Qualcomm's Snapdragon 835 is Its First 10-Nanometer SoC (engadget.com)

An anonymous reader writes: Chipset maker Qualcomm has unveiled its next-gen Snapdragon 835 flagship SOC and confirmed rumors that it will be built by Samsung using its 10-nanometer FinFET process. Compared to the current 14-nanometer Snapdragon 821 (also built by Samsung), the new CPU packs 30 percent more parts into the same space, yielding 27 percent better performance while drawing up to 40 percent less power, the company says. It also improved the design, which will yield "significant" improvements to battery life. The new chip comes with Quick Charge 4, which supports 20 percent faster charging than Qualcomm's last-gen tech. That, the company says, will give you up to five hours of extra battery life with just a five-minute charge. In just 15 minutes, it'll give Snapdragon 835 phones a half-full battery.

Comments

5 hours?

[sims 2]

How long is this thing supposed to run?
Most smartphones only run like 3 hours tops if your actively using the screen.

Re:5 hours?

[Anonymous Coward]

Obviously it depends on battery capacity and screen brightness

Re:5 hours?

[Dorianny]

How long is this thing supposed to run? Most smartphones only run like 3 hours tops if your actively using the screen.

screens and LTE radio's are the 2 biggest drains on battery life in mobile devices. A more efficient SOC is not going to do anything to change that

Re:5 hours?

[Hadlock]

In most cases that I've seen, the person forgets to charge it overnight, and then needs the phone to navigate to a new address when traveling, and check in on emails periodically. Typically if you can get the phone to 50% you can nurse it the rest of the day, or at least until you can plug it in periodically throughout the day to get you back up above ~30%

Re:5 hours?

[guruevi]

Even Apple phones don't drain that quickly. Mine is 3+ years old and still holds a good 8h of "active" time (regularly checking mails, a few phone calls), it used to be two full days when I purchased it although my activity has increased.

Re:5 hours?

[K. S. Kyosuke]

What if the LTE electronics is on the SoC?

Re:5 hours?

[swillden]

What if the LTE electronics is on the SoC?

Doesn't matter. The big drain is powering the radios. It takes juice to transmit signals strongly enough that distant towers can pick them up, and to boost weak incoming signals. More efficient local processing doesn't make a significant difference.

Re:5 hours?

[Archangel Michael]

Trust me when I say this, Apple iPhones aren't all that special. Given the "Ingress" test is the most brutal of "Performance" testing, iPhones do not perform all that much better than similarly equipped Android phones.

The iPhone people I know do not spend significant amounts of time on screen, except when watching movies.

Re:5 hours?

[K. S. Kyosuke]

Boosting weak incoming signals definitely shouldn't be power-hungry. Transmitting is. But as features get shifted into digital circuitry, power-efficient digital circuitry should definitely help with power-efficient communication no matter what direction we're talking about.

Re:5 hours?

[kangsterizer]

tldr: sorry, good intentions but you're wrong, I'm right, etc.

The way the main SOC (ie the snapdragon SOC) decides to power devices on/off (such as radio) and the efficiency of their regulators does change battery life significantly though when the device is not under active use.

But in addition to that, the way the SOC manages it's own cores, and the way the core are built very significantly changes battery life under active use.

But you don't have to believe me, because Samsung builds the SAME phone with 2 chipsets: the galaxy S7/S7Edge comes with the same specs and either a Snapdragon 802 from Qcom or an Exynos from Samsung themselves.
The Exynos Galaxy S7 Edge lasts 2-3 HOURS more than than the Snapdragon version, under active use.

Random test: https://www.youtube.com/watch?... (there's plenty more)

Re: 5 hours?

[K. S. Kyosuke]

I'm not missing anything, it's just that 1) the RX pre-amplifier is a low power device and 2) replacing analog processing with digital enables to lower power consumption in step with digital fabrication process improvements, so if you, for example, come up with ways of improving communication that lower energy per bit transferred in the analog part but require more processing in the digital part (improved self-correcting codes, for example), then digital process improvements may be a prerequisite for such improvements.

Quick Charge?

[Yvan256]

Maybe they could sell their technology to Tesla, Nissan, Chevrolet, etc.

Re:Quick Charge?

[OrangeTide]

Right, because scale isn't relevant here.

Re:Quick Charge?

[GabeGhearing]

You still have to get the power into the Lithium cells quickly. The Power Management Module(PMM) inside the CPU regulates the rate at which power enters/leaves the lithium cells. This Snapdragon supports taking in larger amounts of power and shoving it into lithium cells faster.

Re:Quick Charge?

[ChunderDownunder]

I would settle for a phone that didn't explode!

Re:Quick Charge?

[GNious]

Well, clearly it isn't, since 15 minutes will charge any phone to half, no matter the battery, or usage during those 15 minutes.

Re:Quick Charge?

[thegarbz]

I would settle for a phone that didn't explode!

So buy literally any phone on the market?

Re:Quick Charge?

[OrangeTide]

Well the differences between cell phones is probably only a factor of 4.
A Tesla battery is a factor of 8000.

Re:Quick Charge?

[kangsterizer]

A phone battery is 1S lipo/lion of about 3000mah/12Wh. At 1C (3A) it takes 1h to charge. At 4C it takes 15min. So Qcom charges at 4C or 12A.
A Tesla car battery is 27777777mah/85-100kWh. At 1C it is 27777A. thats.. thats a lot of energy.

A small household uses about 100A/day (fridge, oven, burns, heating, tv, etc. Rough estimate). So you'd have to charge, in 15min as much energy as the household would spend in 277 days. That should give an idea of how much more energy is stored in a Tesla car battery vs a smart phone and why the smartphone can charge in 15min...

Heck my RC batteries can charge at 10C (i.e. more than twice faster than the qualcom phones that aren't even out yet!)
The Tesla supercharges charges at 125A IIRC, which is way more than .. your typical house fuse can deal with.

Just for fun, charging a 100kWh Tesla battery pack at 4C is 400kWh. That's more than 1000A (3.6V per cell).

Re: Quick Charge?

[Anonymous Coward]

Do you do units ? What do mah and A/day mean ?

27% better performance?

[110010001000]

27% faster than the 821 which was 10% faster than the 820. Moore's law is alive and well I see.

Re:27% better performance?

[alvinrod]

Moore's law was about the amount of time it took to double the transistor density in a given amount of area. It's still going, it's merely that its no longer doubling quite as fast as it used to. Part of this is due to the difficulty of shrinking the process features when they get closer to the physical limits, but also because CPUs have reached a point where the performance is more than most consumers need so there's less demand for more powerful chips. I've still got an old Core 2 Duo machine that's mainly used for light web browsing. The only real reason to replace it would be to get something more power efficient.

Mobile still has some of that drive for better performance as there's still a demand for the same order or performance, but there are some design constraints that made previous SoCs less powerful than desired. I think we'll get to another saturation point soon, but there'll be some new product in the future to replace phones that requires an even smaller package so the demand for performance will begin anew.

Re:27% better performance?

[Anonymous Coward]

How is Moores Law still going if it is no longer doubling as fast as it used to? That was the whole point of Moores Law. Transistor count hasn't doubled in a long time.

Re:27% better performance?

[willy_me]

Specifically, Moores Law states that the complexity (ie, number of transistors) for minimum component cost increases by a factor of two per year. Translating this to computer speed or process size is easier said then done. Several other factors impact the minimum component cost. For example, the number of defects per unit area of a specific process. And when it comes to the actual parts manufactured, designers could opt for more transistors or a reduced, albeit not as efficient (price/transistor), final price. Either way, Moores Law still applies.

Re:27% better performance?

[K. S. Kyosuke]

Moore's law allows us to go parallel and that really helps with power/performance ratios.

Quick Charge?

[Game Genie]

Okay, 5W@500mA is horrible, and we needed proprietary solutions to get around this on USB 2.0, but now that there is actually a standard (USB-C Power Delivery) why are they still messing around with this silliness?

Re:When will a phone be out that has this?

[Game Genie]

I stumbled on my original first gen iPhone last week. It's so nice and small. Fits in my hand great. Disappears in my pocket. Almost makes me wish I could get a flagship smartphone with a screen that small (surely I would feel differently if I actually powered it on). Sadly I couldn't find an old 30 pin cable to power it up and see if it still works.

Re:When will a phone be out that has this?

[known_coward_69]

the SE is just a little bigger than the 4S and is basically an iphone 6S with no force touch and a smaller screen

First to announce a 10nm SoC

[edxwelch]

Qualcomm is the first to announce a 10nm SoC, but there are in fact several 10nm SoC's currently in production.
There's Samsung's Exynos and MediaTek's Helio X30 and Apple's A10x which will appear in the new iPad, all due out in first half of 2017.

Re:First to announce a 10nm SoC

[RightSaidFred99]

None of them are "real" 10nm, it's a big marketing con. It's closer to Intel's 14nm.

Re:First to announce a 10nm SoC

[edxwelch]

"Real" 10nm would be where the gate length is 10nm. For that definition Intel's 10nm isn't real 10nm either.

Re:First to announce a 10nm SoC

[RightSaidFred99]

Not really, there are a bunch of different pieces you can look at.

It's all marketing. The only thing that matters is price and performance (and dependencies on those things like yield).

Quick Charge 4

[Anonymous Coward]

The standard Google said is dangerous, and likely causing phones to catch fire.

Oh, the chip's made by Samsung, well, there you go, explosive news.

Quick Charge? Hope they fixed it.

[bobbied]

Wasn't that the major problem with the Note 7 catching fire? Overheating the battery during a "quick charge" causing damage then poof, phone up in smoke?

Re:When will a phone be out that has this?

[unixisc]

You could get one of those micro-USB to 30 pin adapters for $20-25, and use it w/ any micro-USB cable that you have.

Can it be?

[archer, the]

The first settler of Qualcomm!

Intel???

[Anonymous Coward]

This article is about Qualcomm. Why on earth is Slashdot using the Intel logo?

Re:Intel???

[Rockoon]

maybe because intel doesnt know how to make 10nm production work yet

Re:Intel???

[Rockoon]

Fine, Intel can't get their "10nm" production working yet.

You did know that Intel was the first semiconductor manufacturer to lie about feature size, right?

..or perhaps the Intel fanboys have convinced you that everyone but Intel is doing it? Intel fucking invented it 5 or 6 enerations ago. None of their 14nm chips have any feature that is even close to 14nm in size.

Wow 10nm... almost at the theoretical limit of 7nm

[Proudrooster]

Wow, this is just unbelievable. A 10nm die. It is so amazing how far we have come in just 30 years. I am truly impressed and amazed that we can carry around a 2Ghz Quad core in our pocket that is a fully functional computer. It is truly impressive, congratulations everyone!

27% faster OR 40% less power

[viperidaenz]

From TFA:

Samsung’s new 10nm FinFET process, for instance, allows up to a 30 percent increase in area efficiency with a 27 percent improvement in performance or up to 40 percent less power consumption compared to the previous version.

Re:27% faster OR 40% less power

[aglider]

I would prefer 20% less power AND 10% faster!

Re:27% faster OR 40% less power

[Agripa]

And I would prefer 100% less fire.

Intel

[aglider]

Why this article has been labelled as Intel?
I know Intel would like to buy Qualcom but, as far as I know, this hasn't happened yet.

Re:Wow 10nm... almost at the theoretical limit of

[LordWabbit2]

The first computer my parents bought had a 4.7Mhz processor (single core of course) and a whopping 20mb hard drive, display had a max of 4 colours, I forget the resolution. And no it was an 8088, not a zx etc.
Compared to that I have a super computer in my pocket, lets not even bother comparing it to my gaming rig at home.