Mediatek Helio P10 and X20 adoption is set to be sky-high for early 2016 but for phones coming later in the year they might receive the 16nm Helio P20.
The Helio P20 is set to come at some point this year, but with Helio X20 manufacture not kicking off until April (ish) don’t expect the P20 in a phone until at least summer.
Gizchina News of the week
So a long wait, however when the chipset does arrive it will be built on a 16nm process to ensure a faster more efficient SoC. Although its still very early days yet and the P20 specs haven’t been released we expect Mediatek to give it some sort of combination of A72 and A53 cores (perhaps 6 x A53 and 4 x A72?).
If Mediatek continue at this pace a Helio X30 should be on the cards for the last quarter of 2016.
It will be the same bullshit 3 cluster combo as on X20 (2x A72 & 2×4 A53) probably with lowered clocks compared to X20 & naturally cut down GPU.
They are starting even to cannibalize their own production with vast majority of similar products not even to mention how they won’t be able to maintain all of them.
3 clusters make possible to build a stable 20nm soc more powerfull than all other unstable 20nm soc (except Apple A8 ones witch are stable but lower even on singlecore)…
It make possible to have a very close to CPU singlecore power compared to 16 or even 14nm.
It make possible to have a better multicore capabilities than 16 or 14nm.
It allow not so bad gpu framerate in-game compared to throttling gpu (such as Adreno or mp8/mp12…see realjjj posts!)
It allows good price for manufacturers.
Bullshit is 28nm A72 rev0 on 650/652…throttling on s810’s 20nm A57 wasn’t enough?
Or even put underclocked Kryo as LITTLE core.
Not very higher (and not very stable, si lower on 25% of the bench), very lower multicore capabilities…with bigger die-size!
Hypothetics shorter pipelines to reduce consumption due to lack of real LITTLE core…so maybe/probably lower JavaScript and html5 capabilities (like on Krait)
Core usage and scheduling is good on mediatek’s.
See how they make work A53 with corepilot on 6752 compared to simple clusters migration on S615… Same CPU, same clocked RAM… Far far better real-life usage.
In my opinion, CPU on X20 in 20nm will be better and more stable than S820 in 14nm
Waiting for real-life tests, not not partial benchmark comparisons.
Kirin 950 seems to be good and stable too in 16nm…
Time to Qualcomm and Samsung to stop promising with benchmarks and give real-life what is on the paper!
As I dislike Apple, I must admit that they keep their promises : the power you pay (too expensive) is the power you get…you’re are not paying for 2.0ghz and have 960mhz instead!
Two A53 clusters is a stupid power consuming thing that will just drain battery faster!
Multicore capabilities depends on hardware & software interconnect!
The Mali MP8 or MP12 will consume less energy to deliver same performance like MP4 & I believe in this case on this SoC it will be in a best case a puny T820 MP4. Not even to mention Adrenos that are much more power efficient than Multicore Malis.
Go & sleep with him (realjjj).
Wake up & get real!
Compare NUU Z8*performance with Moto G (S410 @1.4GHz) & if you wish Huaweis.
http://media.bestofmicro.com/P/W/555620/original/honor-5x-chart-2.png
Then compare a power consumption between NUU & Moto G skip the Huawei this time to make it fair (because of battery size & cetera).
http://media.bestofmicro.com/P/V/555619/original/honor-5x-chart-1.png
A MTK hotplug aka Core Pilot is good but it’s still based on work queues & task migration. Don’t write about things you don’t have any idea of! The MT6752 is not on the same clock rates as S615 it is in rev 2 S615 & S617 by the way S617 works better. However their is a difference in the A53 rev 1 (Snapdragons) & rev 2 MTK but this whose major rev & the difference between A72 rev 0 & rev 1 is a small one. However MT6752 also uses more power then Snapdragon contrapart’s.
Shorter pipeline means faster response but also smaller possible top frequency.
The S652 will really give users a stellar real life experience.
Aple raised ARM to be where it is now & they are not raping silicon with to big frequencies so much as others do (excluding Intel). Honey golden frequency to stay on the grean side is about 1GHz. I hate Apple for other reasons & in my case it goes way back (20+) years ago.
I think low frequency A53 is a low thermal gap between A72 and other hot part of the soc. It make a cold zone to prévention heating
Pcmark one of the most incapable benchmark…
Results depends if the soc is compatible with the app.
I saw h265/4k capable vpu make shitty score on video part…
This one is even worst than antutu!
In your exemple mt6752 is capable of h265/1080p…s410 can’t do such a thing, hevc is limited to 720p!…pcmark said 410’s video decoder is better, how it be possible that a bench say such a stupide thing…and how can you show such an exemple…maybe because you don’t understand the weakness of some benchmarks!
Second is writing score…it resieze text, compare score between fullhd ans HD screen is non sense!
You can’t compare “experimental” rev0 cortex A53 un S410 with rev2 in mt6752…
615 is rev 1, not 2!
617 is rev 4…very close performance to rev 2 (geekbench 1.52ghz : 745 on 617, 1.49ghz : 730 on mt6732 quad core)
Even if you underclock mt6752 to the same frequency than S410 it will still providence 15 or 20% more power un single core test.
410 isn’t bad to use but it’s a technical failure due to rev 0 cortex A53…even 1.0ghz A53 rev3 on mt6735M sometime beats 1.2ghz rev0.
The more fun is that the new 412 still seem use these rev0! Qualcomm laughting at their customers!
Your remark about consumption is silly too…not the same frequency, not the same screen, not the same battery
652 will be 25% less CPU performance in web browsing than x20 due to lower frequency…quad A72 28nm 1.8ghz is too much.
800, 805, 810, how many throttling soc will they release before customers stop accepting…i thing 652 will ne one more…i’m not Alone!
In my case, problemes with apple Start when they stop using things such SCSI and powerPC and make x86, they lying to customers who think they buy something else than a PC.
Such as Qualcomm lying when they say s810 is 2.0ghz!
I have problèmes with Intel’s mobile i7U because they performance like desktop i3!
All that add confusion to customers is bad in my opinion
PC Mark is not perfect but it’s probably the best one currently out there. It’s not PC Mark’s fault if some manufacturers didn’t use reference coding (as for instance Allwiner is doing) so it won’t work with what it should. Even old QDSP6 on S410 – S615 is better solution than pore ASICS on MTK so PC Mark result is actually right!
LOOK AT THE PICTURES AGAIN!
It’s not all about hardware you know their is much about software optimizations too. I wrote correct A53 revisions use eyeglasses next time.
Again look at the pictures!
My remark about power consumption is on the place & with more kernel optimizations (thanks to Linaro, CAF & cetera) things get even worse for MTK.
As 1.8GHz whosent to much for S808 it certainly won’t be to much for S650 – 652. Their really isn’t much difference between 28nm HLP & 20 nm considering power consumption (look for comparation of 20 nm Exunoses).
Use CAF chromium on Snapdragon & 1.8GHz Snapdragon will become faster than 2.2 MTK.
I hated them for “Big brother is watching you” comercial (original Mac) & I hated them even more when they trow out Power RISC (Power PC) arhitekture, will probably hate them to the end of my days.
You must really assume whole SoC as a sum of its parts & not just as one part of it. Qualcomm is still far out of the reach when it comes to celular radios (which is by the way second biggest power consumer in real usage as it always eorks), New DSP blocks are far superior to implementations MTK is using & Adrenos consume less power per W so yes S652 will be much better SoC for consumers in many ways.
What did you expect from a Ultra low power CPU? Still it menages to achieve same performance as i3 with 2+ less power consumption. Their is no magic out there just a good or bad engineering.
Problem with ulp is naming, not performances…they’re quite good for mobile usage…I didn’t expect more than an other name.
When most of video player works with soc as Allwinner or Rockchip (VLC, Kodi, mxplayer…), Yes it’s a shame that PCmark isn’t able to use them properly. These players represent reality of mobile video reading, so that PCmark doesn’t represent reality.
You can’t throw benchmark on the table without understanding what it mean…and if the result doesn’t represent reality…the benchmark is bad, it’s simple.
An other exemple is using geekbench to compare RISC and CISC CPU …geekbench is unfair with x86…if you review an atom phone you shouldn’t talk about geekbench score.
Even between Arm cores some results are funny…
Take side by side a mt6752 (even in fullhd) and a s410…obviously, without any benchmark you will see.
You can even take 615, it reacts worst than 6752 by far…more than the frequency difference…same apps often open about 50% faster on 6752 when it’s only 15% bigger on clock, no need to benchmark!
Agree with cellular radio superiority of qualcomm.
The second good thing on 820 is UFS 2.0…
All other things aren’t so impressives outside benchmarks…as always with Qualcomm, good but not as good in reality than in benchmarks.
And I have doubt about stability as Qualcomm often have problem with it…nearly always they doesn’t use lowpower core.
I have doubt about their custom core solution…Krait worst than A15/A17 specialy in JavaScript/html5, Scorpion worst than A9… Their avantage : they were trully available before standard Arm.
Even vellamo (edited by…Qualcomm) show the avantage (real one, you can see it in your webbrowser) of cortex A17 on a Krait 450…(I burn my nand due to the worst disk access I ever seen on rk3288…even worst than on Nexus 7 2012! but wow even at 1.8ghz A17 are damn fast in webbrowsing compared to 2.5ghz krait!)
One of the biggest problem is…Google…always promote Qualcomm even when products are badly designed such as 810!
Keep Mediatek low end image with Android One project (well is Google really sustain this, they’re not often talk about it)…
And ignore intel. Are there economic links between Qualcomm and Google?
What a shame when we expect diversity from Android
Their is something that is called a “Symphony menager” on last QC SoC’s in reality it’s task dividing scheduler that picks between possible execution units to with one it will send a task to be executed picking a right core or accelerator to be used & it alone will save about 20% of the juice SoC is using. Most of players won’t work good with newer Allwiner SoC’s & their is a lot of fuss in Linux community about Allwiner not using a proper video for Linux implementation & not publishing the sources that it’s obligated to do by the licence. Roock chip is entirely another story & old Rk3288 works wonderfully concerning this (it’s probably best choice for TV box) & even Ubuntu works good enough with it. From ones with good suport I can only mention Amologics more. Geekbench is pore example & reason why it works badly with X86 (Cisc) is because it uses only essential (SIMD) optimizations. If you want real estimations you need to use open industrial grade tests and compile them from the source with all needed optimizations & best tool chain for the architecture available (compiler, libs & other tool’s). How much can performance differ I will show you whit example of this essential math lib:
http://www.yeppp.info/benchmarks.html
It’s small academic project but easy to implement.
Krait design whose cross mix of A7 & A15, whose really power efficient & rightly on pair with performance of A53 (but still more power efficient mainly thanks to its 32 bit design). Scorpions where bad & based on A8 design (altered), QC never managed to push it’s specific hardware design changes in the main stream (concerning software).
I don’t use shit like Velamno for Web benchmarking, I am still at the Peacekeeper even it’s development stopped, find it best & not optimized for particular architecture or Web engine also like it doesn’t glorify rarely used components over usually used ones in real world (like Web GL over traditional html tasks).
Google promotes what it have open stack to work with, it will never ever accept to work with closed property only one. Unfortunately now their isn’t much under open stuck this day’s (in smartphone world) only QC & Exunos as Texas Instruments & Broadcom left the market (mostly thanks to QC).
You see Google doesn’t have much choice at all especially if you think about it how rarely Samsung collaborates with other manufacturers to use their SoC’s & I don’t think that will change.
Hmmm I prefer Amlogic S812 box despite weaker A9 and lack of 10bits h265 than rk3288…
All benchmark where better on rk3288, but real-life usingis better on S812.
If when tweaking IO-Scheduler and cache, a simple mp3 on external hard drive stutter when system attempting to access emmc…I don’t know if problem comes from rk3288 or emmc…but I have a rk3188 and nand access aren’t good too on it
It comes from neither of them but poor optimized caching along with badly implemented block & Asynchronous IO’s on kernel side. Do try with deadline IO scheduler. It can also be cased by pore implementation of mutex locks, rq queue expires & cetera but deadline should resolve it anyway.
Edit: It’s also important who’s set top box you used, only couple of manufacturers did a good job in mantinance & optimizing it.
Yes I try deadline, and bigger cache, it improves a little, but problem was still there…now nand and box are fully dead…even try to shortcut the nand’s pins…it’s probably due to many plug/unplug
Manufacturers was Sunchip under Ugoos name…
My Tronsmart s812 is cool with 1To usb-harddrive on it (with FLAC and Blu-ray h265 ripped by me – originals only – great codec!), Kodi and Kore to pilot it from tablet
Did you herd a sad news Kodi is switching to default Android acceleration lib (under Android of course) so we can kiss goodbye manual contrast setting & other costume settings.
I’m still on Isengard 15.2. I didn’t try Jarvis beta and candidate release…
I heard that libstagefright support is suppressed in favor of Mediacodec…it’s sad that they give not the choice to userusers like previously releases.
But I’m on Amlogic, so I will use AmCodec…they still support it…
Mediacodec seem to have problem with 4k video display (green screen)
Well Jarvis is coming, they released a second release candidate last night.
Mediacodec is being polished out concerning that problem & with it skins can finally be rendered in different (lower) resolution than movie (witch whose a grater performance problem). Good news is that DSP pack will also be landing on Android so it’s not all bad. ?
You are seriously damaged.
You attack me here for no reason and you make idiotic claims suggesting i ever argued that MP4 is good, or enough or better than MP12.
You just talk without thinking. You are not incapable of thinking, you just refuse to count to 10 before opening your mouth and going off the rails.
So just go forth and multiply.
Em I?
Do you want me to post an url to comments of yours where you claimed how MP4 Mali is better than MP12 or fresh one where you still claim how MP4 is better than MP8?
I didn’t attacked you just told your friend to get litle closer to you if he wants.
At the end me & him had a civilized normal conversation that even ended friendly.
When I am not right I admit it & I am very capable of critical thinking with formal & un formal logic. I can multiple this with trigonometric progression but I don’t see any reason for that. Even an idiot knows difference between multiplication of: 1, 2, 3 with an possible exception of you.
Now go & wash out that dirty mouth of yours!
You know comments like this actually tells us all about your; good education, good behavior & how intelligent you really are.