Qualcomm sends out invites for the Snapdragon 660 unveiling
- ?
- Anonymous
- j5$
- 05 May 2017
AnonD-646953, 05 May 2017It is well known that kryo280 is based on A73 and kryo is b... moreKryo 280 seems based upon a73, but the prior kryo seems to be a new design
- ?
- Anonymous
- j5$
- 05 May 2017
ZolaIII, 04 May 2017Not only that they mostly use floating point instructions b... moreI'm making two points: 1. simd (not fp) is fine 2. the jpeg specification calls for only integers, starting with the quantization step on the encode side. certainly there are non-normative fp implementations because using fp allows for greatr precision and speedups
- D
- AnonD-646953
- 0V{
- 05 May 2017
ZolaIII, 04 May 2017& you know that by pulling one from your ass then shrin... moreIt is well known that kryo280 is based on A73 and kryo is based on krait , as I told you before.
Anyway, look at this image and tell me what makes you think "A57 and kryo is similar".
http://images.anandtech.com/graphs/graph11088/kirin_960-performance-gb4_fp.png
- Z
- ZolaIII
- 3iH
- 04 May 2017
AnonD-646953, 04 May 2017I don't forget you were insisting that kryo280 is based on ... more& you know that by pulling one from your ass then shrinking your self to micron level & taking a tour around. So please enlighten us how you do that. It's evident that Kryo is based on A57 & we still don't know much about Kryo 280 design.
I thought this chat have some rules regarding what is prohibited.
- D
- AnonD-646953
- 0V{
- 04 May 2017
ZolaIII, 03 May 2017Well to set it all up into the right light; integer perform... moreI don't forget you were insisting that kryo280 is based on A72 and kryo is based on A57, and when I told you that you were wrong, you began writing technical detail with your f*cking English like a bulldozer in order to hide your mistake.
- D
- AnonD-646953
- 0V{
- 04 May 2017
AnonD-666662, 03 May 2017Anandtech is a very reliable source, so I'd trust that over... moreI have read the analysis of anandtech, and it helped me a lot to make my app. And you can find in my site, I refer to anandtech's post.
If you have one of the devices which has one of those cores: [A53,A57,A72,A73,kryo] , try my app and compare the score with the graph in my site, you will see the accuracy of my app.
My app has nothing to do with the optimisation, because the same runtime environment runs the same code. When we build an Android app we produce the dalvik-bite-code and dalvik runs dalvik-bite-code. If you are in the same Android version, you have the same dalvik. You can see in my test, AndroidN has a big difference with others. But others have similar scores.
I'm planning to make a native ver. The native code will have nothing to do with the optimisation too, if you don't compile it with optimisation options. If you compile without these options, the same ISA produces/runs the same code.
For your example of GalaxyS7 and OnePlus3T, the answer is "because OnePlus3T was overclocking when it detects the geekbench".
source: https://www.xda-developers.com/benchmark-cheating-strikes-back-how-oneplus-and-others-got-caught-red-handed-and-what-theyve-done-about-it/
(This is one of the reason I wanted to create my own app to check)
In the end, Qualcomm switched to A73 because kryo was too bad in power consumption, not because of its problem of UX. Battery life is more and more talked today. Kryo's DNA existed since more than 5 years ago. Krait had the same feature ; beast FP performance. But it didn't make qualcomm switch the micro-architecture.
- z
- zodiacfml
- J7R
- 04 May 2017
Nice but most manufacturers make the mistake or bad intention to use the SoC in a phone while pricing it at flagship levels. They may have a point that performance is the same when there's little impact on battery life but still they are milking money from people without knowledge of these stuff.
- Z
- ZolaIII
- S9u
- 04 May 2017
Anonymous, 04 May 2017Jpeg and png have int based datastreams, not fp.Not only that they mostly use floating point instructions but they also use SIMD instructions and for a very long time now (development started 1998). Today dominant in use jpeg lib is jpeg-turbo. It's pretty same story regarding PNG and most of the time NEON instructions aren't beneficial at all. On Android as we are mostly stuck still using soft abi it basically disables GGC to produce any floating point code. So your argument stands there.
- D
- AnonD-564191
- vG7
- 04 May 2017
be super ultra happy cuz i'm using sd400
- ?
- Anonymous
- j5$
- 04 May 2017
ZolaIII, 03 May 2017Well to set it all up into the right light; integer perform... moreJpeg and png have int based datastreams, not fp.
- M
- Mr. Electronics
- 0fL
- 04 May 2017
Mr. Electronics, 04 May 2017u kidding? I wish I had a phone like yours! Even though my ... moreWhoops! What I meant to say was that your a lucky guy or woman, not a guy and woman :) sorry for the mistake
- M
- Mr. Electronics
- 0fL
- 04 May 2017
Anonymous, 03 May 2017MY SD625 Feels outdated :/u kidding? I wish I had a phone like yours! Even though my isn't that bad at a snapdragon 617, it still doesn't provide as much battery performance as yours would. You're a lucky guy/woman
- ?
- Anonymous
- vx6
- 04 May 2017
AceShot, 03 May 2017I made some test. Most graphical intense games are using up... moreGood for you, but Today, Most people do Play Online Hardcore Graphics Game while on the background There are Chat App, Downloading Torrents, Youtube Videos for WalkThrough, Connected to Charger for Zombie Mode ^_^ and Bluetooth for Audio Satisfaction and ALL APPS THAT WAS NOT PROPERLY CLOSED .. (lazy)
And Some are not satisfied to Download only one hardcore Game on the Phone
I Think that kind of 1.3 GHZ Phone have no build-in Apps that Optimized Performance ...
So that being said , for 6 months of use of that king of Phone, Phone could Stubborn,, u Know ..
- ?
- Anonymous
- HjA
- 04 May 2017
I don't think it will be a 14nm version of 4xa73+ 4xa53, how will qualcomm sell 835 if this benchmarks the same as that?
- ?
- Anonymous
- tPG
- 04 May 2017
Sometimes I think 65X series is faster than 82X series
- ?
- Anonymous
- m}3
- 04 May 2017
Anonymous, 03 May 2017Be happy bro I'm using sd 430 😀Be happy bro, I'm using Snapdragon 200.
- Z
- ZolaIII
- S9u
- 04 May 2017
AnonD-666662, 03 May 2017Anandtech is a very reliable source, so I'd trust that over... moreIf you want to see relative performance metrics you use Open Source benchmarks that are or whose industrial standard like good old dhrystone and whetstone, Linpack or newer ones like core mark from EEMBC (all do as non member you won't have source access) you don't use popular bullshits.
My recommendation goes to the Roy Longbottom collection regarding traditional ones and on Android.
http://www.roylongbottom.org.uk/android%20benchmarks.htm
Geekbench isn't particularly optimised for anything it's actually rather not optimised for anything which particularly best shows on its X86 end. Anandteach also knows to miss sometimes & certainly aren't anymore on the level they once whose regarding in depth analysis of SoC's. Benchmarks are that that they are & their only to show relative performance & even that much is only possible with a good ones.
- ?
- Anonymous
- thg
- 03 May 2017
Peter-B, 03 May 2017Ah, sorry! The link was correct but I typed "Mix" instead o... moreNone taken, Peter. None taken. Homophonic naming scheme of Xiaomi models just took a heffty efect
- D
- AnonD-666662
- pTk
- 03 May 2017
AnonD-646953, 03 May 2017Yes, you're right in most terms.
Kryo280 is actually an A7... moreAnandtech is a very reliable source, so I'd trust that over anything else:
http://www.anandtech.com/show/11201/qualcomm-snapdragon-835-performance-preview/2
Now they haven't done a deep dive yet but the results speak for themselves.
Most benchmarks (done by both Anandtech and others) I've seen have shown the 820 being consistently outclassed by A72 and A73 cores in typical workloads except in FP.
And no, I don't have either an SD810, SD820 or SD835 device, so I can't really test anything myself. And the site you link to is in Japanese and Google Translate isn't helping.
Any benchmark can be made to showcase a particular processor doing well depending on how you develop it and optimize it. That comes down to both SoC, device and kernel as well as apps. That's why you see differing results with the same chip for example (eg OnePlus 3T vs Samsung Galaxy S7(820)).
For example Apple consistently showcase incredible JavaScript performance. Another example is how Geekbench had to revise how they scored devices because they favored Qualcomm chips to a degree that wasn't corresponding to actual performance.
In the end, it's all about user experience and it seems that A73 better serves that purpose than Kryo does otherwise it wouldn't make sense to switch architecture especially since they could as well have spend R&D resources on improving Kryo but for some reason it apparently wasn't feasible - at least not with their timetable.
- ?
- Anonymous
- JBH
- 03 May 2017
Anonymous, 03 May 2017MY SD625 Feels outdated :/Be happy bro I'm using sd 430 😀