Samsung’s 14nm in mass production?

[astilo]

Readers will decide if the 14nm Exynos indeed delivers very limited advantages.

Lol, first of all, in Antutu, the 14nm Exynos got around 69K. There are plenty of links out there, just have a look.
Can you show us a mobile SoC getting a better score? Tegra X1 has a 10W power consumption scenario. At best you can put it in a tablet, for sure not in a mobile phone.
The Snapdragon 810, to make a fair comparison between high end SoCs, built on TSMC 20nm, managed a quite good 55K. Can you feel the difference between 69K and 55K?
Well, it is just a disappointing 25%!!!

 

[user_2013101]

Yes the 4 A57 ARE as big as Apple's A8. No the A53 play no major role.

Again due to lack of understanding of benchmarks you resort to flawed conclusions, the memory scores come from ARM's memory architecture: ....

The dual cores of A8 are equivalent to Exynos’ four A57. However, the additional four A53 plays no role? This proud poster should tell Samsung to get rid of the unnecessary A53’s to save space and power.

Wherever Exynos scores poorly, it is due to our lack of understanding or someone else’s fault, such as ARM’s problematic memory architecture.

What about the Browsermark, the benchmark that better resembles real-world user experience? Why Exynos fails to show advantage on this benchmark?

Exynos’ 70K Antutu score is only from some media reports. The formal Antutu database has not yet updated. And, Samsung’s PR machine is notorious for media misinformation.

For example: since Samsung had started A9 production months ago, according to Samsung propaganda, when Samsung will ship A9s?

Conclusion: Samsung’s superiority cannot be questioned, or you lack understanding and are biased.

 

[astilo]

What about the Browsermark, the benchmark that better resembles real-world user experience? Why Exynos fails to show advantage on this benchmark?

Only the iphone6 is on par, but with a much smaller screen resolution: 2560x1600 vs 1334x750.

Exynos’ 70K Antutu score is only from some media reports. The formal Antutu database has not yet updated. And, Samsung’s PR machine is notorious for media misinformation.

Sure, just propaganda: [video=youtube;YlV0naJ3ZHo]https://www.youtube.com/watch?v=YlV0naJ3ZHo[/video]

For example: since Samsung had started A9 production months ago, according to Samsung propaganda, when Samsung will ship A9s?

Guess what, maybe with the next iphone?

 

[user_2013101]

The fair assessment of Galaxy S6 will arrive in due time. The ultimate arbiter is the market. We will find out if S6 can re-gain the market shares that had lost to iPhone 6.

In the past, Samsung had been caught inflating benchmark numbers. See a report below:

July 31, 2013
the international version of the Galaxy S4 (the one equipped with Samsung’s Exynos 5410 Octa) will boost the GPU clock to 532MHz, from 480MHz, if it detects that GLBenchmark 2.5.1, Antutu, or Quadrant is running.

Samsung caught artificially inflating Galaxy S4 benchmark results, fires back non-explanatory explanation | ExtremeTech

 

[astilo]

The fair assessment of Galaxy S6 will arrive in due time. The ultimate arbiter is the market. We will find out if S6 can re-gain the market shares that had lost to iPhone 6.

In the past, Samsung had been caught inflating benchmark numbers. See a report below:

Sad but true. It is an old story anyway. At that time, most of the top brands used to cheat, moreover in antutu.
It is also one of the reasons why I have always been a google nexus fan.
Most Android brands cheat in benchmarks - PhenomTech News and Tutorials

 

[user_2013101]

March 12, 2015

Daiwa Securities‘ Rick Hsu and Olivia Hsu came out this morning saying “the A9 contest is coming to an end” and TSMC has won. The analysts believes TSMC will get 70% of Apple’s business:

TSMC Will Eat Most Of The A9 Apple Pie: Daiwa - Asia Stocks to Watch - Barrons.com

Nothing new to me. I have been saying so since last November, despite the relentless Samsung media misinformation to the contrary.

 

[astilo]

Nothing new to me. I have been saying so since last November, despite the relentless Samsung media misinformation to the contrary.

And do you really believe that Apple had not yet decided? These chips are supposed to go inside the Iphone 6s next September.
Do you even have a rough idea about a 14nm full wafer cycle time? Guess? Days, weeks,..., nope. Several months.

 

[Daniel Nenni]

And do you really believe that Apple had not yet decided? These chips are supposed to go inside the Iphone 6s next September.
Do you even have a rough idea about a 14nm full wafer cycle time? Guess? Days, weeks,..., nope. Several months.

The fab capacity must also be built in advance so it is longer than the typical design cycle.

 

[Daniel Nenni]

Nothing new to me. I have been saying so since last November, despite the relentless Samsung media misinformation to the contrary.

Okay, I figured it out. Apple is currently designing to TSMC for the A10 which reporters have confused for the A9. Morris Chang said TSMC will lose 16nm business in 2015 but will make it up in 2016, remember? It's all about Apple... Just my opinion of course. But remember it takes 12 months or so to design an SoC and 6 months to get it into production. The A10 needs to tape-out in Q4 so it is well on its way.

 

[user nl]

Okay, I figured it out. Apple is currently designing to TSMC for the A10 which reporters have confused for the A9. Morris Chang said TSMC will lose 16nm business in 2015 but will make it up in 2016, remember? It's all about Apple... Just my opinion of course. But remember it takes 12 months or so to design an SoC and 6 months to get it into production. The A10 needs to tape-out in Q4 so it is well on its way.

I am no expert in this topic but just came upon this item on deepchip where one of the replies was:

A10 is being designed on Cadence EDI in TSMC 16FF.

A9 trying early yields now. Process issues. A10 might go Samsung 14
if TSMC can't fix this.

A10X is to be on Samsung 10 nm.


Engineering comments point to SNPS vs. CDNS PNR shakeout at Apple


user nl

 

[Daniel Nenni]

I am no expert in this topic but just came upon this item on deepchip where one of the replies was:

A10 is being designed on Cadence EDI in TSMC 16FF.

A9 trying early yields now. Process issues. A10 might go Samsung 14
if TSMC can't fix this.

A10X is to be on Samsung 10 nm.

user nl

Remember there are two very different chips, one for the iPhone and one for the iPad. At 20nm it was the A8 and A8x in 2014. So expect an A9 and A9x at 16/14nm 2015, A10 and A10x again at 16nm/14nm 2016. There is no way 10nm will be ready in time for HVP in 2H 2016, my opinion.

Wikipedia:
The A8 features an Apple-designed 1.4[SUP][66][/SUP] GHz 64-bit[SUP][67][/SUP] ARMv8-A[SUP][67][/SUP] dual-core CPU, and an integrated PowerVR GX6450 GPU in a four cluster configuration.[SUP][66][/SUP] The A8 is manufactured on a 20 nm process[SUP][68][/SUP] by TSMC,[SUP][69][/SUP] which replaced Samsung as the manufacturer of Apple's mobile device processors. It contains 2 billion transistors. Despite that being double the number of transistors compared to the A7, its physical size has been reduced by 13% to 89 mm[SUP]2[/SUP] (consistent with a shrink only, not known to be a new microarchitecture).[SUP][70][/SUP]

The Apple A8X is a system on a chip (SoC) designed by Apple, introduced at the launch of the iPad Air 2 on October 16, 2014.[SUP][71][/SUP] It is a high performance variant of the Apple A8. Apple states that it has 40% more CPU performance and 2.5 times the graphics performance of its predecessor, the Apple A7.[SUP][71][/SUP][SUP][72][/SUP] Unlike the A8, this SoC uses a triple-core CPU, a new octa core GPU, dual channel memory and slightly higher 1.5 GHz CPU clock rate.[SUP][73][/SUP] It uses an integrated octa-core PowerVR GXA6850 graphics processing unit (GPU) running at 450 MHz and a dual-channel memory subsystem.[SUP][73][/SUP] It's manufactured by TSMC on their 20 nm fabrication process, and consists of 3 billion transistors.

 

[user nl]

............................. There is no way 10nm will be ready in time for HVP in 2H 2016, my opinion.

Daniel, as said I cannot judge the credibility of these predictions/rumors, just my `2 cents contribution (?)` to your very nice semiwiki website/ecosphere that I joined recently.

One of those overviews from an analyst at KGI Research gives very specific info on nodes/manufacturers for the various A9/A10 and x-variant chips of Apple; 10 nm Samsung is mentioned for A10x:

Samsung is expected to split orders for the iPhone's next-generation A9 with Global Foundries, while TSMC would handle the A10 in 2016. The A9X — which would power both the iPad and a new low-end Mac — would be fabbed on TSMC's 16nm line, with the A10X moving to Samsung's 10nm plant.

Apple may shun Intel for custom A-series chips in new Macs within 1-2 years



This analyst at KGI Research predicted Apple to have sold 73 M iPhones in last quarter of 2014, and 3 days later the facts came out as 74.5 M in the official Apple press release. So maybe this analyst has some credibility and may be well connected in the Apple ecosphere? As said I have no expertise on this particular topic of Apple chips and nodes/manufacturers, just linking these reports as observed on various sites like deepchip (seems to have credibility amongst many (subscribed) chip engineers from the leading companies?) and appleinsider:

KGI's Ming-Chi Kuo: Apple shipped 73 million iPhones in Q4

Apple Reports Record First Quarter Results - Yahoo Finance


user nl

 

[Daniel Nenni]

Daniel, as said I cannot judge the credibility of these predictions/rumors, just my `2 cents contribution (?)` to your very nice semiwiki website/ecosphere that I joined recently.

One of those overviews from an analyst at KGI Research gives very specific info on nodes/manufacturers for the various A9/A10 and x-variant chips of Apple; 10 nm Samsung is mentioned for A10x:

Samsung is expected to split orders for the iPhone's next-generation A9 with Global Foundries, while TSMC would handle the A10 in 2016. The A9X — which would power both the iPad and a new low-end Mac — would be fabbed on TSMC's 16nm line, with the A10X moving to Samsung's 10nm plant.

Apple may shun Intel for custom A-series chips in new Macs within 1-2 years



This analyst at KGI Research predicted Apple to have sold 73 M iPhones in last quarter of 2014, and 3 days later the facts came out as 74.5 M in the official Apple press release. So maybe this analyst has some credibility and may be well connected in the Apple ecosphere? As said I have no expertise on this particular topic of Apple chips and nodes/manufacturers, just linking these reports as observed on various sites like deepchip (seems to have credibility amongst many (subscribed) chip engineers from the leading companies?) and appleinsider:

KGI's Ming-Chi Kuo: Apple shipped 73 million iPhones in Q4

Apple Reports Record First Quarter Results - Yahoo Finance


user nl

Great research, I appreciate the post.

This chart and I agree, it is what I have been saying with the exception of the A10x @ 10nm in 2016. That is really going to be a stretch but it could certainly happen. The 10nm PDKs are just now out (Revision 0.1) so anything is possible. I do see the wisdom in Apple splitting the chips by versions between foundries. It is much easier than splitting the same chip between two foundries. It also keeps two foundry sources enagaged so you can get competitive business terms. It is a shame the iPhones outsell the iPads 70/30 though. Hopefully Apple can put an Ax SOC in a MACBook soon to get the volumes up.

In regards to modems, it is my hope that the A10 will have an integrated modem.

 

[Daniel Nenni]

FYI: https://www.semiwiki.com/forum/content/4422-apple-leaks-chip-sources.html

 

[loekf]

The A9X — which would power both the iPad and a neouw low-end Mac — would be fabbed on TSMC's 16nm line, with the A10X moving to Samsung's 10nm plant.???

UH ???? A new low-end Mac with an ARM architecture ? Though iOS and OSX share many components, I don't see it happen. Architecture and hardware wise it's a different story. If you see the new Mac notebook with its shrinked PCB (thanks to Intel's Core M) it's almost a iPad with a keyboard and touchpad.

 

[user nl]

The A9X — which would power both the iPad and a neouw low-end Mac — would be fabbed on TSMC's 16nm line, with the A10X moving to Samsung's 10nm plant.???

UH ???? A new low-end Mac with an ARM architecture ? Though iOS and OSX share many components, I don't see it happen. Architecture and hardware wise it's a different story. If you see the new Mac notebook with its shrinked PCB (thanks to Intel's Core M) it's almost a iPad with a keyboard and touchpad.

Loekf, as said I'm no Apple A9/A10 nor OSX expert. I always stayed away from these relatively expensive `toys`, but instead used since 1992 free Linux on cost effective x86 `toys', starting with the first SLS distribution . I have academic working background so I always had to be prudent with our (finite) tax payers money.....

But in the Apple ecosphere there are quite some stories about Apple`s potential move from Intel to ARM,
and the merging of iOS and OSX, e.g. this Aug 2014 story quoting a former Apple executive:

Former Apple executive Jean-Louis Gassée — himself a longtime skeptic of Apple’s transition to ARM chips for its desktop and laptop computers — says he’s recently been convinced, and even believes that Apple could release ARM-based Macs as soon as 2016.
.........................
According to Gassée, he even thinks we could see a desktop-class A10 processor, which would put the eventual migration of the Mac from x86 to ARM to begin at 2016. Which means that in the next two years, iOS and OS X could eventually converge.

Read more e.g. here:

http://www.cultofmac.com/289773/ex-apple-exec-says-macs-run-arm-processors-2016/

When will Apple merge OS X and iOS? | TechnologyTell


User nl

 

[user_2013101]

And do you really believe that Apple had not yet decided? These chips are supposed to go inside the Iphone 6s next September.
Do you even have a rough idea about a 14nm full wafer cycle time? Guess? Days, weeks,..., nope. Several months.

My outlook that TSMC to build most, or all, of A9s has not changed. Apple probably had made the decision around November last year.

Details with supporting evidences were posted under another thread:

https://www.semiwiki.com/forum/f302/buying-opportunity-tsm-5690-post19718.html#post19718

 

[Andrei]

My outlook that TSMC to build most, or all, of A9s has not changed. Apple probably had made the decision around November last year.

And as proven in the last few weeks and discussions in this thread, you continue to be wrong.

 

[user_2013101]

Latest piece of evidence, from checks on equipment vendors, that TSMC has started the mass production of most of A9s, posted under another thread:

https://www.semiwiki.com/forum/f302/buying-opportunity-tsm-5690-post19731.html#post19731

 

[ta152h]

The A9X — which would power both the iPad and a neouw low-end Mac — would be fabbed on TSMC's 16nm line, with the A10X moving to Samsung's 10nm plant.???

UH ???? A new low-end Mac with an ARM architecture ? Though iOS and OSX share many components, I don't see it happen. Architecture and hardware wise it's a different story. If you see the new Mac notebook with its shrinked PCB (thanks to Intel's Core M) it's almost a iPad with a keyboard and touchpad.

You're confusing hardware and software. A Mac is just a PC in drag, if you look at it as hardware. But, it's not how it's perceived in the market. Also, the software experience is radically different.

A MacIntrash on ARM does not mean a Mac on iOS, or a merged OS. There would be some components that would be similar, or shared, like the APIs and such, but the inner workings would likely remain radically different. Think in terms of Windows 95 and Windows NT, but perhaps moreso because the form factor is tangibly different, whereas Win 95/98/ME was the same, and just for better compatibility with older applications that behaved in a manner NT couldn't allow.

Apple has a history of jumping instruction sets, from 68K to POWER, to x86. ARM is inevitable, as it makes software development less expensive, allows better cross platform compatibility without awkward emulators, and gives Apple much more control over the processor than they have with Intel.

I wouldn't be shocked if Apple chose AMD's K12 as their first ARM Mac processor, as it's not clear to me the Mac line is currently big enough to warrant the development of a high performance processor by Apple. By using K12, they could put that in the higher-end Macs, while using the A9/A10 for the less powerful devices. It would not cause the market confusion of having x86 for high performance Macs, and ARM for lower performance models, which would be a logistical nightmare as well.

Mac using ARM is inevitable, it's only the timing that is in question. It's also highly uncertain they'll choose K12, as that's pure speculation. It is plausible though. But, that assumes Apple doesn't feel Mac volume warrants a new processor (somewhat likely), A10 does not have the performance (very likely), and the K12 is a very good processor (well, this one is iffy, based on AMD's recent history with big processors. But, it's possible, with it being ARM, and Keller giving it a shot).