Samsung Exynos Thread (big.LITTLE Octa-core)

[Trumpstyle]

60 fps offscreen for 880 mali just seems unreal. Intel skylake 530 scores 70 fps and xbox one gpu get around 150 fps. Would be nice surprise if true though and if true no way intel ever getting into mobile.

 

[Sweepr]

For what it is worth:

Google's translation:

Thanks. Single-thread performance on par with Apple's A9 and more than twice the MT performance if the leaks are accurate. 65-95% better graphics performance too.

Apple A9 leak
Geekbench 3 single-core: 2090 points
Geekbench 3 multi-core: 3569 points
GFXBench T-Rex offscreen: 66 FPS
GFXBench Manhattan offscreen: 30.3 FPS

Samsung Mongoose Exynos leak:
Geekbench 3 single-core: 2136 points
Geekbench 3 multi-core: 7497 points
GFXBench T-Rex offscreen: 108.9 FPS
GFXBench Manhattan offscreen: 59.4 FPS

Of course efficiency and sustained performance metrics could look different.

60 fps offscreen for 880 mali just seems unreal. Intel skylake 530 scores 70 fps and xbox one gpu get around 150 fps. Would be nice surprise if true though and if true no way intel ever getting into mobile.

True, ARM itself promises:

In all, the Mali-T880 has 1.8 times more performance and is 40% more energy efficient when compared to 2014 Mali-T760 GPU-based devices.

1.8x the performance of Galaxy S6's Mali T760MP8 would provide ~47 FPS.
As always take the scores with a huge grain of salt.

 

[Trumpstyle]

True, ARM itself promises:



1.8x the performance of Galaxy S6's Mali T760MP8 would provide ~47 FPS.
As always take the scores with a huge grain of salt.

Dude could it be possible that samsung will use there 14nm finfent lpp instead of lpe which can give slight boost and maybe with wide i/o 2 to reduce size of the chip they could squeeze and 1-2 extra gpu cores maybe?

 

[Hans de Vries]

Maybe "too many leaks" over there , how about this one:

老黄的下一代处理器，Tegra X1的下一代丹佛版本样品跑分曝光： CPU方面，Geekbench3单线程：2599，多线程：4691 GPU方面：GFXbench3.0 曼哈顿离屏：117fps 霸王龙离屏：206.9fps 四季度或明年初试生产，工艺1xn Finfet，帕斯卡架构彪悍，新丹佛架构依然侧重整数运算。 老黄放弃手机平板，性能开满血

google's translation:

Huang's next-generation processor, Tegra X1 Denver version of the next generation to run sub-sample exposure: CPU aspect, Geekbench3 single-threaded: 2599, multithreading: 4691 GPU aspect: GFXbench3.0 Manhattan offscreen: 117fps Tyrannosaurus offscreen: 206.9 fps in the fourth quarter or next year the first test production, process 1xn Finfet, Pascal sturdy architecture, still focusing on the new Denver architecture integer arithmetic. Huang to give up the phone flat performance full of blood

compared with:

28nm 2.5GHz Tegra K1 with Denver ( HTC Nexus 9 )
GB3 single-threaded: ~2000
GB3 multi-threaded: ~3500

20nm Tegra X1
Manhattan offscreen: 66 fps
Tyrannosaurus offscreen: 124 fps

 

[soccerballtux]

still perfectly happy with my LG G2's S800 Cpu...

not sure why people need/want faster.

for now, [well, I don't game] I think 4GB RAM is more important than the faster CPU.

If you're kind to your phone, good case and all, I could see it lasting a good 5 years at this point.

 

[Systems analyst]

ARM gave some performance data in :
http://www.anandtech.com/show/9184/arm-reveals-cortex-a72-architecture-details

Comparing A57 and A72, with both at 28nm:
"... what we are looking at is a general 16-30% increase on IPC depending on the kind of workload."

It seems that ARM are making rapid progress in developing their micro-architectures, over and above the improvements in manufacturing nodes. My impression is that mobile-phone SoCs are by no means mature; there is plenty of room for improvement, as evidenced by the first figure in the above reference:

A15 2014: 1.0 reference level performance.
A57 2015: 1.9 times performance of A15.
A72 2016: 3.5 times performance at 2.5 GHz in a mobile phone, using the target production process of 16nm FF+.
The latter implies 4.2 as fast at 3GHz, in a larger form.

The conclusion is that these leaked benchmarks are preliminary implementations; we need to wait until next year to see the full potential of the ARM A72, especially in larger form-factors, with large screens.

 

[Nothingness]

The figures ARM gave were for power efficiency, not for raw perf. This was confirmed by Peter Greenalgh, an ARM engineer, on RWT.

 

[DeathReborn]

I wonder if ARM gets any info on the architectural changes Qualcomm, Apple & Nvidia etc make to their custom chips for possible use in future standard ARM cores.

 

[Systems analyst]

The figures ARM gave were for power efficiency, not for raw perf. This was confirmed by Peter Greenalgh, an ARM engineer, on RWT.

See the 'performance' tab on:

http://www.arm.com/products/processors/cortex-a/cortex-a72-processor.php

which says:

"The Cortex-A72 delivers 3.5x the sustained performance in the smartphone power envelope over 2014 28nm Cortex-A15 processor designs. The processor features several major micro-architectural improvements which build on the current generation of ARMv8-A cores. The enhancements in floating point, integer and memory performance improve the execution of every major class of workload.

The processor is optimized for the 16nm FinFET process technology, enabling the Cortex-A72 to clock up to 2.5GHz in the mobile power envelope and leading to even higher total delivered performance.

On top of these key performance improvements, the Cortex-A72 CPU also benefits from significantly lower power consumption. This improved efficiency combined with the 16nm FinFET process technology enables the Cortex-A72 processor to achieve a 75% power reduction in representative premium mobile workloads. "

This seems to clearly distinguish between performance and power consumption/efficiency.

I could not find any reference to A72 or Peter Greenhalgh on www.realworldtech.com; can you give me a specific reference?

 

[Nothingness]

Never trust marketing, never

http://www.realworldtech.com/forum/?threadid=147766&curpostid=147801

 

[krumme]

Notice its 3.5 times the performance in THE SMATPHONE POWER ENVELOPE
Notice it also spans 28nm over 20nm to 16nm finfet.

But anyway the raw performance increase for a72 with less power (how much 2% lol) over the a57 in same process node (as claimed) is extreme compared to x86 improvements.

 

[Systems analyst]

Peter Greenhalgh said:

"Firstly, there are micro - architectural enhancements throughout the Cortex - A72 design which improve both IPC and power. In fact, the Cortex - A72 power improvements are achieved on the same process with the same library as Cortex - A57.We aren' t relying on a process shrink to achieve the power improvement or boost performance purely through frequency.Depending on the workload we' re seeing anywhere between 10 - 50 % more clock - for - clock performance than Cortex - A57 under identical system conditions while also reducing power.I' m talking about a range of decent sized, representative workloads, not micro - benchmarks.More information on the micro - architecture and said workloads will be disclosed in due course!

With respect to the performance uplift numbers over Cortex - A15 and Cortex - A57 that we quoted at the launch.These do factor in process technology.Cortex - A15 is predominantly a 28 nm product, Cortex - A57 is mainly on 20 nm and Cortex - A72 we expect to see on 14 nm/16 nm.While the audience here on RWT is very technical and understands the nuances of process geometry, many of the press we speak to want an appreciation of how much faster the processor in their favourite phone, tablet, may be in the future.To that end we quote performance uplift as a combination of the IPC improvement and the extra frequency afforded to our silicon partners due to advances in process technology.Overall we' re extremely pleased to have increased IPC while meaningfully decreasing power across various use cases.It' s going to be great to see Cortex - A72 in partner platforms.As we' ve got more than 10 partners extremely active with the processor I hope the wait won' t be too long![in a big.LITTLE configuration with Cortex - A53 of course]"

My reading of this is that 'performance' clearly refers to throughput, and 'power' clearly refers to the rate of dissipating energy. It seems clear to me that the A72, at 16nm FF+ should have a remarkable performance - next year. We will see.

 

[Sweepr]

The Mobile CPU Core-Count Debate: Analyzing The Real World

http://www.anandtech.com/show/9518/the-mobile-cpu-corecount-debate

What is clear though albeit there are corner-cases, is that the vast majority of applications do seem to be optimal for quad-core SoCs. This is why traditional 4-core and 4.4 big.LITTLE designs still appear to make the most sense in terms providing a balanced configuration and making most use of the hardware at hand. For big.LITTLE, even if there were no use-cases where all cores are concurrently used, it's not a big deal as what we are aiming for in heterogeneous systems is power efficiency gains.

This is also the point of the discussion where the debate of the potential detrimental effect of having more cores comes into play: The fact that a SoC has more cores does not automatically mean it uses more power. As demonstrated in the data, modern power management is advanced enough to make extensive use of fine-grained power-gated idle states, thus eliminating any overhead there might be of simply having more physical cores on the silicon. If there are cases (And as we've seen, there are!) which make use of more cores then this should be seen purely as an added bonus and icing on the cake.

Includes a lot of Exynos 5433 and Exynos 7420 data. Andrei proposes an interesting discussion, weighing the benefits of 4, 8 and 10-core mobile SoC designs.

 

[soresu]

In response to the earlier post about fishy mali t880 score, remember that it is also 40% more efficient than t760 according to ARM, which I think allows them to clock it even higher (t880 max = 850mhz, t760 max = 700mhz?).

Allowing for extra clock and likely driver optimisation too, an extra 10 fps or more (over 47 fps) is plausible?

 

[Sweepr]

In response to the earlier post about fishy mali t880 score, remember that it is also 40% more efficient than t760 according to ARM, which I think allows them to clock it even higher (t880 max = 850mhz, t760 max = 700mhz?).

Allowing for extra clock and likely driver optimisation too, an extra 10 fps or more (over 47 fps) is plausible?

I hope you are right but Mali T760MP8 already operates at 772MHz (boost to 852MHz).

 

[Fjodor2001]

The Mobile CPU Core-Count Debate: Analyzing The Real World

http://www.anandtech.com/show/9518/the-mobile-cpu-corecount-debate



Includes a lot of Exynos 5433 and Exynos 7420 data. Andrei proposes an interesting discussion, weighing the benefits of 4, 8 and 10-core mobile SoC designs.

There's a separate thread on this topic on the forum:

http://forums.anandtech.com/showthread.php?t=2445110

 

[lopri]

Interesting quote by Samsung with regard to their follow-up 14nm process.

Samsung’s 14nm FinFET Process Offering ⁃ 14LPE and 14LPP.

14LPE (Early edition) targets the early technology leaders and time-to-market customers such as mobile application SoCs to meet the latest mobile gadgets’ aggressive schedule and improved performance/power requirements. 14LPE is the first foundry process technology manufactured in the foundry industry with the successful volume ramp-up. 14LPE offers 40% faster performance; 60% less power consumption; and, 50% smaller chip area scaling as compared to its 28LPP process.

14LPP (Performance boosted edition) is the 2nd FinFET generation which the performance is enhanced up to 10%. 14LPP is the single platform for every application designs with the improved performance for computing/Network designs and the lowered power consumption for Mobile/Consumer designs. 14LPP will be the main process technology offering in 2016 and after.

Both of 14LPE and 14LPP finished its product level process qualification with real silicon data. Multiple customer designs were taped-out and some of them are under mass production now.

http://www.samsung.com/semiconductor/foundry/process-technology/14nm/

 

[Sweepr]

Exynos 5433's performance is still pretty solid for a 8'' tablet (should be able to keep up with a hypothetical A8-based iPad mini 4), though Exynos 7420 or a custom tablet chip would be a better pick for the 9.7'' version IMHO (Galaxy Tab S2). A welcome bump from last year's Exynos 5430 1st generation model.

Samsung Galaxy Tab S2 Gaming + Performance Test
www.youtube.com/watch?v=RidL6W1WGx8

Samsung Galaxy Tab S2 Gaming Performance
www.youtube.com/watch?v=z8aOoDbtcNs

Samsung Galaxy Tab S2 - Playing NOVA 3 - Performance test
www.youtube.com/watch?v=p3w2n5uwJHQ

Samsung Galaxy Tab S2 - Gaming Review
www.youtube.com/watch?v=gDKcBfz8g-c

Samsung GALAXY Tab S2 - Benchmark (1st Generation vs 2nd Generation)
www.youtube.com/watch?v=onzPYDZ8Tvk

Reviews with test results:
www.gsmarena.com/samsung_galaxy_tab_s2_97-review-1290p3.php
www.laptopmag.com/reviews/tablets/samsung-galaxy-tab-s2-8-inch

It's a real shame they didn't bother to enable 64-bit support for this chip.
Waiting for AnandTech's review, will there be one?

 

[Nothingness]

(...) though Exynos 7420 or a custom tablet chip would be a better pick for the 9.7'' version IMHO (Galaxy Tab S2). (...)
It's a real shame they didn't bother to enable 64-bit support for this chip.

You summed it up so well The lack of 64-bit OS is what will make me not buy it...

 

[krumme]

You summed it up so well The lack of 64-bit OS is what will make me not buy it...

I have the s2 tab 8.0 and its fast. For browsing it newer gets to 100% in cpuz and the a57 cores very seldom goes to 1900Mhz.
The lack of 64bit anoys me as a nerd but it frankly doesnt make a difference in general usage and browsing as the integer perf is the same if not a little better on the 32bit set.
Sujectively its very much faster than qcom s800 in android 5.0.
I have a 50mbit very fast line and asus 68 router and its not the cpu lacking.
What it needs is more efficiency. Better cpu govenor and later migration between cores as the newer 14nm variant.

Fantastic screen btw.

 

[Sweepr]

Some updates about Exynos.

1 - Samsung will be using their own SoCs for some mid-range phones as well. We might have different Exynos SoCs across their lineup in 2016.

Exynos 7880 is a brand new processor which will be used for Galaxy smartphones next year. It’s most likely going to be used in premium mid-range smartphones such as the Galaxy A series. The 64-bit chip has 1.8GHz ARMv8A Cortex-A72 + 1.3GHz ARMv8 Cortex-A52 in big.LITTLE configuration coupled with an ARM Mali-T860MP4 GPU.

As per the report Samsung also has the Exynos 7650 in the pipeline, based on the 28nm process it’s a 64-bit chip with 1.7GHz ARMv8A Cortex A-72 + 1.3GHz ARMv8 Cortex-A53 in a big.LITTLE configuration and an ARM Mali-T860MP3 GPU. It’s expected that Samsung will use both processors for 2016 Galaxy smartphones, most likely the ones that are targeted towards the mid-range segment of the market.

www.sammobile.com/2015/10/19/exynos...sors-will-be-used-in-2016-galaxy-smartphones/

2 - Galaxy S7 is rumoured to be based on the long waited Mongoose Exynos M1 SoC in some countries (Korea, Japan, Europe?) while the rest gets Snapdragon 820. There's also an India version running Exynos 7422 (supposedly an upgraded Exynos 7420), probably lower price.

www.sammobile.com/2015/10/07/samsun...y-s7-variants-each-with-a-different-processor

3 - Custom iGPU is the next step.

Samsung is quite capable of making its own processors, the biggest endorsement of this came from the company itself when it decided to use its own Exynos chipset in the Galaxy S6 and the Galaxy S6 edge as opposed to the Snapdragon 810. There have been rumors that in the future Samsung will also develop its own GPUs or graphics processor, according to a new rumored roadmap Samsung is going to start using its own GPUs with Exynos chipsets by 2017 or 2018.

4 - Latest 'leaked' Mongoose Exynos M1 Geekbench scores (take it with a grain of salt):

Coming towards today’s scores, the Exynos M1 will be coming with a frequency of 2.3 GHz with a couple of different performance modes on board. These will be a standard performing mode, a power saving mode and an ultra power saving mode, for when you’re really out of batter. In single core results for standard performance, the chip manages to score 2294, while the score for multi-thread performance is 6908.

Coming towards the power saving mode on the M1, the processor manages to score 1701 for single and 4896 for multi-threading. The ultra power saving mode sees scores dip down further to 1100 and 3209 for single and multi-threading respectively, both of which are effectively half of what the processor is able to achieve with standard performance mode. These results are quite impressive when compared to current chipsets, particularly Samsung’s own Exynos 7420 and Apple’s A9.

http://weibo.com/u/5673255066

 

[tipoo]

60 fps offscreen for 880 mali just seems unreal. Intel skylake 530 scores 70 fps and xbox one gpu get around 150 fps. Would be nice surprise if true though and if true no way intel ever getting into mobile.

The time for mobile to surpass this generation will be significantly shorter than it took to pass the 7th gen, for a few reasons, it seems.

 

[Sweepr]

Samsung is trying to fix Snapdragon 820 chipset

As we reported earlier, Samsung might release three processor variants of the Galaxy S7. After exclusively using the homegrown Exynos 7420 chipset in its high-end smartphones that were released in 2015, the South Korean smartphone giant might have made up its mind to bring back Snapdragon chipsets in its upcoming flagship smartphone for 2016.

According to a new report from Business Korea, Samsung is making every effort to fix heating issues in the Snapdragon 820 chipset and trying to optimize it. The company is trying to release a patch before the end of this month by modifying the microprocessor control program. If this step doesn’t solve the issue, the company is planning to use heat radiating pipes to improve heat dissipation, which means that Samsung is hell bent on using the Snapdragon 820 in the Galaxy S7 at all costs.

The company is trying to stabilize the chipset because the Snapdragon 820 is being manufactured using Samsung’s 14nm FinFET process. The better the performance of the Snapdragon 820, the higher the sales of the Galaxy S7 as well as Snapdragon 820 chips. This will in turn be advantageous to Samsung. The South Korean electronics giant is also working on its next-generation chipset, the Exynos 8890, which has custom 64-bit CPU cores.

After the poor performance of Qualcomm’s first high-end 64-bit mobile processor, the Snapdragon 810, the company has been under a lot of pressure from its investors. The mobile processor giant has decided to lay off 15 percent of its workforce in July. The company claims that the Snapdragon 820 offers upto 40 percent faster graphic performance while being 40 percent power efficient than the Snapdragon 810.

www.sammobile.com/2015/10/27/samsung-is-trying-to-fix-snapdragon-820-chipset
www.businesskorea.co.kr/english/new...ries-stabilize-qualcomms-snapdragon-820-chips

 

[ShintaiDK]

I think Samsung knows its not fun to fight with vanilla cores in a flagship product.

 

[vissarix]

Samsung is trying to fix Snapdragon 820 chipset



www.sammobile.com/2015/10/27/samsung-is-trying-to-fix-snapdragon-820-chipset
www.businesskorea.co.kr/english/new...ries-stabilize-qualcomms-snapdragon-820-chips

i dont know why samsung is trying to get at any cost an s820 on the upcoming s7?!

is qualcomm giving them for free?

the s810 was a complete failure and even watercooled it still performs worse then an exynos 7420 as showed by linustechtips.