Snapdragon 835

[Blacklizted]

SCREEN
Size: 6" - 7"
Resolution: 2K for longer screen time.
Bezels: Uniform size. Remove logo.

SPEAKERS/MICS
Speakers: Front-firing and doubles as mic depending on orientation

OPERATING SYSTEM
OS: Full Windows 10 or phone UI with landscape mode. 64bit. Demo of W10 on ARM is so cool. Can't wait to run UWP, Desktop Apps, and Android emulators on my phone

BATTERY
4500mAh or bigger. Removable.

MEMORY
4Gb or bigger. Needed if you're planning to run emulators.

STORAGE
64Gb or bigger. Expandable.

CAMERA:
Aperture: 1.4/f
Megapixels: 32mp (max)
Dual: No thanks. I have an iP7+ and its not that great
Flash: Xenon or multi LED. Not into flash photography that much
JPG: Improved or same algorithm as on the 950 XL

PROTECTION
Doesn't have to be waterproof but include some weather-sealing tech.

These and all the goodies coming from this chip like better bluetooth, wifi, video compression and playback, directx support, better image stabilization, lower battery consumption, faster processing, etc.

 

[kwright62]

I really hope there is a 7" screen coming with this processor. This would move Windows out of the phone business and into a completely new category.

 

[Vittorio Vaselli]

You will never run android emulators on a phone. it is slow on surface 3, on emulated x86 will be worst.

 

[startac7868]

Any possibility of the upcoming 10 nm Canonlake CPU from Intel in late 2017?

Yeah, I know MS has ported Windows to ARM and all this talk about the next Redstone Windows update but an Intel processor which doesn't need any kind of emulator to run full Windows (32 and 64 bit versions) yet almost as power efficient would be better, no?

 

[Vittorio Vaselli]

Any possibility of the upcoming 10 nm Canonlake CPU from Intel in late 2017?

Yeah, I know MS has ported Windows to ARM and all this talk about the next Redstone Windows update but an Intel processor which doesn't need any kind of emulator to run full Windows (32 and 64 bit versions) yet almost as power efficient would be better, no?

no

 

[Cruncher04]

but an Intel processor which doesn't need any kind of emulator to run full Windows

ARM does not need an emulator to run full Windows either. Emulation is just for the apps wich are compiled for Intel only.

 

[startac7868]

Okay, fine, software compiled for Windows 10 ARM will include Microsoft Office and browsers.

But these I use regularly. They are available typically for Windows, OSX and Linux. Where it's Windows 10, none are ARM native.

Webstorm, Eclipse
Filezilla
7-Zip
Adobe Acrobat
Balsamiq Mockup
Snag-It, Camtasia
GIMP
Inkscape
SketchUp
Synology CloudStation
WinHTTrack
Netlimiter

There are others. The point that I'm making is that almost all "useful" apps are not, and will not likely be for quite some time if ever, be available for ARM. Emulation is possible but running natively, without emulation overhead, is better.

 

[Cruncher04]

Webstorm, Eclipse
Filezilla
7-Zip
Adobe Acrobat
Balsamiq Mockup
Snag-It, Camtasia
GIMP
Inkscape
SketchUp
Synology CloudStation
WinHTTrack
Netlimiter

Just glancing over the list many of the apps are open source and can be compiled for ARM. So they will run native rather sooner than later. Similarly for .Net/CLI and Java apps like Eclipse as they do not need to be compiled for a specific processor architecture.

But that is not even the point. You need to have the devices on the market in order for developers target them. As such emulation is just a bridging solution and not the be all end all. Intel is far from being competitive in the mobile space so Microsoft has to support ARM with full Windows if they want to stay relevant.

 

[a5cent]

Okay, fine, software compiled for Windows 10 ARM will include Microsoft Office and browsers.

I also think the open source stuff could become available for ARM quite quickly. I also wouldn't be surprised if Office 365 ends up including binaries for ARM. Software that isn't compiled for ARM obviously will perform better on a real x86 CPU. As far as I can tell the advantages the SD835 retains (compared to Intel's upcoming 10nm chips) are mobile features (gps, radios, sensor hub, etc) and price... assuming Intel sticks with their current pricing policies, the ARM SoC will be about $200 cheaper than the comparable Intel Core M chip.


That leaves the SD835 with enough of a price advantage that we can't call one outright better. If you're judging only by performance with no regard for cost or battery life, then Intel will be better. If you care about perf/dollar or perf/watt, then things start getting tricky. We'll need real tests, with x86 and ARM software, before we can really determine that.

 

[Cruncher04]

If you're judging only by performance with no regard for cost or battery life, then Intel will be better.

I guess this statement is not true generally as is. If you look at SD835 single core CPU performance, it is above all Atoms and below Core M. And even this is only half the truth, because Core M is 2 core while SD835 is 4 core, so overall CPU performance of SD835 is already better than Core M.
In addition GPU performance is better than both.
perf/watt is generally much better with ARMs, that was true when Intel had the technological advantage and is even more true now, when the technological advantage is diminished.

 

[a5cent]

I guess this statement is not true generally as is. If you look at SD835 single core CPU performance, it is above all Atoms and below Core M.

As you said, single core performance of an SD835 lies below that of a Core M. The person you quoted explicitly asked about Cannonlake and there will be no Cannonlake based Atom. The statement is therefore generally true in this context, particularly since it is true even without figuring in the overhead of emulation which will in many cases (at least occasionally) be necessary.



As I've said a few times already, we really can forget about Atoms at this point. They are a thing of the past. Intel no longer designs or produces them.

 

[Cruncher04]

As you said, single core performance of an SD835 lies below that of a Core M. The person you quoted explicitly asked about Cannonlake and there will be no Cannonlake based Atom.

Well you mentioned Intel will be faster without being specific about which chip. Cannonlake is pretty much speculation right now because it is away at least yet another year. And even then it is doubtful that a 2 core Cannonlake M with around 4.5W TDP outperforms a 4 core SD835 with multithreaded code. Thats not even considering, that we would have a 7nm Snapdragon successor at the time Cannonlake arrives.

Now of course i did not talk about emulation.

 

[a5cent]

^ I didn't think I had to be specific, since the person you quoted already was. I was replying in that context. Sorry if that wasn't clear. Anyway, that includes the topic of emulation, as that was that person's concern, and that will be required at least some of the time. That can't just be ignored. The specifics of Cannonlake are certainly speculation, but so is most anything related to the SD835. Both will become available within a few months of each other (Cannonlake is scheduled for 2H 2017) so they will be direct competitors for most of their lifecycle. IMHO it makes sense to compare them. That assumes an SD835 successor isn't already released this year, but I could be wrong about that.

 

[Cruncher04]

The specifics of Cannonlake are certainly speculation, but so is most anything related to the SD835. Both will become available within a few months of each other (Cannonlake is scheduled for 2H 2017) so they will be direct competitors for most of their lifecycle.

According to the latest roadmap it is 4Q17 and rumor is, that due to problems with 10nm it is delayed to 1Q18. What is known already is that the Y series is only 2-core. In addition we are talking 4.5-5W TDP, which is too high for small tablets or phones.
And then of course Intel would need to upgrade the GPU significantly in order to be in the same league as SD835.

 

[a5cent]

^ Core M's have always been somewhat configurable. All are rated at 4.5W but have been able to go as low as 3.5W or even lower. That's fine for a phone sized device. I'm assuming 10nm and Cannondale based Core M's will have the potential to go even slightly lower.


I have not yet seen the Q1 2018 rumour. That would be pretty bad for something originally planned for 2016. Wait and see I guess...

 

[cactuspete23]

10nm is pretty amazing ! Wonder how much further technology will go; and how much more power reduction we can expect.

 

[startac7868]

10 nm is supposedly pretty much the physical limit for silicon. Intel has stated that they think they can go as far as 7 nm before having to switch from silicon to another substrate such as gallium arsenside. But given how long it has taken to go from 14 nm to 10 nm, it will be a long time to go from 10 nm to 7 nm silicon. And then, a totally new material like gallium arsenside will take years to perfect.

At 10 nm, I think we should be satisfied with living in the last golden age of silicon based chips.

 

[cactuspete23]

Wow, That's cool !
- Wonder what someone from the 1950's would think, or even understand, if they found a 10nm chip and tried to inspect/understand what it was for....

 

[a5cent]

^ they probably would have just barely been able to analyze and understand it in 1950. Just not produce it. The first transistor was invented in 1948. There are billions of transistors on a modern chip. The fist electron microscope, which would have been required for analytical purposes, was available in 1940. There have been no fundamental changes in all that time. Everything has just gotten very much smaller.

 

[startac7868]

There have been no fundamental changes in all that time. Everything has just gotten very much smaller.

And by physical limitations, it can't get any smaller unless a fundamental switch (transistor OR something else) and the material (semiconductor silicon TO something else) changes. Hence, as I said, enjoy the wonderful sunset of the age of silicon chips.

10 nm silicon chips will likely be the norm for at least 5 years. How's that for futureproofing. Architecture may change but the process won't. It used to be process-architecture tick-tock model. It recently became tick-tock-semitock (process-architecture-architecture enhancement) when attempts at the 14 to 10 nm process (the tick) hit a wall. It'll be tock-tock-tock for the foreseeable future.