iPhone 6s Chipgate: TSMC vs. Samsung

Gweenz

[H]ard|Gawd
Joined
Dec 18, 2003
Messages
1,216
No more -gate references to scandals please. It's not clever any more, never really was.
 

MoFoQ

Gawd
Joined
Sep 18, 2002
Messages
849
Thing is...smaller the nodes go, the hotter chips can get so heat dissipation/thermal management must also be adapted to the smaller node size.

It would be interesting to know if the Apple engineers designed the thermal management around the larger node or the smaller.
One way to test this would be to slap on your own heatsink to a "franken-6S" and see if it will perform better than stock (battery life, etc.)
 

Armisael

Weaksauce
Joined
Apr 19, 2013
Messages
118
As a general rule when you implement the same design on a smaller node you produce less heat. Working on a smaller node allows you to make design decisions that lead to a hotter chip, but the node itself doesn't inherently do that.

The two chips won't be otherwise identical because the two processes almost certainly have different design rules, but if anything I'd bet on the samsung chips being cooler.
 

Factum

2[H]4U
Joined
Dec 24, 2014
Messages
2,466
16nm?
14nm?

PR FUD

Cell-SizeComparison.png


It's shame only Intel is sticking to their normal naming convention...and everybody else is renaming their process...even if gate and picth dosn't alter...sad, sad state of affairs :(
 

rat

Supreme [H]ardness
Joined
Apr 16, 2008
Messages
4,915
Thing is...smaller the nodes go, the hotter chips can get so heat dissipation/thermal management must also be adapted to the smaller node size.

Actually, it's the other way around. Die shrinkage means less heat for the same design.
 

cbutters

Gawd
Joined
Dec 30, 2005
Messages
514
Thing is...smaller the nodes go, the hotter chips can get so heat dissipation/thermal management must also be adapted to the smaller node size.

Actually, it's the other way around. Die shrinkage means less heat for the same design.

@techrat... yes die shrinkage means less heat for the same design, however MoFoQ said "hot" which is different than heat.

While you are right that smaller nanometer processes on the same design produces less heat, it also means you have a design doing the same amount of work in a much denser area. This can lead to a chip running "hotter" in terms of temperature, but produces less overall heat. (Like we have seen with haswell vs. sandybridge for example)
 

Armisael

Weaksauce
Joined
Apr 19, 2013
Messages
118
That's a bad comparison - intel started using crappy TIM with ivy bridge. People who've delidded ivy/haswell procs and replaced the TIM with solder have seen huge temperature drops.
 

extide

2[H]4U
Joined
Dec 19, 2008
Messages
3,494
16nm?
14nm?

PR FUD

Cell-SizeComparison.png


It's shame only Intel is sticking to their normal naming convention...and everybody else is renaming their process...even if gate and picth dosn't alter...sad, sad state of affairs :(

Wow thats a really cool image
 

rat

Supreme [H]ardness
Joined
Apr 16, 2008
Messages
4,915
@techrat... yes die shrinkage means less heat for the same design, however MoFoQ said "hot" which is different than heat.

While you are right that smaller nanometer processes on the same design produces less heat, it also means you have a design doing the same amount of work in a much denser area. This can lead to a chip running "hotter" in terms of temperature, but produces less overall heat. (Like we have seen with haswell vs. sandybridge for example)

Not exactly comparable.

Haswell was a 'tock', an architectural redesign. The rules with Sandy and Ivy don't apply here as it's not strictly a die shrink.

Both TSMC and Samsung are fabricating the A9 at two different process node sizes. There (afaik) should be no difference in the actual mask between the two.
 

pxc

Extremely [H]
Joined
Oct 22, 2000
Messages
33,064
16nm?
14nm?

PR FUD

http://www.extremetech.com/wp-content/uploads/2014/12/Cell-SizeComparison.png[/I MG]

It's shame only Intel is sticking to their normal naming convention...and everybody else is renaming their process...even if gate and picth dosn't alter...sad, sad state of affairs :([/QUOTE]Both TSMC and Samsung had explained their hybrid node process a couple of years ago. That is what was expected.

Even I was snarking on it 2 years ago based on its characteristics: [url]http://hardforum.com/showpost.php?p=1040313878&postcount=38[/url]

[quote="pxc, post: 1040313878"]... The Alliance's upcoming 14/16nm node... is a hybrid node with performance characteristics and density not improved much from the 20nm node.[/QUOTE]
 

evilsofa

[H]F Junkie
Joined
Jan 1, 2007
Messages
10,078
As a general rule when you implement the same design on a smaller node you produce less heat. Working on a smaller node allows you to make design decisions that lead to a hotter chip, but the node itself doesn't inherently do that.

The two chips won't be otherwise identical because the two processes almost certainly have different design rules, but if anything I'd bet on the samsung chips being cooler.

Yet it's the smaller Samsungs that are running hotter and using up more battery life.

There's speculation that the TSMCs might be throttling and doing less work than the Samsung, but I have yet to see anyone talk about running a test that shows which version of the phone has faster performance, unless I have missed something obvious.
 

shansoft

Supreme [H]ardness
Joined
Oct 20, 2008
Messages
5,076
Yet it's the smaller Samsungs that are running hotter and using up more battery life.

There's speculation that the TSMCs might be throttling and doing less work than the Samsung, but I have yet to see anyone talk about running a test that shows which version of the phone has faster performance, unless I have missed something obvious.

Actually, it's the Samsung one that is throttling to keep the temperature in control.

There has been a lot of post regard to this issues and show that Samsung one consistently getting lower score than TSMC on benchmark.

My first 6S had Samsung chip, and it was noticeably hotter and uncomfortable to use.
 

DrNick

Limp Gawd
Joined
Jul 19, 2006
Messages
276
People always find something to flip over about. We have a 6s (sammy) and a 6s plus (tsmc) in use and I have yet to notice any real difference between the two even if they really cant be directly compared due to obvious differences. Even that geekbench app shows them pretty much identical in terms of performance no matter how many times i run it
 

pxc

Extremely [H]
Joined
Oct 22, 2000
Messages
33,064
You're quoting FUD that has been disproved. Look to an expert like Chipworks for facts.

https://www.chipworks.com/about-chipworks/overview/blog/a9-is-tsmc-16nm-finfet-and-samsung-fabbed

http://www.chipworks.com/sites/default/files/blog%20image6.jpg[/I MG][/QUOTE]You're comparing real apples and banana shaped rocks. The other image shows the dimensions of M1 and gate pitch as described by each manufacturer. As I quoted myself above, the 14nm and 16nm modes from Alliance members are hybrids and didn't get full geometry benefits expected in a node shrink.

Showing that chipworks die size comparison actually supports the other image. lol
 

Freebo

Limp Gawd
Joined
Oct 26, 2012
Messages
252
I read a lot regarding this, some people are saying its the other way around that its the TSMC chip thats causing it to over heat and shutdown. Some are saying apple did this to make TSMC look better to cut out samsung.... who knows at this point...

All I know is the phone is overheating and its shutting down... every release they have some major issue that people just ignore and keep on buying them and calling them awesome..
 

pxc

Extremely [H]
Joined
Oct 22, 2000
Messages
33,064
People always find something to flip over about. We have a 6s (sammy) and a 6s plus (tsmc) in use and I have yet to notice any real difference between the two even if they really cant be directly compared due to obvious differences. Even that geekbench app shows them pretty much identical in terms of performance no matter how many times i run it
The problem is differences in battery life.

Apple Confirms 'Good And Bad' iPhone 6S Chips Offer You Shorter Battery Life

article said:
Speaking to TechCrunch, Apple has pointed out that real-world usage is far more reflective of how an iPhone is used that the benchmarking tests, and that in normal use the variance between the two handsets is on the order or two or three percent. Given tolerances in manufacturing, this performance is within the margin of error that many would expect in consumer electronics.

A common use where this power drain difference would show outside of benchmarking is in mobile gaming, especially if it's more than minimally taxing.
 

mi7chy

2[H]4U
Joined
May 22, 2013
Messages
3,985
You're comparing real apples and banana shaped rocks. The other image shows the dimensions of M1 and gate pitch as described by each manufacturer. As I quoted myself above, the 14nm and 16nm modes from Alliance members are hybrids and didn't get full geometry benefits expected in a node shrink.

Showing that chipworks die size comparison actually supports the other image. lol

What part don't you understand that the true die size as shown by Chipworks shows a narrower and shorter die size invalidating the first picture which purports to show the same vertical dimension for TSMC 20nm/16nm and Samsung 14nm?
 

mi7chy

2[H]4U
Joined
May 22, 2013
Messages
3,985
Even Applenandtech was propagating this FUD but corrected itself:

http://www.anandtech.com/show/9330/exynos-7420-deep-dive
A great deal of discussion ensued over whether Samsung’s 14nm process really represented a “true” die shrink over its 20nm predecessor. We were ourselves surprised to see Chipworks announce that the piece came in at only 78mm² compared to the Exynos 5433’s 113mm². This 31% shrink was beyond what we expected, as we previously reported that Samsung’s 14nm process was to continue to use the 20nm’s BEOL (Back-End-Of-Line, a chip’s largest metal layer) and thus make for only a minor progression.
 

DrNick

Limp Gawd
Joined
Jul 19, 2006
Messages
276

mi7chy

2[H]4U
Joined
May 22, 2013
Messages
3,985
Probably right but 2 extra hours seems insane. But you'd need two identical phones to compare, I cant test or judge that on these

2 hours delta is very fishy. I know there are a lot of variables due to second sourcing components, batteries, etc. that may make a small single digit % difference but 2 hours I blame on nationalistic Chinese media supporting Chinese TSMC fab.
 
D

Deleted whining member 223597

Guest
Samsung produces worse products than some "garbage Chinese manufacturer"? Impossibru!!!!
 

shansoft

Supreme [H]ardness
Joined
Oct 20, 2008
Messages
5,076
2 hours delta is very fishy. I know there are a lot of variables due to second sourcing components, batteries, etc. that may make a small single digit % difference but 2 hours I blame on nationalistic Chinese media supporting Chinese TSMC fab.

Correction: TSMC is Taiwanese, not Chinese.

Full name - Taiwan Semiconductor Manufacturing Company

;)

Samsung produces worse products than some "garbage Chinese manufacturer"? Impossibru!!!!

LOL, he just gotta jump in every thread to trash talk anything related to Samsung or Apple....
 

mi7chy

2[H]4U
Joined
May 22, 2013
Messages
3,985
That's about as ignorant as saying Texas is not USA. What do you think the C in Taiwan ROC/PRC stands for?
 

jimmyb

2[H]4U
Joined
May 24, 2006
Messages
3,165
What part don't you understand that the true die size as shown by Chipworks shows a narrower and shorter die size invalidating the first picture which purports to show the same vertical dimension for TSMC 20nm/16nm and Samsung 14nm?

Each chip is going to have a totally different physical design, so it's not a particularly effective way to compare process area/density. Same goes for power consumption.
 

pxc

Extremely [H]
Joined
Oct 22, 2000
Messages
33,064
What part don't you understand that the true die size as shown by Chipworks shows a narrower and shorter die size invalidating the first picture which purports to show the same vertical dimension for TSMC 20nm/16nm and Samsung 14nm?
You don't understand the picture or what I wrote AT ALL. lol

Seriously, lol.
 

jimmyb

2[H]4U
Joined
May 24, 2006
Messages
3,165
Each chip is going to have a totally different physical design, so it's not a particularly effective way to compare process area/density. Same goes for power consumption.

I should add that they'll certainly also have differences in their schematic and RTL implementations, beyond just physical implementation.
 

Factum

2[H]4U
Joined
Dec 24, 2014
Messages
2,466
Both TSMC and Samsung had explained their hybrid node process a couple of years ago. That is what was expected.

Even I was snarking on it 2 years ago based on its characteristics: http://hardforum.com/showpost.php?p=1040313878&postcount=38

It's sad...not only are they being dishonest...they also fool not so knowledgeable people (just look in this thread).

It's really sad...they use PR instead of engineering.
They might as well call it 2FF++ nm FUD process.
 

Jagger100

Supreme [H]ardness
Joined
Oct 31, 2004
Messages
7,700
That's about as ignorant as saying Texas is not USA. What do you think the C in Taiwan ROC/PRC stands for?

That's like reasoning the United States of America refers to all the States in the American Continents.

Taiwan is filled with Chinese who fled from the rule of the current mainland government. They choose not to flaunt their independence and anger the much larger dictatorship just a stone throw away and loose that independence.
 

Jagger100

Supreme [H]ardness
Joined
Oct 31, 2004
Messages
7,700
That's about as ignorant as saying Texas is not USA. What do you think the C in Taiwan ROC/PRC stands for?

That's like reasoning the United States of America refers to all the States in the American Continents. Technically anyone from North or South America are Americans but Call a Canadian an American and see his reaction, likely not positive and definitely will be corrected. Because the term is based on Country which is likely the Usage you are responding to is country based and not race based.

Taiwan is filled with Chinese who fled from the rule of the current mainland government. They choose not to flaunt their independence and anger the much larger dictatorship just a stone throw away and loose that independence. Its unfortunate that some take advantage this low-key tact from Taiwan to spread misinformation to make it easier for Mainland China to one day assert full control.
 
Top