Don't bother with battery comparisons on the i9500, the phone is unfinished.

Search This thread

AndreiLux

Senior Member
Jul 9, 2011
3,209
14,598
So I got my i9500 and already did some foolery with it.

Fine device, but I hate the raised lip around the screen edge. Something I definitely did not miss on the S3 and something very annoying.

Other than that small design critique:

THE ****ING PHONE ISN'T RUNNING FINAL FIRMWARE!

Basically the CPU is running on the cluster migration driver, meaning it switches all four cores from the LITTLE to the big cluster, as opposed to the core migration driver who does this in an individual core-pair manner.

You can pretty much throw all battery comparisons out of the window: it's completely unfinished and unoptimal.

I already compiled the kernel and flashed it without the cluster migration tidbit, but the phone won't boot. So yea. Current sources also useless.

Cleverly enough: you can't really distinguish between the two drivers apart from one manner: if /sys/devices/system/cpu/cpufreq/iks-cpufreq/max_eagle_count is present, you're running an IKS driver. If it's not, then you're running the sub-optimal IKCS driver.

So yea. We'll see what Samsung does about this, currently the advantages of big.LITTLE are pretty much unused.

Another nail in the coffin on how rushed and unprepared this phone has been.
 

Xdenwarrior

Senior Member
Mar 23, 2007
263
14
Wow, this is seriously turning out to be a fiasco.

This is EXACTLY why at the end I don't care for technical details about socs but was rather waiting for real world usage first. As much I wanted to agree with Andrei Lux on how intelligent BigLittle is, I sort of felt that it wont be same at the end.
Question is now: Is this possible to fix in the near future?? So that maybe buying the Exynos will be beneficial if the devs take over. I wont bet on Samsung introducing mind-blowing improvements in that department in upcoming firmwares :(
 
  • Like
Reactions: WarEagleUS and AMT2

AndreiLux

Senior Member
Jul 9, 2011
3,209
14,598
Question is now: Is this possible to fix in the near future?? So that maybe buying the Exynos will be beneficial if the devs take over. I wont bet on Samsung introducing mind-blowing improvements in that department in upcoming firmwares :(
The code other driver is there in the kernel, it's just not used. No idea. It's not like we need Samsung for it: I already talked to a developer at Linaro about some incomplete switcher code that's being currently getting the green-light to be made public. But who knows how long that will take.

Whatever the case, I gather that they can't just let it be in the current state.
 

Xdenwarrior

Senior Member
Mar 23, 2007
263
14
The code other driver is there in the kernel, it's just not used. No idea. It's not like we need Samsung for it: I already talked to a developer at Linaro about some incomplete switcher code that's being currently getting the green-light to be made public. But who knows how long that will take.

Whatever the case, I gather that they can't just let it be in the current state.

Any way to just disable cortex a15 altogether yet just to see how well cortex a7 will perform in simple texting, browsing, calling and to see what the battery life will be like on that?? (cause cortex a7 only uses like 200 something mw as opposed to 1000mw for snapdragon). I know u wont be able to game. How often does Cortex A15 hits in? cause I would suspect a much worse battery life with incomplete drivers doing the switching if its very often on. But PocketNow reports very similar battery results to snapdragon variant which I find odd :confused:
 

AndreiLux

Senior Member
Jul 9, 2011
3,209
14,598
Any way to just disable cortex a15 altogether yet just to see how well cortex a7 will perform in simple texting, browsing, calling and to see what the battery life will be like on that?? (cause cortex a7 only uses like 200 something mw as opposed to 1000mw for snapdragon). I know u wont be able to game. How often does Cortex A15 hits in? cause I would suspect a much worse battery life with incomplete drivers doing the switching if its very often on. But PocketNow reports very similar battery results to snapdragon variant which I find odd :confused:
Use any app to limit the CPU frequency to 600MHz. That'll limit it to the A7 cores running to 1200MHz. Basically you can just use CPU-Spy. Everything <= 600 are A7's mapped at half frequency, everything above it are A15's at 1:1 frequency.

As for PocketNow: irrelevant. The difference is what could be instead of what is, the Snapdragon doesn't play a role in the discussion here.
 

Xdenwarrior

Senior Member
Mar 23, 2007
263
14
Use any app to limit the CPU frequency to 600MHz. That'll limit it to the A7 cores running to 1200MHz. Basically you can just use CPU-Spy. Everything <= 600 are A7's mapped at half frequency, everything above it are A15's at 1:1 frequency.

As for PocketNow: irrelevant. The difference is what could be instead of what is, the Snapdragon doesn't play a role in the discussion here.

Hey thanks, but I don't have the S4 to test it with since i'm still debating on which to get. I live in Canada and so the only version here which I can get a lot cheaper on a contract is LTE snapdragon, but I wont mind getting the Exynos since it got potential. Besides 16GB internal isn't enough for me. So that's why asking if u seen any improvements in battery when only cortex a7 ran? If a7 doesn't do much in power consumption, then no point spending 800 bucks and loosing LTE altogether...
 

BoneXDA

Senior Member
Oct 9, 2012
1,739
1,234
Is it possible that we're having a simpler Exynos 5 system technically closer to Exynis 5 Quad (plus 4 A7 cores) than a real seamless Octa-core system? It was strange reading that "Octa-core manufacturing starts in Q2" (April-June) then see Octa-core versions hitting reviewers early April, that's way too low time frame. Maybe this is a 1st-gen 5410. In any case, performance and current-state battery life beats the Snapdragon version, even if only just.
 
  • Like
Reactions: mw86

karlesg

Senior Member
Mar 1, 2011
207
160
Basically the CPU is running on the cluster migration driver,

wtf? Well done Samsung... This is ridiculous...

Oh wow. Just got word (without further in-depth explanation) that this might actually be a hardware limitation. Coming from a reliable source.

No words...

WHAT THE F@CK??!!

Actually WTF is a massive understatement here....!!!


Please can you give more info about this matter whenever is possible? This is very serious...
Is it a specific hardware limitation? Something that Samsung specificly did in GS4 (I9500) ?

Because this can't be a generic exynos octa limitation. It makes no sense... Unless everything we've read from Samsung and ARM about exynos octa, are completely misleading...
 
Last edited:

msavic6

Senior Member
Jun 28, 2010
2,007
575
Vancouver
A hardware limitation..? They advertised the functionality and to then release a device without it, is just plain stupid. Hopefully it is a just a kernel issue and can be resolved quickly.

Sent from my Galaxy Nexus using xda premium
 

rakeshishere

Senior Member
Feb 15, 2011
2,633
1,016
Probably Samsung will implement it in their Note 3 device? It's a conspiracy so that people buy their next Note phone but this news is sad.


Sent from my GT-I9100 using Tapatalk 2
 

rkial

Senior Member
Nov 25, 2012
751
192
So its either all A15s or all A7s?
so would the 'octa' really be a better choice than the S600? That should be powerful enough.. and the S600 is pretty power efficient too
 

bala_gamer

Recognized Contributor
Mar 29, 2011
12,350
7,035
Madurai-Chennai
Xiaomi 13
Xiaomi 13 Pro
So its either all A15s or all A7s?
so would the 'octa' really be a better choice than the S600? That should be powerful enough.. and the S600 is pretty power efficient too

What I understood is its either the full cluster of a7 or a15 is used/ functional based on the load, dynamically turning on one or two cores of a15 to work along with a7 may not be possible it seems.

I may be wrong, waiting for an elaborate exp from andrei

Sent from my GT-I9500 using Tapatalk 2
 
Last edited:

-]Megacharge[-

Senior Member
Nov 16, 2010
409
119
Montreal
What I understood is its either the full cluster of a7 or a15 is used functional based on the load, dynamically turning on one or two cores of a15 to work along with a7 may not be possible it seems.

I may be wrong, waiting for an elaborate exp from andrei

Sent from my GT-I9500 using Tapatalk 2

I was always under the impression this was the intention of Samsung's particular implementation of it. I thought it was common knowledge that Samsung's version worked on a 4 or 4 (A15) or (A7) basis.

Maybe he was talking about the ability to change that.
 
Last edited:

Top Liked Posts

  • There are no posts matching your filters.
  • 92
    So I got my i9500 and already did some foolery with it.

    Fine device, but I hate the raised lip around the screen edge. Something I definitely did not miss on the S3 and something very annoying.

    Other than that small design critique:

    THE ****ING PHONE ISN'T RUNNING FINAL FIRMWARE!

    Basically the CPU is running on the cluster migration driver, meaning it switches all four cores from the LITTLE to the big cluster, as opposed to the core migration driver who does this in an individual core-pair manner.

    You can pretty much throw all battery comparisons out of the window: it's completely unfinished and unoptimal.

    I already compiled the kernel and flashed it without the cluster migration tidbit, but the phone won't boot. So yea. Current sources also useless.

    Cleverly enough: you can't really distinguish between the two drivers apart from one manner: if /sys/devices/system/cpu/cpufreq/iks-cpufreq/max_eagle_count is present, you're running an IKS driver. If it's not, then you're running the sub-optimal IKCS driver.

    So yea. We'll see what Samsung does about this, currently the advantages of big.LITTLE are pretty much unused.

    Another nail in the coffin on how rushed and unprepared this phone has been.
    25
    Managed to properly enable the IKS driver instead of the IKCS one. The system presents itself as if it would actually work and the CPUFreq layer actually shows two different concurrent frequency planes as one would theorize.

    ...

    But it's all shenanigans, if you actually read the core status registers, the A7's never get switched on, even if the frequency they inside the A7's region. Furthermore the A7's L2 cache is always powered on but never actually does anything.
    Regarding the CCI: it is powered off (Or at least unusable beyond its default state), when you try to enable or do something with it, you can't boot at all.

    I now have info that this is a problem in silicon that won't be solvable, meaning neither core migration nor HMP will be possible.
    24
    "Since the A7 and A15 are equally capable of executing the same ARM instruction set, any applications running on one core can just as easily be migrated to run on the other. In the example above there are a pair of A15s and a pair of A7s on a single SoC. In this particular configuration, the OS only believes there are two cores in the machine. ARM’s own power management firmware determines which core cluster to activate depending on performance states requested by the OS. If the OS wants a high performance state, ARM returns the A15 cores at a high p-state. If it wants a low performance state, the chip will put the A15s to sleep and schedule everything on the A7s. Cache coherency is guaranteed via the CCI-400 interconnect, so any data invalidated by one core cluster will be reflected in the other cluster’s cache. ARM claims it can switch between core clusters in this configuration in as quick as 20 microseconds.

    If everything works the way ARM has described it, a big.LITTLE configuration should be perfectly transparent to the OS (similar to what NVIDIA is promising with Kal-el). ARM did add that SoC vendors are free to expose all cores to the OS if they would like, although doing so would obviously require OS awareness of the different core types."

    Source: http://www.anandtech.com/show/4991/...alcore-more-power-efficient-highend-devices/2

    Please before throwing big words out there read what I said and then read the article. I bold all the things you clearly misinterpreted. As said before it cant use cores from both clusters. Only 1 cluster can be active at a time.
    It can use cores from both clusters at the same time. Please just stop talking and please stop posting 2 year old articles which you don't understand.

    There's cluster migration, there's core migration, and there's full MP. Read up on them before you embarrass yourself in saying you understand the mechanisms.
    Android is not Core aware and wont be for a considerable long time.
    You don't have the slightest clue what you're talking about here. Android has nothing to do with it.
    20
    I remember now why I stopped contributing to XDA. Everyone is a genius here that knows absolutely everything...

    @AndreiLux - You start this tread and 7 pages in all you give as proof is "Im not spoon-feeding you". So now that 20 ppl are involved fighting are you going to get your head out your ass and explain what you found in detail and how it affects everyone and how you got to finding this? or are you just going to say nothing with no proof giving except "um yea you know i found this um yea bro and you just um yea got to believes me bro" cause I would love to see some evidence to your speculation. Thanks bro
    Yea .. I don't see him sharing anything except ... "I found something, but I cant tell you"
    Excuse me but that the hell do you want? I already explained how to differentiate between the two drivers. I already told you I'm not going to spoon-feed you. Anybody capable of reading code will find the IKS CPUFreq drivers in the source code which is publicly available, so I'm going to ask you, what exactly do you want? Do you want me to explain to you how the driver works? Let's arrange some paid private lessons then.

    I am only as good as the information available out there. The information portrait in the articles I read is where my knowledge of this stops so PLEASE enlighten all of us that don't know how the technology works that's stated in 5 articles in total over the internet how YOUR understanding is so much better than the limited information available on the net about the big.LITTLE technology..
    I've already did my job in informing the people on this forum about big.LITTLE in this thread and gladly asked the questions posed there.

    But your first sentence here is hilarious, you're exactly the type of genius you're describing above: absolutely no first-hand knowledge and just regurgitating information from news websites. I take it for a fact that you didn't read my thread above on the matter else we wouldn't be having this discussion at this moment. So I'm going to correct you on that: you're only as good as the information which you consume. If you cannot invest and research more than 5 minutes on the top results of Google, then you have no place in the discussion.

    Also about news sites in general: Most of them are absolutely bloody clueless. Even AnandTech; Anand and Brian himself are no experts on the matter: they are journalists with both of them having some area of expertise. I'm actually the one sharing information with Brian every-now-and-then. And they can be wrong, as demonstrated by Anand reporting and insisting the frequencies switched around on the Octa, just because that's how he was told. Learn to differentiate between first-hand experience and factual reporting (Also known as "development") as opposed to plain journalism.

    Now please do yourself a favor and stop embarrassing yourself by disappearing and not replying back here again. Thank you.

    @AndreiLux there is a thread of a guy who have the Korean S4. there is no 1.8Ghz max frequency, he have 1.6Ghz like the international version. I think Samsung just changed its mind. If you want the link to the thread just let me know.

    Sent from my GT-I9300 using Tapatalk 2
    The 1.8GHz limit is something which I guess is coming from the per-core count limits which I explained once, meaning 1.8GHz is supposed to be hit only on a single A15 core with 100MHz decrements per additional core until 1.6GHz.

    As I've said, the only way to verify this is to check the sysfs entry as I stated in the OP. The report about "hardware limitation" came from a Linaro big.LITTLE developer who is on the IKS development team, his code is running on this phone. He didn't disclose any concrete information but that's what he answered me on the cluster vs core migration topic on the Octa.
    18
    Yes I completely understood what the OP said and I complete understand the technology but yet again my post was to clarify to the people that want to cancel their orders and freaking out how the technology works.

    But what you are asking for is not possible. Hardware infrastructure don't work that way regardless. You want to have 2 cars deliver 1 person to a location by having 2 wheels of each car turning where 1 car could have done the same job and have all 4 wheels turning consuming less power in the process

    Listen to what you are requesting as logically it makes no sense.

    Your arrogance and ignorance are amazing, you said "I completely understood what the OP said and I complete understand the technology" Amazing, you understand big.LITTLE completely, but totally fail to grasp the two main modes in which it can run in.

    This technology, can with future updates to the Linux kernel run all 8 / n# cores together on the same task, it is called heterogeneous multi-processing (HMP) Somehow with your 'enjin (sic) analogy this can't be possible?, on a threaded task, do all threads have the same priority and need of CPU cycles? Intel's hyper-threading technology does not offer equivalent CPU performance to each thread, but it works.

    The feature, we expected from Samsung in their 5410 was IKS with core level granularity, so you could have 2 A15s running, with an A7 also active, for example the A15s could be active while web browsing, whilst the A7 is checking for system / email updates etc. Now the OP is saying that the granularity is at the cluster level, that is a disappointment as it doesn't deliver the technology nuance we were expecting.

    At first, software was thought to be the reason, but now the OP believe that it could be a hardware limitation, I await patiently to see how this plays out, and I can understand people's disappointment, and desire to wait until all the fact are clear before spending a significant amount of money on a phone.