and i have reasone for that. AND the way this very smart youtuber compare the performace of this two chipset is not relavant,you simply cant cant click everything in the same time... and sd820 maybe is slower then exynos in the samsung s7,but that same chipset in the diferrent device with different software can be faster then s7 with exynos.
proper speed test is when you open the same app in the same time,and you see which phone opens first,same thing for ram..