Snapdragon 820 Previews

[JimmiG]

The 6P / S810 are not a good demonstration of big.LITTLE, the SoC is just simply bad in more ways than one. The 7420 has like twice over the perf/w and even the 5433 is far ahead.

In regards to Apple battery life etc, iOS itself is just incredibly more efficient than Android and I would bet a that has a lot to do with the fact that iPhones still compete with their little batteries. Again unless I get an iPhone and measure power in something like SPEC then I doubt large efficiency advantages.

Google should really work more on battery life improvements in the core Android OS. Battery life actually got worse from KitKat to Lollipop. I think they reversed it somewhat with Marshmallow, but it should be top priority for them.

 

[Thala]

For instance, heat generation is an indirect indicator of energy use of a CPU. It doesn't tell if it's resulting from handling the load, leakage or whatever but ultimately that's still expended energy. I'm not one of those who disparage "turbo" type methods but it's readily apparent that Apple's CPU's are able to sustain load without the same levels of heat generation as others. It's possible that Apple has superior heat dissipation but at steady state, it's not going a big difference compared to other aluminium case phones.

You get to the wrong conclusion, as heat dissipation is no indication of the inferiority of BIG.little. BIG.little has no advantage in high performance use-cases, however it shows its benefits in low-performance use-cases when implemented right.
In low performance use-cases power gets more and more dominated by leakage as the CPU load decreases. In high performance use-cases dynamic power is the dominant contributor to power.
So it is a very reasonable assumption that if you reduce powered area you get some reduction in overall power.
It is close to impossible, that a big core consumes the same or less power below a certain load threshold compared to a small core.

 

[vissarix]

what a complete fail for qualcomm, barely faster then an old exynos 7420...i wouldnt have expected that a 200b worth company does this kind of mistakes...they are going the amd route...

 

[ShintaiDK]

i wouldnt have expected that a 200b worth company does this kind of mistakes...they are going the amd route...

Its down to 73B today and still dropping. They will most likely get an EU fine like in China due to predatory pricing on chipsets. Dropping the value even further.

52 week range for the stock says it all. Peaked at 75, now down to 47. Their peak market cap was around 130B or so.

 

[lopri]

So you would argue that Geekbench is not a serious CPU test?

IMO It is still a better test than browser tests. It really makes me cringe time after time when I look at those meaningless-but-susceptible-to-misinterpretation browser tests charts AT put up in their reviews. They could make those a bit more meaningful by:

1. Using a same browser across the platform, such as Chrome or Firefox. This has an effect of evening the playing field and give better representative relative performance of an SOC.
2. Using the most optimal browser for a given platform to measure the true capability of a given SOC. This will take the bias introduced by sub-optimally tuned browsers out of the equation, and give the SOCs the most favorable environment to stretch their legs.

But for some reason they choose not do so, and worse yet they pick and choose browsers without explaining clear justification then mix the results up the presentation. This turns the charts near-meaningless and forces readers to go back to previous reviews to compare numbers in an objective manner of their own. Even that endeavor is a risky and fruitless enterprise because past tests are often run on different OS/browser builds. We know an updated OS can bring about a huge difference in Javascript performance.

I wish AT were more consistent and transparent in presenting these browser tests, or abandoned them altogether.

 

[poohbear]

man, Qualcomm stocks have TANKED the past couple of months weeks. I bought it at $53 thinking it bottomed out and then it drops even further to $47! and to think this stock was averaging $70 a share before the summer. Ugh. Everybody has lost confidence in em.

 

[el etro]

Qualcomm rocking again.

Single threaded bench, Snapdragon S820 versus S810 and especially their last core, Krait400 from S800:

Amazing for Android users phones. Didn't outperform Apple, but took them back in the game again. If the claimed CPU efficiency gains is delivered, they remain winner for one more round.

 

[Nothingness]

Their Chrome score is much worse than 2426, at 5184. We should wait for production devices to see how things turn on.

 

[Arachnotronic]

At this point I don't know why Qualcomm designs its own cores. A good A72 implementation would have done the job just as well and have been cheaper to do.

They really like milking the "custom core" angle because everyone seems convinced that as long as yours is "custom" its better than ARM's which isn't always true.

 

[StrangerGuy]

At this point I don't know why Qualcomm designs its own cores. A good A72 implementation would have done the job just as well and have been cheaper to do.

They really like milking the "custom core" angle because everyone seems convinced that as long as yours is "custom" its better than ARM's which isn't always true.

Samsung and Qualcomm is in deep Apple envy mode, thinking that making a custom chip would turn the tide of top to bottom Android commoditization, when that obviously won't work when they don't even own the software stack.

Android OEMs would be lucky to ship a total 50M high end devices next year. Apple can ship that in just one quarter these days.

 

[el etro]

Their Chrome score is much worse than 2426, at 5184. We should wait for production devices to see how things turn on.

But at chrome optimization is even more poor for the S820. Galaxy note 5(basically the same chipset as Samsung Galaxy S6) also scores 30%-40% less in that bench in the first graph.

 

[linkgoron]

I don't get why people use browser benchmark scores as "CPU" benchmarks.

Safari is basically hand tailored to Apple hardware (metal or whatever APIs iOS uses and the small amount of different hardware specifications), and this gives the iphone 6s a huge advantage from the get-go.

Chrome (or any other Android browser), by the way that Android is built, is much less optimized by default. I think that just the 800/801/805/808/810/exynos are more CPUs than what CPUs that the latest iOS runs on (and you can also add the 6/4 line and mediateks).

BTW I'm not talking about user experience - just talking about a benchmark which is in any way meaningful to compare CPUs.

 

[RussianSensation]

Samsung and Qualcomm is in deep Apple envy mode, thinking that making a custom chip would turn the tide of top to bottom Android commoditization, when that obviously won't work when they don't even own the software stack.

Android OEMs would be lucky to ship a total 50M high end devices next year. Apple can ship that in just one quarter these days.

Do you think it would make much difference even if Android flagships outperformed iPhones? I doubt it would. Almost no one I know who owns an iPhone has actually checked its benchmarks against Android phones before purchase or used CPU/GPU benchmarks as a big reason to buy one over the other.

Wouldn't you say that overall S6/Edge was a faster phone than iPhone 6?
http://www.anandtech.com/show/9146/the-samsung-galaxy-s6-and-s6-edge-review/5

That didn't really help it during Q2/Q3 2015. Not only that, S6/Edge absolutely owned iPhone 6 in 2D and 3D video IQ and screen quality (OLED > IPS) but that didn't matter either. Even now I think S6/Note 5's selfie and rear camera are better than the 6S/6S+'s and their screens are also far superior, but most of the market doesn't care at all. There is a lot more to people buying iPhones over flagship Android phones than CPU/GPU benchmarks.

I know it sounds really bad but I'll just say it -- you can attract young, good looking females in many poor countries simply by owning an iPhone or gifting them one - it acts as an unspoken status symbol that projects wealth. I've never seen a flagship Android phone pull something like that off.

Why do you think people pay $250-300 US for Nike running shoes in Central Asia, Brazil, Eastern Europe or Russia? An iPhone itself is a status symbol, much like driving a Mercedes or BMW since not everyone is a "car guy/gal".

There are other big selling points to iPhones that Android cannot compete with:
- long-term software updates/support
- higher resale value
- can sell an iPhone in almost any country for quick cash if need emergency $ -> iPhone is more liquid 'asset' than an Android phone in my experience
- Certain U.S. designed (i.e., American) products sell MUCH better in the U.S./Canada than elsewhere - Xbox One, Corvette, Mustang, etc.
- iPhone breaks and you can exchange it for a new one at any Apple store location, same day. Samsung, HTC, Asus, Lenovo, BLU, Sony, etc. cannot compete against that.

I also think loyalty is a big part of it. For example, let's say a consumer started with iPhones years ago. To switch, they are thinking I have to learn something new, research all these complicated Android choices. It makes it a lot easier/safer to buy what you know and they have used iOS for a long time + your buying decision is validated since so many other millions of customers own it and like it so it's a safe choice.

That's why I think even if 820/8890 were to smoke the A9/A9X in every benchmark, that's not sufficient for Android flagship sales to take off.

 

[coercitiv]

Open Snapdragon 820 thread: read tutorial on how to bang good chicks with an iPhone in many poor countries.

 

[ChronoReverse]

You get to the wrong conclusion, as heat dissipation is no indication of the inferiority of BIG.little. BIG.little has no advantage in high performance use-cases, however it shows its benefits in low-performance use-cases when implemented right.
In low performance use-cases power gets more and more dominated by leakage as the CPU load decreases. In high performance use-cases dynamic power is the dominant contributor to power.
So it is a very reasonable assumption that if you reduce powered area you get some reduction in overall power.
It is close to impossible, that a big core consumes the same or less power below a certain load threshold compared to a small core.

That's assuming you have only the small core and not the required additional complexity to get big.little to work. Which nobody has yet to get behaving entirely well (not even Samsung's phones in the field, i.e., not testing conditions). Now that I think about it, this is likely the reason why Google Camera is hardcoded to use the little cores only on the Nexus 6P. It's simply easier than letting the automatic mechanisms try to guess (makes me wonder if this is one of those problems that actually can't be optimally done automatically).

Furthermore, when you're in the low CPU utilization range where a little core would be useful, you're talking about a small amount of energy usage overall anyway. It's a case of penny-wise pound-foolish where small gains in low CPU utilization scenarios are negligible compared to the screen, radios and even short bursts of load.



The obvious rebuttal is that the big.little system just has to be implemented properly but it doesn't seem like anyone is even close to getting it to work well enough that you retain performance and have it work without hiccups. Samsung has been working at this since the Galaxy S4 and even with the Note 5 the results aren't really that impressive (it's just great compared to the trainwreck the SD810 is).

 

[dark zero]

At this point I don't know why Qualcomm designs its own cores. A good A72 implementation would have done the job just as well and have been cheaper to do.

They really like milking the "custom core" angle because everyone seems convinced that as long as yours is "custom" its better than ARM's which isn't always true.

And nVIDIA had the chance on ARM, why they gave up on that? their cores were the best along Apple....

 

[Arachnotronic]

And nVIDIA had the chance on ARM, why they gave up on that? their cores were the best along Apple....

Denver wasn't great. It did well in GB3 but it seemed to be a very fragile architecture. Also seems to have been buggy...

 

[Thala]

Furthermore, when you're in the low CPU utilization range where a little core would be useful, you're talking about a small amount of energy usage overall anyway. It's a case of penny-wise pound-foolish where small gains in low CPU utilization scenarios are negligible compared to the screen, radios and even short bursts of load.

It might be penny-wise but it certainly is not pound-foolish. In addition don't assume that the screen is always on, think about mp3 playback, pulling you email etc. And finally leakage becomes more and more dominant as the technology shrinks. So the effect of BIG.little will become more apparent as the technology progress.

 

[Hans de Vries]

Great Antutu V6.0 scores: 132.268 for the Snapdragon 820 (msm8996)

compared with

More GeekBench scores coming in now.

GeekBench Snapdragon 820 vs. Kirin 950
GeekBench Snapdragon 820 vs. Apple A9

This Leak from November 15th seems spot on:

2016: The year of the Desktop Replacement Phone.

Imagine the Snapdragon 820 in the coming Lumina phone,
liquid cooled when USB-C docked... or in this phone

Microsoft seems to be at least a year ahead of the pack when it
comes to Desktop Replacement Phones. Actually the A10 rumor with
six cores might make good sense for Apple.

Think about a USB-C docked, liquid cooled iPhone 7, running MAC
OS-X on 4 cores while still functioning as a phone with IOS on the
other 2 cores. Imagination virtualized the PowerVR 7 series to serve
multiple operating systems with display graphics simultaneously.

and then do the compile trick using LLVM IR bitcode in the App store...

 

[poohbear]

Great Antutu V6.0 scores: 132.268 for the Snapdragon 820 (msm8996)

compared with

More GeekBench scores coming in now.

GeekBench Snapdragon 820 vs. Kirin 950
GeekBench Snapdragon 820 vs. Apple A9

This Leak from November 15th seems spot on:

2016: The year of the Desktop Replacement Phone.

Imagine the Snapdragon 820 in the coming Lumina phone,
liquid cooled when USB-C docked... or in this phone

Microsoft seems to be at least a year ahead of the pack when it
comes to Desktop Replacement Phones. Actually the A10 rumor with
six cores might make good sense for Apple.

Think about a USB-C docked, liquid cooled iPhone 7, running MAC
OS-X on 4 cores while still functioning as a phone with IOS on the
other 2 cores. Imagination virtualized the PowerVR 7 series to serve
multiple operating systems with display graphics simultaneously.

and then do the compile trick using LLVM IR bitcode in the App store...

I remember Motorola tried that 5 years ago with their atrix phone....they simply didn't have the software ecosystem to male it popular. But with Windows, this can really take off if they make it idiot proof! Good on them for thinking ahead.

 

[Arachnotronic]

That A9 AnTuTu score is low. My 6s Plus gets 131362.

 

[ShintaiDK]

You also got a TSMC chip

 

[Arachnotronic]

You also got a TSMC chip

The Samsung-built A9 in my iPhone 6s (non-plus) scores 129446. I ran this test with battery down to just 30%.

 

[Arachnotronic]

My 6s Plus results:

 

[Face2Face]

My 6s Plus results:

I just got this score with my Samsung made A9

iPhone 6s Plus