Samsung Exynos Thread (big.LITTLE Octa-core)

[Sweepr]

S810 geekbenchscores here
http://m.gsmarena.com/ces_2015_lg-review-1187p3.php
2ghz a57

Nice find. For some reason the score is a bit lower than Exynos 5433, here's my latest results (running at stock 1.9/1.3GHz, factory default settings @ Android 4.4.4 32-bit):

Anyway, let's wait for more results from different devices, that's just the first of many S810-based flagships launching this year. I wonder if there will be some new leaks about Exynos 7420 and Sammy's future SoCs @ CES 2015.

Edit: Here's a list of Geekbench 3 scores from different ARM/x86 chips for those interested in comparisons.

 

[Nothingness]

I posted Galaxy Note 4 S810 Geekbench score weeks ago

 

[krumme]

I remember . This LG is in the hands of gsmarena.com. I asume its comming to market sooner than the note 4 but we will see.

Why do the scores wary so much for the devices? The lg is running 5.01 btw

 

[Sweepr]

I remember . This LG is in the hands of gsmarena.com. I asume its comming to market sooner than the note 4 but we will see.

Why do the scores wary so much for the devices? The lg is running 5.01 btw

Trouble for Qualcomm?

The Snapdragon 810, however, faces concerns about technological issues.

Problems such as overheating at certain voltages and performance degradation caused by memory controller problems have been reported, and its clock rate, an index representing a processor's performance, was estimated to be lower than its predecessor, the Snapdragon 805.

Back to the issues, sources from Korea and analysts with US investment firm J.P. Morgan are convinced that the Snapdragon 810 is suffering from crippling overheating issues. Apparently, this problem is caused by the high-performance Cortex-A57 cores overheating when clock speeds reach 1.2 to 1.4GHz, which is a surprising problem for a core designed to run at speeds approaching 2GHz. This then causes the chip to throttle back on performance, to prevent the whole system from overheating.

However, Samsung’s Cortex-A57 powered Exynos 5433 does not suffer from overheating issues, suggesting that this is a problem specific to Qualcomm’s Snapdragon design rather than a problem with the Cortex-A57 itself. This leaves the finger pointed at Qualcomm and TSMC’s 20nm chip design, with several analysts suggesting that a “redesign of a few metal layers” may be needed to fix the issue.

All Geekbench and AnTuTu scores I've seen (both LG Flex 2 and the rumoured S810 Note 4) are way too low for a 2GHz A57 chip (64-bit enabled). Any delay here would most certainly affect the launch schedule of the next Android flagships.

Samsung Electronics plans to release the Exynos 7 Octa series this year. Its flagship smartphone Galaxy Note 4 S LTE-A already uses the octa-core Exynos 7 chipset, and the company will expand use of the product manufactured in the 14-nanometer process for its handsets in the future.

Nvidia also unveiled its new 64-bit mobile chip Tegra X1 and promoted its top-tier graphic processing capability.

14nm Exynos 7 Octa coming soon, as expected.

www.koreatimes.co.kr/www/news/tech/2015/01/133_171341.html
www.androidauthority.com/Snapdragon-810-overheating-issues-579284

 

[jhu]

No Blender benchmarks?

 

[monstercameron]

No Blender benchmarks?

lol, you and blender

 

[jhu]

lol, you and blender

hehe, rendered this on my phone just because I could:

 

[krumme]

Trouble for Qualcomm?







All Geekbench and AnTuTu scores I've seen (both LG Flex 2 and the rumoured S810 Note 4) are way too low for a 2GHz A57 chip (64-bit enabled). Any delay here would most certainly affect the launch schedule of the next Android flagships.



14nm Exynos 7 Octa coming soon, as expected.

www.koreatimes.co.kr/www/news/tech/2015/01/133_171341.html
www.androidauthority.com/Snapdragon-810-overheating-issues-579284

Lets give qcom a week then we know. And a 810 have tons of qualities besides the cpu. Besides they have denied it december but it keeps popping up for some reason.
As it is ss is bragging about their processes and looks to actually have been first here by a long stretch now. A57 20nm perhaps in small numbers but i agree 16nm finfet a57 is probably very close now. Qcom have already missed half of the party and messed this up.
Cant really see the huge benefit for ss though for beeing first but for each day...damn.

 

[Sweepr]

Don't want to start a new thread so I'll post here:

Samsung drops Snapdragon S810 from Galaxy S6, will use its own Exynos 7 Octa chips

Samsung Electronics Co. (005930) will use its own microprocessors in the next version of the Galaxy S smartphone, dropping its use of a Qualcomm Inc. (QCOM) chip that overheated during the Korean company’s testing, people with direct knowledge of the matter said.

Samsung, the world’s largest smartphone maker, tested a new version of Qualcomm’s Snapdragon chip, known as the 810, and decided not to use it, the people said, asking not to be identified because the issue hasn’t been discussed publicly. Qualcomm shares fell.

The decision is a blow to Qualcomm, the world’s largest maker of semiconductors used in phones, which has been supplying Samsung with chips that run the company’s best-selling handsets. The South Korean company, the world’s second-largest chipmaker, is trying to become more self-reliant and boost its own processor-making division as it spends $15 billion on a new factory outside Seoul.

“Samsung may release the next Galaxy S as early as March, and it can’t dare to take the risk to use any of the chips in question for its most important model,” said Song Myung Sup, a Seoul-based analyst at HI Investment & Securities Co.

www.bloomberg.com/news/2015-01-20/s...ualcomm-chip-from-next-galaxy-smartphone.html

That brings up interesting questions. What kind of problems is Qualcomm facing with S810? Is Samsung ready to launch millions of Galaxy S6 phones using a brand new 14nm SoC by March (apparently yes)? Bloomberg also mentions that they want to boost their CPU division, which makes sense, rumour says they are developing custom CPU/GPU cores for quite some time (could be ready by Galaxy Note 5's launch).

About a week ago DigiTimes mentioned 80-90% of Galaxy S6 phones would be based on Exynos, now Samsung is expected to ditch Qualcomm's S810 completelly.

Samsung will initially have 80-90% of processors used in its Galaxy S6 manufactured using its in-house developed 14nm process technology, according to the sources..

http://www.digitimes.com/news/a20150114PD213.html

 

[witeken]

Don't you mean

Apparently no?

 

[III-V]

That brings up interesting questions. What kind of problems is Qualcomm facing with S810? Is Samsung ready to launch millions of Galaxy S6 phones using a brand new 14nm SoC by March (apparently yes)? Bloomberg also mentions that they want to boost their CPU division, which makes sense, rumour says they are developing custom CPU/GPU cores for quite some time (could be ready by Galaxy Note 5's launch).

They're having overheating issues.

I've long desired Samsung to replace Qualcomm completely in their phones, but only if they went custom. It seems like they're finally going to do this, as was leaked sometime back... was it a year ago? Two? Can't remember.

Anyway, Samsung, like Intel, has a foundry, and I think they could really leverage that as a strength. They're a huge company too, and if they'd just wake up, they'd realize they have the resources to beat Qualcomm at their own game, especially after Qualcomm's dropped the ball on the CPU portion of their SoCs time and time again.

Don't you mean

Apparently no?

No, as in having 14nm phones on the shelf in March? You're almost certainly correct. They'd surely be 20nm-based.

 

[Sweepr]

Don't you mean

Apparently no?

No, as in having 14nm phones on the shelf in March? You're almost certainly correct. They'd surely be 20nm-based.

That would contradict all leaks and rumours about Exynos 7420 to date. Also:

We have started mass production wafers in 14-nanometer. So, in 14-nanometer, our progress is well on track, including process and the yields. Here we're talking when I mention yields being on track, this is for a very advanced product. It'll be our most advanced applications processor. Yielding at full specifications targets with all of the characteristics for speed and power as well.

Samsung said that 2 months ago and I don't think they were talking about the Q4-2015 Galaxy Note 5.

 

[ancientarcher]

That would contradict all leaks and rumours about Exynos 7420 to date.

I guess he is joking. 14nm = 20nm FinFet :\
Sure, it is 20nm FinFet but performance characteristics are 30-50% better than 20nm planar, so that will make a huge difference! The improvement in transistor performance in moving to FinFet from Planar is equivalent to a node move. So, why not a different name!!

 

[III-V]

I guess he is joking. 14nm = 20nm FinFet :\
Sure, it is 20nm FinFet but performance characteristics are 30-50% better than 20nm planar, so that will make a huge difference! The improvement in transistor performance in moving to FinFet from Planar is equivalent to a node move. So, why not a different name!!

I wasn't joking. I remember reading that they'd gone into mass production, but 14nm seems a bit soon for them. I guess a March paper launch with April or May availability isn't too crazy.

 

[witeken]

I guess he is joking. 14nm = 20nm FinFet :\
Sure, it is 20nm FinFet but performance characteristics are 30-50% better than 20nm planar, so that will make a huge difference! The improvement in transistor performance in moving to FinFet from Planar is equivalent to a node move. So, why not a different name!!

Because density is for all intents and purposes the same.

28HPM also had the efficiency increase of 1 node compared to 28LP, so should they have called that one 20nm? So then 20nm would have been called 16nm and their 20+FinFET 10nm and they could have said they're 1 node ahead of Intel!

 

[ancientarcher]

Because density is for all intents and purposes the same.

28HPM also had the efficiency increase of 1 node compared to 28LP, so should they have called that one 20nm? So then 20nm would have been called 16nm and their 20+FinFET 10nm and they could have said they're 1 node ahead of Intel!

And what part of Intel's 14nm transistors measures exactly 14nm can you please elaborate? i.e. what is the exact transistor dimension referred to when Intel decided to call the node 14nm?


With the 1281 single core Geekbench 3 score posted for 20nm Exynos 5433, I am waiting to see what the upgrade to FinFet will bring!! And we are just talking about vanilla cortex A57 here

 

[witeken]

And what part of Intel's 14nm transistors measures exactly 14nm can you please elaborate? i.e. what is the exact transistor dimension referred to when Intel decided to call the node 14nm?

Doesn't matter. The name is 0.7x the previous one because that's how they scaled the transistor. 16nm isn't.

Even that is irrelevant. The name is misleading for most tech folks because they think the nodes are the same because the names are the same.

 

[Arachnotronic]

Doesn't matter. The name is 0.7x the previous one because that's how they scaled the transistor. 16nm isn't.

Even that is irrelevant. The name is misleading for most tech folks because they think the nodes are the same because the names are the same.

Come now, Intel is guilty of this same practice

 

[dawheat]

They're having overheating issues.

I've long desired Samsung to replace Qualcomm completely in their phones, but only if they went custom. It seems like they're finally going to do this, as was leaked sometime back... was it a year ago? Two? Can't remember.

Anyway, Samsung, like Intel, has a foundry, and I think they could really leverage that as a strength. They're a huge company too, and if they'd just wake up, they'd realize they have the resources to beat Qualcomm at their own game, especially after Qualcomm's dropped the ball on the CPU portion of their SoCs time and time again.

No, as in having 14nm phones on the shelf in March? You're almost certainly correct. They'd surely be 20nm-based.

Going by the performance of the Exynos SOC in the Note 4,Samsung should be totally fine with the S6, especially with how underwhelming initial S810 benches seem to be.

20nm or 14nm, with integrated modem it should be very competitive any other phone SOC in 1H 2015.

 

[witeken]

Come now, Intel is guilty of this same practice

Actually, you're right. A lot of people really think there's significance in the number 14, 20, 22, 28, etc. But there isn't. But for comparative purposes, TSMC's, Global Foundries' and Samsung's moves are misleading.

 

[III-V]

And what part of Intel's 14nm transistors measures exactly 14nm can you please elaborate? i.e. what is the exact transistor dimension referred to when Intel decided to call the node 14nm?

Wanna let us know what part of TSMC, GloFo, former IBM, UMC, Samsung, and everyone else's process nodes actually measure exactly their named nanometers?

Going by the performance of the Exynos SOC in the Note 4,Samsung should be totally fine with the S6, especially with how underwhelming initial S810 benches seem to be.

20nm or 14nm, with integrated modem it should be very competitive any other phone SOC in 1H 2015.

I'm very excited to see FinFETs showing up in the market that needs them most. If Samsung does indeed use 14nm with their upcoming S6, I would be ecstatic.

 

[krumme]

That would indeed be fantastic if true and BIG news. I want my note 5 with 14nm finfet now
One thing is 810 overheating issues but perhaps the real reason is Samsung have fixed their 14nm yield issues.

 

[Idontcare]

That would indeed be fantastic if true and BIG news. I want my note 5 with 14nm finfet now
One thing is 810 overheating issues but perhaps the real reason is Samsung have fixed their 14nm yield issues.

Even if they haven't fixed their yield issues, it is far easier for them to run more wafers as they aren't capacity limited nor are they operating at full utilization. They need to run the wafers anyways, just to have product flowing for the process engineers to do tests with in order to improve yields. Kinda a win-win, just not the cheapest kind.

If this is true then it will certainly give Samsung and Apple a leg up over the competition that depends on Qualcomm.

 

[III-V]

C. Demerjian over at Semi-Accurate claims a "major SoC" is in mass production on a 16/14nm FF process, and that the foundry in question has fixed their yield issues... lines up pretty well with the Exynos rumor.

 

[jpiniero]

C. Demerjian over at Semi-Accurate claims a "major SoC" is in mass production on a 16/14nm FF process, and that the foundry in question has fixed their yield issues... lines up pretty well with the Exynos rumor.

It could be Apple. There is talk of Apple doing two iPhone releases this year; much like iPad 3 and 4.