Comments Locked

65 Comments

Back to Article

  • Lodix - Wednesday, April 4, 2018 - link

    Nice work, and shame on Samsung.
  • skavi - Wednesday, April 4, 2018 - link

    How do huge companies fucking up this badly?
  • skavi - Wednesday, April 4, 2018 - link

    *fuck
  • Toss3 - Wednesday, April 4, 2018 - link

    Because they aren't run by engineers and have to work together with other departments that may have a better understanding of what will sell more devices. Pretty sure Anandtech is the only site that use PCMark as a benchmark.
  • Spunjji - Thursday, April 5, 2018 - link

    So having a device with decent battery life /doesn't/ sell the device? You're right about there being many factors selling a device, but the issue here isn't about one thing being sacrificed for another - it's about proper functionality being left on the table because of rushed and/or buggy code.
  • Speedfriend - Wednesday, April 4, 2018 - link

    And this is why I read Anandtech.

    Thanks for saving me from buying an E9810 S9+ too.
  • admnor - Wednesday, April 4, 2018 - link

    Same here. Currently waiting for the OnePlus 6 launch and review.
  • johnnycanadian - Thursday, April 5, 2018 - link

    Pixel 3 XL here, but yes and yes.
  • Quantumz0d - Thursday, April 5, 2018 - link

    S9+ is the best package you can get. The Op6 is a clone now, and for 700USD you are just paying for the OxygenOS and BL unlock and development. But the features and price, design are way off.

    I'm running an 820 chip. I like it, but if I had to pick one now I'd happily chose this 9810 S9+, Flar2 is developing I'm pretty sure the reason being its full package and no absurd notch.
  • lilmoe - Thursday, April 5, 2018 - link

    Yes, the S9 is the best package. I'd recommend it to everyone. Thing is, I wouldn't touch it myself. It's too early for this kind of chip on Android at this process node. Maybe with Android P, and maybe at 7nm EUV.
  • jjj - Wednesday, April 4, 2018 - link

    But you don't have a battery life result for S9 with SD845. Comparing it with the S+ is not ok and not highlighting that is also misleading as many won't notice.
    You can measure power for both like you did for SPEC but in browsing since , in this case, the goal is to look at the SoC.

    Will you compare the small cores too in part 2? Just to see if part of the difference comes from cores and SoC. Somewhat related to this , the graphs with power for 1 to 4 cores were nice, maybe they stage a comeback?
  • Andrei Frumusanu - Wednesday, April 4, 2018 - link

    > Comparing it with the S+ is not ok and not highlighting that is also misleading as many won't notice.

    I'm not comparing it to the S9+ anywhere here in this article. All the highlighted devices are the E9810 S9. And we don't have a S845 S9, simple as that.

    > Will you compare the small cores too in part 2?

    I'll try at some point but it will involve some tricks for the Snapdragon variant as it can't be rooted.

    > Somewhat related to this , the graphs with power for 1 to 4 cores were nice, maybe they stage a comeback?

    The single core power posted in the S9 review seems to just scale near linearly with cores as far I've seen.
  • jjj - Wednesday, April 4, 2018 - link

    Andrei, you do compare it when you say that you think you can't catch up. You don't claim to have the result but I haven't even noticed in the review that the SD845 result is for the + (it is highlighted there). Now I had to check 3 times and look for the S9 with SD845 to make sure I am not missing it, then went to the review to see if you list it there and forgot it here. Not saying it's intentional but it is confusing.

    The core count scaling is tested with the power virus and how does memory bandwidth scale with that?
    If it's not the small cores and uncore, that leads to the difference in efficiency in SPEC vs browsing, maybe it's the scheduler and DVFS or maybe the task, hopefully you figure it out.
  • jjj - Wednesday, April 4, 2018 - link

    One unrelated question, any idea how OLED power consumption varies between the S9 and S9+? Same number of pixels but different area and never thought about this before. Are the pixels same size so you end up with some power?
  • jjj - Wednesday, April 4, 2018 - link

    nm on the display question, luminance is expressed on a per area unit so power will scale with area at same luminance.
  • lopri - Wednesday, April 4, 2018 - link

    They use identical components sans display and battery. It isn't much of an ask to extrapolate it yourself when the reviewer clearly stated that he did not have the device in the foregoing article.

    As to multi-core power scaling, I do believe OEMs set hard limits on power per core usage, so it is more of a device test than a power or bandwidth scaling test.
  • eastcoast_pete - Wednesday, April 4, 2018 - link

    Great work Andrei - shame on you, Samsung! Andrei just fixed or at least improved a sizable number of your self-inflicted shortfalls, maybe you, Samsung, should bring him on board as a (paid!) consultant.
    Apart from that, this article confirms my view of Samsung: do the hardware, stay away from the software - please!
  • [ECHO] - Wednesday, April 4, 2018 - link

    Amazing work Andrei!
  • [ECHO] - Wednesday, April 4, 2018 - link

    Kinda makes you wonder if, with somewhat simple tweaking like you've done, this wouldn't make a fantastic Chromebook chip given a higher TDP and better thermals...
  • Spunjji - Thursday, April 5, 2018 - link

    If they don't use it as such I'll be bitterly disappointed. Looks to have the chops to rival those godawful number-mangled m7/5/3 chips from Intel!
  • zer0hour - Wednesday, April 4, 2018 - link

    Another amazing article on the E9810 S9. Keep up the good work guys, there's nothing else like this anywhere else.
  • Toss3 - Wednesday, April 4, 2018 - link

    Amazing work as always! Pretty sure the engineers over at Samsung knew all of this already; they just had to push the performance high enough to gain those high Geekbench and Antutu scores to make them appear to have closed the gap to Apple, and to sell more devices (most 'reviewers' on Youtube only run those two benchmarks and call it a day).
  • Quantumz0d - Thursday, April 5, 2018 - link

    True. Of all the stupid sites everyone say about those only. They forget how th ROM and Kernel is tuned. The HTC10 was severely limited and unleashed with the help of XDA. Same for 3Ts limits and unstable 821 Sultan fixed it.

    Custom Software is always da best. 9810 running EX is really superb, A11 is a failure too. You can read the article on the 845 QRD. And S9 Review you'd see how it fails to hold clocks against the 835 - 40% loss in clocks (A11) vs 10%(835). Add the Power hungry voltage ripping cheat and battery disaster its all shiny and fallacy.
  • ZolaIII - Wednesday, April 4, 2018 - link

    If you want help or at least an advice on tuning the complete mechanism I will be more than happy to help. It's not a straightforward as lithography is new & it depends what type of scheduling is used alongside with hotplug that is used & if it's configurable but I am certain we can do it in couple of days much better than any OEM ever did (including Sony) as I have experience doing so & managed achieving so in the past. So if interested please contact me.
  • Andrei Frumusanu - Wednesday, April 4, 2018 - link

    You're welcome to reach out to me via email. I'm relatively confident in what I'm doing as I had been messing around with device kernels, power management, and schedulers for well over five years now.
  • ZolaIII - Wednesday, April 4, 2018 - link

    I contacted puch so that they can give you my email as I have hard time finding yours.
    I developed unique approach regarding tuning which differs from traditional ones as load balancing and no task packing, hotpluging & cetera (CAF) or tight task packing, minimum active core's, hotpluging, more conservative CPU scheduler scaling (Sony).
    You can check it out along with user experience & achieved results for the Snapdragon S650 - 652 SoC's over at XDA.
    https://forum.xda-developers.com/mi-max/how-to/pow...
    As a hint it beats the much more power efficient SoC pared with bigger battery & same display on OEM's implementation which to start with whose my primary goal.
    Best regards and I hope we will soon start working on it together.
  • tsk2k - Wednesday, April 4, 2018 - link

    andrei(at)anandtech.com
  • ZolaIII - Wednesday, April 4, 2018 - link

    Thanks!
  • serendip - Saturday, April 7, 2018 - link

    I recommend Zola's work too, I've been using it on my Mi Max for a while now. I did some monkeying around with scheduler settings and kernel tuning in that device but I always go back to Zola's scripts for long battery life with almost-stock performance.
  • K_Space - Wednesday, April 11, 2018 - link

    Any chance we'll see the custom kernel in xda? :-D
  • ZolaIII - Wednesday, April 11, 2018 - link

    For whose work? Mine is there for a long time now as a script only for a adjusted kernel & build tree (so that tunables are exposed to the user space in this manner). We never putted them as defaults nor builded the franken kernel for it it's still based on what used to be standard included sched kernel infrastructure and with default MSM in kernel hotplug solution (which is now separate kernel ko module). Kernel, build tree, device tree, vendor tree and script are all present on XDA and or Github. I don't believe Andrei is interested in pushing it so far and I am certain he will detailed explain what & how he did in the future article after he is finished.
  • flar2 - Wednesday, April 4, 2018 - link

    I've taken a different approach with ElementalX kernel. Instead of disabling the hogplug mechanism, I boosted the hotplug limits a bit. With two cores, the limit is increased from 2314 to 2496MHz and with three or more it is increased from 1794 to 1924MHz. This provides a substantial increase in multicore performance while maintaining single core performance. Overall, the subjective experience is a bit snappier and it doesn't destroy battery life.

    I experimented with disabling hotplugging, but the phone quickly becomes unstable unless the big cluster max frequency is reduced. It looks like when all four cores run at freqs over 2.0GHz, they draw too much power. Reducing the max frequency obviously reduces single core performance substantially.

    Although Samsung's hotplug code is a bit convoluted, they are trying to achieve a compromise, balancing battery life, performance and stability. Ultimately, this is a hardware problem, the Exynos chip is too power hungry at the highest frequencies. The hotplug mechanism allows it to run at higher frequencies for single tasks. The alternative would have been to release the phone with a max frequency of somewhere between 1794 and 2314 MHz. Given that Samsung needs to be conservative due to chip variations, they would have struggled to go much higher than 2GHz (imagine the scandal if the hardware was unstable!)

    No doubt more gains can be made, so I will continue experimenting as well.
  • MrCommunistGen - Wednesday, April 4, 2018 - link

    Thanks for your continued work on the ElementalX kernel. I haven't used it since my Nexus 5, but it was really helpful back then.

    On another note, I think you've corroborated my knee-jerk suspicion that the reason for Andrei's device crashing when running 4x M3 @ 2.7GHz was a power delivery issue -- but I don't want to be putting words in your mouth or drawing conclusions you didn't intend.

    Hypothetically (assuming power is the issue), I wonder if you could run the M3 cluster at a low enough voltage that it would be stable at 4x2.7GHz. Maybe you'd need to win the silicon lottery in a big way to do it. I'm just academically curious whether it is possible with this combination of chip design/lithography/power delivery/etc.
  • flar2 - Wednesday, April 4, 2018 - link

    I'm quite sure it's a power delivery issue. They really had to push the voltages to reach freqs over 2.3. Heat could be a factor too, but I've had many phones that get a lot hotter than this.

    My device can't even complete a benchmark without rebooting at 4X2.4GHz.
  • Quantumz0d - Thursday, April 5, 2018 - link

    This is what I thought, to supply more voltage and current to the processor the SoC hits its ceiling (voltage for heat, current for the HW limit), the glass make of the device I guess can be attributed to this. Also the conservative approach on the GPU arch. Limiting its power vs the power hungry CPU arch.

    But when efficiency drops down we have to either make sacrifice for the battery or the heat. In these small confined spaces I guess we are limited as well, but given your EX reputation I guess we can have a better final product by the customization. Unlike the Apple throttle battery gate and disastrous A11.

    Also from what I remember unlike the SD800 days the newer 820 and up the voltage planes are complicated and hard for end user to tune, I used to UV/LiveOC on my Hummingbird in i9000 and OMAP in Gnex, seems like we cannot tune them much nowadays but still we can do it on these complex chips is great !

    Much thanks Flar2. For the insight and your work. Always a admirer of great devs at XDA..If Iwanted to buy a successor to OP3, I will either see the 9810 S9+ or the 9810 Note9 if you develop for the latter.
  • ZolaIII - Wednesday, April 4, 2018 - link

    What's the maximum sustainable frequency for two big cores running with high utilisation (on a long run)? That's also the answer how you would limit it so that it doesn't reach the point when it throttles down & reaches higher thermal leaking. This is actually a upper performance limit. The sustainable optimal one (or little higher then that) is the top one Samsung used on the little core's, really ideal should be about 1.3GHz so you could adjust those together by per frequency putin a load limit to interactive CPU governor. Regarding the hotpluging I still found two + two big - little core's active all the time as best possible solution, it's not a most power efficient one (which would be one big & all small ones) but is most suitable for all possible tasks. This is just add hock.
  • MrCommunistGen - Wednesday, April 4, 2018 - link

    WOW. Thanks Andrei! I'm looking forward to the rest of your updates! Hoping you can find a good happy median with performance and power with your DVFS and scheduler tunings (and running the M3 cluster at peak clocks higher than 1794MHz?)

    I think this really illustrates how hardware specs aren't everything. I know I'm not alone among enthusiasts in being disappointed that Qualcomm hasn't been able to compete with Apple SoCs in terms of IPC or single-threaded workloads. Nothing I do on my phone *needs* the performance, but I WANT IT ANYWAY! Similarly, I'm disappointed that their driver support for older SoCs doesn't last for very long. BUT at least they seem to get the hardware/software tuning *moderately* right on day-1 -- which is no small feat for a company as large as Qualcomm. I can see why they are so appealing for OEMs as a turnkey solution.

    I guess my point is: We can complain that Qualcomm is holding back the Android ecosystem, but think of where we'd be without them.

    On another note:
    The last time I messed around with the likes of schedulers and DVFS was before big.LITTLE on my Nexus 5. My unit would thermally throttle pretty badly under any sustained load (long sessions of Temple Run) so I lowered voltages across the board as much as I could while maintaining stability and lowered peak clockspeeds. The gains were nothing drastic like you've demonstrated here, but I was pretty proud of myself at the time.
  • A5 - Wednesday, April 4, 2018 - link

    Great article, looking forward to the other parts, even if I'm not in the market for this phone.
  • Lorem Ipsum - Wednesday, April 4, 2018 - link

    Great write up! Will you guys be willing to release your modified kernel? As one of those enthusiast people into modifying my mobile device I’d love to see some of these improvements.
  • flar2 - Wednesday, April 4, 2018 - link

    ElementalX has been available for a couple weeks already, with the changes I describe above
    https://forum.xda-developers.com/galaxy-s9/samsung...
  • x4nder - Wednesday, April 4, 2018 - link

    hi, that was a good read. What I don't understand though is why you said it's unlikely that significant changes introduced through firmware updates. Any reasons why Samsung won't try to fix it if it's a software problem?
  • Andrei Frumusanu - Thursday, April 5, 2018 - link

    In past Galaxy devices I've never seen Samsung change anything in regards to the SoC behaviour and they stuck with the BSP settings for the lifetime of the device.
  • serendip - Saturday, April 7, 2018 - link

    I'm surprised they won't change BSP settings because they're also the SoC vendor. Or maybe the chipset division people don't talk to the mobile devices people.
  • darkich - Thursday, April 5, 2018 - link

    Actually it just hit me what this whole mess up could be about..it's just Samsung deliberately crippling their chip to keep it on par with the SD845..and going too far with it/doing a bad job.
  • Quantumz0d - Thursday, April 5, 2018 - link

    Did you read the review, I think its not just limiting factor of the 9810 to match the 845. The way its designed is bad. Its a power hungry chip, they only made waves with the new GB scores, if you look past that bench and see the size of the die and the power hungry nature like A11 it falls of a cliff. The wider bandwidth is surely welcoming but at the cost of the thermal efficiency when the phone is made of glass and has limited area to offload the heat it will be a sour grape.

    The GPU also has been given low importance and not being upgraded due to the CPUs power needs. Considering the small area of improvising and the tuning. Samsung deliberately left the behind the screens part.

    Hotplugging is worst when it comes to the >SD810 SoC flagship chips. 820 will choke if you do that. Instead they need to tune it with efficiency.

    Still I like that Flar2 is developing for this device. Given the fact that the SD is locked and OP6 is notched. Exynos S9+ is the best package. Still gotta see how HTC does (usually conservative) and Pixel 3 is long time wait..
  • ZolaIII - Thursday, April 5, 2018 - link

    Don't worry S845 is also horrible design, well S810 whose worse & S82x whosent good one either.

    You are silly. Their is not enough core's that you can turn some down on S82x as a matter of fact best thing you could do is transform it into not heterogeneous one, limit the so cold high performance ones to the frequency of so called efficient ones then you could actually hotplug cores 03 & 04 but I don't think any of that would work as I think it's physical two detached cluster design. S82x is the probably worst design ever from engineering standpoint.
    Size of the core/instructions true output at given frequency = efficiency. If M3 is 33% bigger than A75 while having 30% higher instruction true output it will be equally fast as A75 running at 2.6GHz being on 2GHz while becoming more power efficient as leaking is very bad (better say insane) when you cross 2GHz barrier meaning at 2.6GHz power consumption is double. We can only build wider core's now if we want more performance as silicone limit is reached long time ago.
  • Quantumz0d - Friday, April 6, 2018 - link

    What you wrote is it anywhere near to what I said and you speak like as you are a HW expert but fail to make simple communication.

    Read sultan thread on hotplugging. That's all I can say and 82x is worst ? Better go and join Qcomm and maybe you can design the world's most powerful & efficient soc.
  • ZolaIII - Friday, April 6, 2018 - link

    Now you probably won't ever understand it but hire it goes anyway.
    The scheduling priority goes this way; HMP sched has priority over CPU sched & it places tasks on ready & available cores in the manner of first suitable core idling or if their is no core idling on the least utilised one if all core's are high utilised & they are over implied task packing rule (tunable) it will wait until their is an available core that meats the criteria. He doesn't play poker nor looks in the crystal ball has no knowledge nor cares about cores that are shot down. CPU sched tries to manage CPU frequencies the way it's written or to the applied additional rules regarding utilisation/frequency, there ware hybrid hotplug cpu sched solutions but they never played very well as CPU sched simply doesn't control CPU task placing. Hotplugging is slow not because a software but because every core must do pos when coming up & for both the core & memory sub system belonging to it & only then HPM sched can issue a placement or migration to it this takes up to around 200ms at least (that's in most cases 10 or more ticks from schedulers) compared to the idling core in which case this is down to: 20ms direct placing, 50~70 side migration on the same cluster, 75~160 up and down migration between clusters.
    But hire is the real catch if you have one big or litle core is idling & migrated/started task is SMP 2 (as most of them are) that core will wait until the other one is ready spending power & not doing anything. So that is the answer to your bla bla S82x worst ever... Hotplugging as hotplugging simply doesn't have any sense on 2+2 only SoC which by the way it's the worst one ever as analogy to used core's & typology is trying to kill a fly with a hammer & as their is more fly than things get worse. So it's very inefficient. At the end is the usage of the hotplugging useful and recommended at all? Yes it is on the octa (or hexa core more limited) but it needs to be good configured so that it doesn't cause problems & stalls to the HPM task scheduling because big cores are very power hungry and for some workloads like gaming it's even possible to save a little juice on small core's and what ever additional headroom you provide to GPU it's more than welcome. Bottom line is you need at least two of each core's in active state all the time, if you are a lite user four small ones & one big one on will give better results (reading, scrolling) regarding power consumption but as their is no such a thing as light use as Browser, social network apps, chats are all notorious by their high performance requests burst tasks final results are neglect able with marginal difference in power consumption between the two only 2+2 will be faster to respond (snappier). With good configurable hotplug on up to A55/A75 with cores that use shared L2 cache per pair of them (not applicable to A53's before r02) it's possible to score an additional 5% performance boost with no efficiency penalty using 2+2 setup & on octa big.little SoC's by simply choosing every first core of the pair to be the one always on (all L2 used & with exclusive access all the time + faster pos of other core as their is one less init lv to be done).
  • Quantumz0d - Friday, April 6, 2018 - link

    I understand the first post was exhausted yesterday, so it flew above. About the second post, I appreciate your time for that. I can understand what you mean regarding the Octa with the big.Little vs Quad 82x it seems like similar to the Ryzen CCX, HW limitation.
  • ZolaIII - Friday, April 6, 2018 - link

    Well not exactly unfortunately S82x is heterogeneous two cluster system same as octa big.little one's like S835 would have been better that it whose an ordinary single cluster old fashion SoC like S800 as then you could use hotplug & so on. I tried to explain that in the first post. It's beyond me and common sense why QC used big cores for the power efficient cluster on S82x. Also part about core size/ instruction output also in the first post ties only for OoO vs OoO core's as even best OoO design ever the A73 uses 3x power while providing 1.8x integer performance per MHz compared to in order A53's. On FP tasks it's entirely better to use in order A53's (only 28% performance difference while 3x power difference remains the same) but we can't control task placement like that (at least not yet).
  • Anirudhsarma - Thursday, April 5, 2018 - link

    Will lowering clock speed in settings give better performance and battery life on stock
  • zodiacfml - Thursday, April 5, 2018 - link

    Highly doubtful Samsung would ship a product that benchmarks less than previous SoCs and with such a large chip. Previous Exynos 8895 in the S8+ beats the SD835 consistently. The results here shows lower scores than the SD835.

    Why not get another copy of the device as this thing looks defective.
  • Quantumz0d - Thursday, April 5, 2018 - link

    Thank you Andrei again for this wonderful followup as you promised. Feels really great to see AT come back at full punch. Now with custom kernel, just perfect what I wanted to see !! Please continue with this approach regardless of the SoC claims or whatever.

    Every word is perfect as usual. Top quality stuff. Looking forward for more, big shout out to Flar2 about his approach and the info on the 9810.

    Fun to read and customization is what we should do always, utilize our fundamental thoughts, liberty and choice unlike Apple.

    Thank you.
  • thesloth - Thursday, April 5, 2018 - link

    Thanks for this, very useful and I'll be interested to see how far you can push this. I really hope they address this for the S10.
  • AlyIbrahim - Thursday, April 5, 2018 - link

    Will Samsung fix this problem and the exynos 9810 become better than the SD845?
  • AlyIbrahim - Thursday, April 5, 2018 - link

    Will Samsung fix this problem and will the s9+ exynos 9810 outperform SD845?
  • Monty1401 - Saturday, April 7, 2018 - link

    As an interesting side note, I was tracking the cpu speeds throughout my daily usage - the cores very rarely seemed to spin up past 1.79mhz (only hit speeds above this for around an hour total in the past 8 days) with relatively heavy usage (PUBG, slow-mo camera etc.). I imagine this might explain real-life battery is not as bad as the tests would suggest, but also why performance is not hampered hugely by limiting all cores to this speed? Further adds to my dilemma as to whether I return the device!
  • santz - Wednesday, April 11, 2018 - link

    superb article. thank you. i just purchased a note 8 only 2 days ago after finding out about the s9 exynos issues. oh well..
  • moinahmedqadri - Friday, April 13, 2018 - link

    This is a huge and limited offer to giveaway the iPhone X, iPhone 8, iPhone 7 Plus, iPhone 7 Special Edition [RED] and Samsung Note 8. Just signup to get your favorite phone.
    <a href="http://www.storenshop.com" rel="nofollow">Store N Shop</a>
  • moinahmedqadri - Friday, April 13, 2018 - link

    This is a huge and limited offer to giveaway the iPhone X, iPhone 8, iPhone 7 Plus, iPhone 7 Special Edition [RED] and Samsung Note 8. Just signup to get your favorite phone.
    www.storenshop.com
  • show41 - Friday, June 8, 2018 - link

    Solid article. I think there may be a mistake in the Fig of Performance efficiency, where the left bar was plotted as Joules/SPECspeed not SPECspped/Joules. If plotted in SPECspped/Joules, it should be "more is better".
  • JohnUitm - Monday, June 18, 2018 - link

    Have you considered release your mods to the public. I would gladly pay for it
  • hanhan - Wednesday, November 14, 2018 - link

    Please unleash the gpu power, what if you overclocking the mali g72 to 1.100Mhz per core?
  • Piereligio - Monday, March 18, 2019 - link

    Hi, I rooted recently my S9. I was wondering what kernel you used and what app (if you used any app) to edit its settings. Thank you very much in advance
  • supersohn - Thursday, September 10, 2020 - link

    how to remove hotplugging mechanism from exynos 9810 dvfs/scheduler?

Log in

Don't have an account? Sign up now