Samsung Announces Mass Production of 14nm LPP

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
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.
Which only 100% confirms, that there was mass production for 14LPE products at that time and that 14LPP has at least finished the qualification with provided data. Due to the overlaying of statements and the known timing of the processes, it doesn't contain a confirmation that there was 14LPP mass production going on.
 
Mar 10, 2006
11,715
2,012
126
If we are to believe that 14LPP is just now going into mass production, then this means TSMC beat Sammy to the punch in terms of 16FF+ mass production.
 

NTMBK

Lifer
Nov 14, 2011
10,208
4,940
136
So we haven't actually seen an LPP part yet, and the Apple A9 was on LPE. Meaning that we have absolutely no real world data on how 16FF+ and LPP compare.

So can we all please shut up about inferior and superior processes? We can't possibly know at this point, so just give it a rest. Jesus Christ.
 
Mar 10, 2006
11,715
2,012
126
So we haven't actually seen an LPP part yet, and the Apple A9 was on LPE. Meaning that we have absolutely no real world data on how 16FF+ and LPP compare.

So can we all please shut up about inferior and superior processes? We can't possibly know at this point, so just give it a rest. Jesus Christ.

My sources still stand by the fact that A9 was LPP even in light of this PR, FWIW.

Interestingly what I'm hearing is that A9 was supposed to be LPE, but in order to hit desired performance levels, Apple needed to move to LPP.

Anyway, if Samsung is going to launch the S7 in February, this means that 14LPP needed to have been in production in late 2015. Cycle time for 14/16nm is on the order of 1.5 quarters (wafer in to packaged and tested chip out) so if they are just starting production now, a February S7 (in any real quantity) launch would be nigh impossible.

One more thing to keep in mind...Samsung put out a PR announcing 14LPE mass production in like February 2015, but they were talking about how mass production began in October 2014 (or was it December?) on investor calls/roadshows.

So I wouldn't necessarily rely on the timing of these PRs for a really accurate sense of when production on a given process -- especially for a customer like Apple that's very secretive -- began.
 
Last edited:
Mar 10, 2006
11,715
2,012
126
Your sources a worthless lairs then. Ever heard of Intel Sophia?

No, they're not. And while technically true that Intel builds chips at TSMC, they don't build their major product lines there (i.e. anything that would compete with AMD).
 

Andrei.

Senior member
Jan 26, 2015
316
386
136
Which only 100% confirms, that there was mass production for 14LPE products at that time and that 14LPP has at least finished the qualification with provided data. Due to the overlaying of statements and the known timing of the processes, it doesn't contain a confirmation that there was 14LPP mass production going on.
^ Correct interpretation of that statement.

Official PR statements always happen later than the actual events. Samsung started volume 7420 production December 2014 but only announced it mid-February. The 8890 was announced in November and I'm sure it was in production since then and only now did foundry PR announce it.

If some people are left to believe they had shipping LPP silicon in the A9 end of September churning out massive volumes in the months before then I admire their determination and optimism in the face of opposing evidence.

As for 14LPP vs 16FF+ performance, 16FF+ should still slightly edge it out on average but the differences are so small it'll probably be out-shadowed by binning variation.
 
Last edited:

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
This confirms what we all heard from Samsung that 14LPE was a time to market process for 2015 and worked for Samsung to get their Exynos 7420 and roughly 2/3rd of the A9 chips. I still think that was a smart strategy given the fact that they were able to take business away from TSMC. 14LPP is the process of choice for 2016. Now with 14LPP powering Snapdragon 820 and Samsung Exynos 8890 it will be driving the bulk of the premium Android smartphones in 2016. If we add the fact that GF will be fabbing multiple AMD products at 14LPP which could include AMD Zen CPUs (not confirmed) and GPUs (confirmed) in 2016 and most likely shrunk PS4 and Xbox One APU and Zen APUs in 2017 it looks like TSMC is losing a big chunk of business at 16FF+. Now all we need is products fabbed at TSMC 16FF+ and 14LPP to hit the market and some good analysis from chipworks and reviews from tech press. I am looking forward to Apple A10 / A10X facing off against Exynos 8890 and S820.
 
Last edited:

NTMBK

Lifer
Nov 14, 2011
10,208
4,940
136
My sources still stand by the fact that A9 was LPP even in light of this PR, FWIW.

Interestingly what I'm hearing is that A9 was supposed to be LPE, but in order to hit desired performance levels, Apple needed to move to LPP.

Anyway, if Samsung is going to launch the S7 in February, this means that 14LPP needed to have been in production in late 2015. Cycle time for 14/16nm is on the order of 1.5 quarters (wafer in to packaged and tested chip out) so if they are just starting production now, a February S7 (in any real quantity) launch would be nigh impossible.

One more thing to keep in mind...Samsung put out a PR announcing 14LPE mass production in like February 2015, but they were talking about how mass production began in October 2014 (or was it December?) on investor calls/roadshows.

So I wouldn't necessarily rely on the timing of these PRs for a really accurate sense of when production on a given process -- especially for a customer like Apple that's very secretive -- began.

Very interesting, thanks :) In that case I withdraw my nerd-rage comment for the time being.

(Seriously though, this argument is getting old already... and 14/16nm GPUs aren't even out for another 6 months :( )
 

VR Enthusiast

Member
Jul 5, 2015
133
1
0
(Seriously though, this argument is getting old already... and 14/16nm GPUs aren't even out for another 6 months :( )

Only if you are taking the average with AMD's being ready in 3 months while Nvidia's are ready in 9.

Trolling is not allowed here
Markfw900
 
Last edited by a moderator:

Exophase

Diamond Member
Apr 19, 2012
4,439
9
81
As for 14LPP vs 16FF+ performance, 16FF+ should still slightly edge it out on average but the differences are so small it'll probably be out-shadowed by binning variation.

If that's true that should mean 16FF+ has a very significant performance advantage over 14LPE. If A9 is 14LPE is that really what we're seeing? If anything this claim supports A9 being on 14LPP more than 14LPE.
 

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
Would a company probably wait for the first wafers to be through the process before making PR statements which might imply soon availability?
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
I noticed Samsung states upto 15% higher performance and 15% percent lower consumption for 14LPP vs 14 LPE. If we combine this with Samsung Octa 8 Exynos 8890 and Snapdragon 820 statements it looks like Samsung 14LPP is very competitive with TSMC 16FF+ with roughly similar transistor performance and power and 10% lower die size.

http://www.samsung.com/semiconductor/minisite/Exynos/w/solution.html#?v=modap_8octa

Exynos 8890 is expected to have 30% more performance over Exynos 7420 and 10% better power efficiency.

http://semiaccurate.com/2015/11/14/qualcomm-shows-us-the-snapdragon-820/

"Weighting on the foundry debate Qualcomm said that it does not believe that there is a significant performance difference between TSMC and Samsung. They chose Samsung’s LPP 16nm FinFET process over TSMC’s 16nm FinFET alternative for cost and volume reasons. Qualcomm also hinted at the potential for some of its chips to be built at Global Foundries in the future, but said nothing had been publicly announced."

The key here is yields and TSMC seems to have the advantage there from Apple's choice of TSMC 16FF+ for A9X which is 147 sq mm and the largest high volume mobile SoC we have seen from any foundry.
 
Last edited:

VR Enthusiast

Member
Jul 5, 2015
133
1
0
Ah fuck this forum, what a joke.


Bad language is not allowed here,
And insulting the entire forum is rude and trolling. If you don't like it here, go somewhere else.
Markfw900
 
Last edited by a moderator:

Andrei.

Senior member
Jan 26, 2015
316
386
136
If anything this claim supports A9 being on 14LPP more than 14LPE.
Why's that? LPE vs FF+ explains the power difference at maximum frequency.

I noticed Samsung states upto 15% higher performance and 15% percent lower consumption for 14LPP vs 14 LPE. If we combine this with Samsung Octa 8 Exynos 8890 and Snapdragon 820 statements it looks like Samsung 14LPP is very competitive with TSMC 16FF+ with roughly similar transistor performance and power and 10% lower die size.

http://www.samsung.com/semiconductor/minisite/Exynos/w/solution.html#?v=modap_8octa

Exynos 8890 is expected to have 30% more performance over Exynos 7420 and 10% better power efficiency.
Don't put any weight onto the 8890 numbers, SLSI has a history of misleading SoC figures (both in terms of overselling (5433) but also vastly underselling (7420) them). That 10% efficiency figure for example is certainly a PR miscommunication as their financial call presentation actually said it was 10% lower power (which together with the perf boost is 44% more efficiency).

The key here is yields and TSMC seems to have the advantage there from Apple's choice of TSMC 16FF+ for A9X which is 147 sq mm and the largest high volume mobile SoC we have seen from any foundry.
Is A9X really that high volume?