Apple Dual Sources A9 SOCs with TSMC and Samsung: Some Extra Thoughts

Subject: Processors | September 30, 2015 - 09:55 PM |
Tagged: TSMC, Samsung, FinFET, apple, A9, 16 nm, 14 nm

So the other day the nice folks over at Chipworks got word that Apple was in fact sourcing their A9 SOC at both TSMC and Samsung.  This is really interesting news on multiple fronts.  From the information gleaned the two parts are the APL0898 (Samsung fabbed) and the APL1022 (TSMC).

These process technologies have been in the news quite a bit.  As we well know, it has been a hard time for any foundry to go under 28 nm in an effective way if your name is not Intel.  Even Intel has had some pretty hefty issues with their march to sub 32 nm parts, but they have the resources and financial ability to push through a lot of these hurdles.  One of the bigger problems that affected the foundries was the idea that they could push back FinFETs beyond what they were initially planning.  The idea was to hit 22/20 nm and use planar transistors and push development back to 16/14 nm for FinFET technology.

View Full Size

The Chipworks graphic that explains the differences between Samsung's and TSMC's A9 products.

There were many reasons why this did not work in an effective way for the majority of products that the foundries were looking to service with a 22/20 nm planar process.  Yes, there were many parts that were fabricated using these nodes, but none of them were higher power/higher performance parts that typically garner headlines.  No CPUs, no GPUs, and only a handful of lower power SOCs (most notably Apple's A8, which was around 89 mm squared and consumed up to 5 to 10 watts at maximum).  The node just did not scale power very effectively.  It provided a smaller die size, but it did not increase power efficiency and switching performance significantly as compared to 28 nm high performance nodes.

The information Chipworks has provided also verifies that Samsung's 14 nm FF process is more size optimized than TSMC's 16 nm FF.  There was originally some talk about both nodes being very similar in overall transistor size and density, but Samsung has a slightly tighter design.  Neither of them are smaller than Intel's latest 14 nm which is going into its second generation form.  Intel still has a significant performance and size advantage over everyone else in the field.  Going back to size we see the Samsung chip is around 96 mm square while the TSMC chip is 104.5 mm square.  This is not huge, but it does show that the Samsung process is a little tighter and can squeeze more transistors per square mm than TSMC.

In terms of actual power consumption and clock scaling we have nothing to go on here.  The chips are both represented in the 6S and 6S+.  Testing so far has not shown there to be significant differences between the two SOCs so far.  In theory one could be performing better than the other, but in reality we have not tested these chips at a low enough level to discern any major performance or power issue.  My gut feeling here is that Samsung's process is more mature and running slightly better than TSMC's, but the differences are going to be minimal at best.

The next piece of info that we can glean from this is that there just isn't enough line space for all of the chip companies who want to fabricate their parts with either Samsung or TSMC.  From a chip standpoint a lot of work has to be done to port a design to two different process nodes.  While 14 and 16 are similar in overall size and the usage of FinFETS, the standard cells and design libraries for both Samsung and TSMC are going to be very different.  It is not a simple thing to port over a design.  A lot of work has to be done in the design stage to make a chip work with both nodes.  I can tell you that there is no way that both chips are identical in layout.  It is not going to be a "dumb port" where they just adjust the optics with the same masks and magically make these chips work right off the bat.  Different mask sets for each fab, verification of both designs, and troubleshooting the yields by metal layer changes will be different for each manufacturer.

In the end this means that there just simply was not enough space at either TSMC or Samsung to handle the demand that Apple was expecting.  Because Apple has deep pockets they contracted out both TSMC and Samsung to produce two very similar, but still different parts.  Apple also likely outbid and locked down what availability to process wafers that Samsung and TSMC have, much to the dismay of other major chip firms.  I have no idea what is going on in the background with people like NVIDIA and AMD when it comes to line space for manufacturing their next generation parts.  At least for AMD it seems that their partnership with GLOBALFOUNDRIES and their version of 14 nm FF is having a hard time taking off.  Eventually more space will be made in production and yields and bins will improve.  Apple will stop taking up so much space and we can get other products rolling off the line.  In the meantime, enjoy that cutting edge iPhone 6S/+ with the latest 14/16 nm FF chips.

Source: Chipworks

October 1, 2015 | 12:02 AM - Posted by Anonymous (not verified)

GlobalFoundries is using the Samsung 14nm process and not having any problems!

http://www.techpowerup.com/216361/globalfoundries-14-nm-lpp-finfet-node-...

October 1, 2015 | 05:42 AM - Posted by John H (not verified)

Tapeout means 9-12 months to go before high volume..

October 1, 2015 | 09:58 AM - Posted by Josh Walrath

Yeah, having good yields on test wafers is one thing... getting product out in a timely manner is another.  Lot of work to do yet before we start seeing chips flow from GF's 14nmFF.

October 1, 2015 | 10:18 AM - Posted by Anonymous (not verified)

Apple are a bunch of retailer pricks and form over function style mavins, they do not even provide proper technical information on their CPU/SOC products! What kind of real technology company keeps CPU core specifications hidden, and is so secretive about its products, so secretive that they hired the one technology journalist outside of a pay wall that could even figure out their CPU's inner workings to shut him up!

October 1, 2015 | 10:04 AM - Posted by Anonymous (not verified)

It's Samsung's 14nm process, and GlobalFoundries(GF) is probably paying Samsung some money for fab process engineering help in addition to licensing the process from Samsung. But GF is not behind in getting the process online from the time they licensed it from Samsung. The more 14nm work GF gets the more licensing money Samsung makes, and there will be enough work to keep all the chip fab companies in the game. Samsung makes everything including the kitchen sink.

October 1, 2015 | 12:10 AM - Posted by Anonymous (not verified)

In other news.

"Tear teardown down, roars Apple: iFixit app yanked from store"

http://www.theregister.co.uk/2015/10/01/apple_drops_ifixit/

iFixit comments from article:

"Not too long ago, we tore down the Apple TV and Siri Remote. The developer unit we disassembled was sent to us by Apple," Wiens explained.

"Evidently, they didn't intend for us to take it apart."

October 1, 2015 | 05:47 AM - Posted by John H (not verified)

Very nice info Josh!

I believe Sammy already had high volume Exynos running through its 14nm process. Is A9 the first retail product on TSMC?

October 1, 2015 | 10:01 AM - Posted by Josh Walrath

Yeah, Samsung has been fabbing 14nm parts since last year.  They launched S6 with plenty of product.

A9 sure does look to be the first big, major product that has good visibility from TSMC's 16FF+.  There likely are other products already out there, but we just don't see them.

October 1, 2015 | 10:32 AM - Posted by Cyclops

It would be ideal if apple had their own foundry. They have a lot of money, but not enough to start one and be competitive with the rest of the market.

They apparently have close to 200 Billion USD in reserves, but that's only a fraction of the net worth of a company such as TSMC.

October 1, 2015 | 03:42 PM - Posted by Anonymous (not verified)

Why would Apple need to own a foundry and have all the up-keep expenses of maintaining the fab equipment, when it's much better to competitively contract out the chip fab work to a dedicated foundry. Apple just contracts its work out to the third party chip foundries and saves billions on not having to pay for physical plant and fab equipment upkeep and fab process R&D.

Samsung, GlobalFoundries, TSMC, others have to compete for Apple's business and Apple gets the fab work through these various third party chip foundries and second sources its A series chip production to more than one chip foundry to better maintain production security. Hell Intel's fab problems at 14nm caused Apple to forgo some updated processors for its Macbook line of products, and Apple was probably not to happy with Intel about that, and anyone can have chip foundry issues once in a while!

Apple has a market cap of 622.40 Billion dollars as of Oct. 1, 2015, TSMC's market cap is $102.37 Billion dollars, Apple is worth 6 times as much as TSMC, so your calculations are a bit off, and who wants the expenses of maintaining a chip fabrication plant when they can save money and contract out the production and save billions.

October 1, 2015 | 08:32 PM - Posted by Cyclops

You're right, who wants all that expense and upkeep for the fabs. Intel should sell theirs and outsource their processors, right?

October 1, 2015 | 09:32 PM - Posted by Josh Walrath

Heh, I get your point.  There is a reason why AMD shed its fabs.  They are expensive to maintain, R&D on new processes is insane, and Apple very likely would not be able to keep the lines filled.  Intel is a different case because they have so many products that they manufacture.  They also have been slowing down their fab business because they simply don't have enough product to keep all of those fabs busy.  That is why they started doing limited foundry work for a handful of companies that they don't view as competitors.

So, it needs to make financial sense to operate your own fabs, otherwise any lul in sales will cause some significant penalties when it comes to shutting down production at the fabs.  Hiring and firing skilled people who work there is not a good option for anyone when boom and bust cycles invariably come around (like for AMD).

October 1, 2015 | 11:13 PM - Posted by Cyclops

Yeah, true. Some fabs keep open even when they have outlived their usefulness. Sandy bridge and Ivy bridge were 32 and 22nm respectively. Their PCH (P67,Z77, etc.) however, was of the 65nm variety.

When did Intel build their 65nm fabs? I think it was back in 2006. so 2006 to 2013 is a healthy lifespan for a fab, although at lower production rates. 28nm fabs I think have been one of the most profitable in history. They started in 2010-2011 and still going strong, thanks to immature 20nm finfet.

They could save more money in the long run if they switch to 450mm wafers but the retooling costs are insane.

October 2, 2015 | 09:38 AM - Posted by Josh Walrath

A lot of thought has gone into 450mm wafers, and the last I heard is that all cost analysis covering the next few years shows that in the end there will still be excess capacity at the major fabs for 300mm.  Sure, we will have limited capacity at the cutting edge nodes for the first half year or so of their existence, but that evens out pretty quickly... especially when a node will last as long as they do now!  So, going 450 will only exacerbate the problem, plus drain a lot of money on retooling costs.  Perhaps eventually we will get to a point where 450 makes sense, but right now there is no movement in that directions.

October 2, 2015 | 10:44 PM - Posted by Cyclops

Czochralski is rolling in his grave. I guess the 200mm to 300mm transition put quite a big dent in their pockets.

October 2, 2015 | 12:13 PM - Posted by Anonymous (not verified)

What about that brand new Fab facility (Fab 42) that Intel had built, and then before the equipment was purchased Intel halted the work and put the building in mothballs! Chip fabs bleed money if they are not operating at as close to 100% capacity as possible! Even Intel takes a hit when its costly fabs can not be run at full capacity.

Intel still has to pay upkeep on the empty building, I'll bet that Intel uses the building to store all those unsold and unwanted ultrabook SOC SKUs that the OEM's are still trying to pawn off on consumers, I don't see Apple offering any quad core Mac Minis anymore! Do the auto makers still make all the parts for their cars anymore! The automakers have not been fully vertically integrated for decades now, they get their parts made by third party parts suppliers and save money in the process, what is the difference with the Chip makers. Running car parts plants is expensive if those plants sit idle, the same for chip fabs!

October 1, 2015 | 03:00 PM - Posted by Anonymous (not verified)

A lot of information that could be useful here with some downloads too.
http://www2.techinsights.com/l/8892/2015-09-28/zxx9c

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.