Chip giant partners with GlobalFoundries to fabricate Apple
chips
Samsung strikes chip deal with Apple
chips
Samsung strikes chip deal with Apple
Samsung strikes chip deal with Apple
- Thread starter astilo
- Start date
user_2013101
Banned
So with rumors favoring TSMC for Apple's future business, we now see Samsung is aggressively upping their game in the press.
...
First they were rumored to have had the deal in hand in October. Now it's a fresh rumor that no, the deal was really signed in November. At some point you realize that it's just a reoccurring rumor to sell to Wall Street based on an "unidentified customer," which could be anybody or no one. So did Samsung really land a signed deal with Apple giving them the majority of their future chip business? While I highly doubt it, only time will tell.
Did Apple Really Sign a Deal with Samsung for 80% of their 14nm Processors? - Patently Apple
...
First they were rumored to have had the deal in hand in October. Now it's a fresh rumor that no, the deal was really signed in November. At some point you realize that it's just a reoccurring rumor to sell to Wall Street based on an "unidentified customer," which could be anybody or no one. So did Samsung really land a signed deal with Apple giving them the majority of their future chip business? While I highly doubt it, only time will tell.
Did Apple Really Sign a Deal with Samsung for 80% of their 14nm Processors? - Patently Apple
I'm pretty sure that the current Apple idea is to go ahead with both the options. Also Qualcomm probably will do the same. There are few companies that can really afford the costs of a double development anyway.
The other "big question" is: is there anybody willing to move to Intel?
The other "big question" is: is there anybody willing to move to Intel?
In order for TSMC or Samsung to deliver wafers to Apple in time for the iProduct refresh next September the design must have been taped out in Q4 0f 2014 (which is now). Figure on a 12 month leading edge SoC design cycle at a minimum and you will see that the wafer agreements have been signed long ago. There are usually milestones in early access wafer agreements and if they are not hit new agreements can be made.
Apple has strict policies in regards to confidentiality of their suppliers so anything said here are anyplace else is a rumor and rumors out of Korea are even more unreliable than most (my opinion).
One thing I can tell you is that, like the A8, there will be two versions of the A9, one for the iPhone and one for the Ipad (and hopefully a MAC Air). The word in Silicon Valley was that the Samsung 14nm LP process is targeted for the A9 and the TSMC FF+ process for the A9x. This was based on both price and availability since the Samsung FinFET process was reportedly 6 months ahead of TSMC at the beginning of this year.
The most recent word in Silicon Valley is that TSMC has gained on Samsung and will have FinFET wafers out at about the same time. However, I still say that Samsung has the iPhone and TSMC has the iPad given the tape out schedules. I don't know if that will be a 80/20 split. In Q3 2014 Apple sold about 33.8M iPhones and 14.1M iPads so that is closer to a 60/40 split.
I agree, but it will probably cost only 1/10th of the A8. Guessing the size, it could be a 2k chips per wafer (assuming a 300mm wafer in 28nm node), so taking as good the 40M figure, we are talking about 20K wafers. Not bad, sure, but the only one making good money will be eventually Apple.
user_2013101
Banned
Apple’s A9 processors will be mostly built at TSMC. Samsung may, just may, get a small portion of the orders, mainly because Apple may wish to have a second source as leverage for price negotiation with TSMC.
It’s purely PR that Samsung will produce most, if not all, of the A9 chips.
I don’t know exactly why Samsung unleashed such ridiculous PR in recent months. For example, after Apple’s A8 was confirmed, via Chipworks teardown, to be built from TSMC’s 20nm, IHS filed a report said that Samsung still produces 40% of A8 chips, without giving any evidences. Then, many news media cited the report. That’s big a lie.
During 2012 and 2013, when many rumors of Apple diversifying part of processor productions to TSMC, Samsung didn’t launch PR offensives. Samsung was producing chip for Apple and ignored the rumors.
This year, on Oct 2, 2014, a news story cited Samsung’s president, Kim Ki-nam, “told reporters at headquarters in Seoul that once the company begins to supply Apple with chips using its latest technology, profits ‘will improve positively,’ ZDNet reports.”
This is highly unusual, because Apple’s contracts include confidentiality clauses and Samsung or TSMC had never publicly commented on Apple contracts. It is not an indication of Samsung having the contract; instead, on the contrary, more likely Samsung had lost the contract and is not restricted by confidentiality any more. Samsung can conduct its PR and mention Apple directly.
It takes Apple three years to diversify away from Samsung. TSMC’s 20nm improves A8’s power consumption, and enables larger displays. A8’s unusually high transistor density surprises even the die-hard Intel loyalists. IPhone 6 turns out a bigger success than expected. Since the iPhone 6 release, Apple has been taking market shares from Samsung. But, now Apple suddenly decides to abandon the winning partnership with TSMC, take a U turn, and go back to Samsung for processor production?
Finally, Apple tends to pick the best technologies for its iDevices. The 16FF+ is better than Samsung’s 14nm in performance, density, and yields. In addition to ditching its 3 years’ effort of diversification, Apple wants to bet its next iDevices on Samsung’s inferior and unstable 14nm process?
16FF+ is NOT late for next year’s Apple releases. Even if the volume production starts in July 2015, with much faster ramping than 20nm and superb yield, it’s not late. This year’s 20nm production from January didn’t deliver high output until late May, anyway.
In reality, the 16FF+ volume production mostly likely will start in early 2Q15 or late 1Q15; that is, a quarter earlier than announced.
My opinion: Samsung will be lucky to receive a small portion of the A9 orders. Vast majority, if not all, of A9s will be manufactured at TSMC.
We will find out for sure in 4-6 months.
It’s purely PR that Samsung will produce most, if not all, of the A9 chips.
I don’t know exactly why Samsung unleashed such ridiculous PR in recent months. For example, after Apple’s A8 was confirmed, via Chipworks teardown, to be built from TSMC’s 20nm, IHS filed a report said that Samsung still produces 40% of A8 chips, without giving any evidences. Then, many news media cited the report. That’s big a lie.
During 2012 and 2013, when many rumors of Apple diversifying part of processor productions to TSMC, Samsung didn’t launch PR offensives. Samsung was producing chip for Apple and ignored the rumors.
This year, on Oct 2, 2014, a news story cited Samsung’s president, Kim Ki-nam, “told reporters at headquarters in Seoul that once the company begins to supply Apple with chips using its latest technology, profits ‘will improve positively,’ ZDNet reports.”
This is highly unusual, because Apple’s contracts include confidentiality clauses and Samsung or TSMC had never publicly commented on Apple contracts. It is not an indication of Samsung having the contract; instead, on the contrary, more likely Samsung had lost the contract and is not restricted by confidentiality any more. Samsung can conduct its PR and mention Apple directly.
It takes Apple three years to diversify away from Samsung. TSMC’s 20nm improves A8’s power consumption, and enables larger displays. A8’s unusually high transistor density surprises even the die-hard Intel loyalists. IPhone 6 turns out a bigger success than expected. Since the iPhone 6 release, Apple has been taking market shares from Samsung. But, now Apple suddenly decides to abandon the winning partnership with TSMC, take a U turn, and go back to Samsung for processor production?
Finally, Apple tends to pick the best technologies for its iDevices. The 16FF+ is better than Samsung’s 14nm in performance, density, and yields. In addition to ditching its 3 years’ effort of diversification, Apple wants to bet its next iDevices on Samsung’s inferior and unstable 14nm process?
16FF+ is NOT late for next year’s Apple releases. Even if the volume production starts in July 2015, with much faster ramping than 20nm and superb yield, it’s not late. This year’s 20nm production from January didn’t deliver high output until late May, anyway.
In reality, the 16FF+ volume production mostly likely will start in early 2Q15 or late 1Q15; that is, a quarter earlier than announced.
My opinion: Samsung will be lucky to receive a small portion of the A9 orders. Vast majority, if not all, of A9s will be manufactured at TSMC.
We will find out for sure in 4-6 months.
hist78
Well-known member
One thing that's really strange to me is that Samsung's blunt attack on Apple for the so-called "iPhone6 bendgate". The attack video was posted on the Samsung's official YouTube channel. Samsung didn't even bother to use a third party to hide that video's origin.
WATCH: Samsung twits Apple over 'Bendgate' in new video | SciTech | GMA News Online
People keep telling me that business is business. Apple doesn't care that much. But I think unless one of the following condition is true, Samsung does care, unless:
A. Only if Samsung already got next year A9 contract signed. So Samsung doesn't care what Apple may respond.
B. Only if Samsung didn't get the A9 contract or hopelessly got a small size of A9 contract. So Samsung doesn't care what Apple may respond.
This is potentially a several billion dollars business for Samsung. Should Samsung's CEO have told their PR/Marketing people be a little bit polite until things settled?
Or am I too logical to understand this industry?
WATCH: Samsung twits Apple over 'Bendgate' in new video | SciTech | GMA News Online
People keep telling me that business is business. Apple doesn't care that much. But I think unless one of the following condition is true, Samsung does care, unless:
A. Only if Samsung already got next year A9 contract signed. So Samsung doesn't care what Apple may respond.
B. Only if Samsung didn't get the A9 contract or hopelessly got a small size of A9 contract. So Samsung doesn't care what Apple may respond.
This is potentially a several billion dollars business for Samsung. Should Samsung's CEO have told their PR/Marketing people be a little bit polite until things settled?
Or am I too logical to understand this industry?
I think there one more thing to remember. Having Apple as a customer is both a blessing and a curse. Forgive me if I do not remember correctly, but in the past, TSMC has said that it wont behave to Apple in an exclusive way nor will it build specific fabs etc just for Apple. In the 20nm node, things are kinda easy. Who else is designing for this node? I think it is practically only Apple. Almost all of the big names, are staying with 28nm and moving then directly to 16nm. So it is relatively easy for Apple now to have all the resources it may need, but in the 16nm era this is not going to be the case. So, no matter how good or bad, early or late, 14nm tech from Samsung is going to be, it seems to me that Apple needs it because it is going to compete for capacity in TSMC. The fact that except Samsung, GF is offering a copy exact process (as I understand it, since GF has licensed the Samsung 14nm process, a design can be migrated from Samsung to GF without having to be redesigned), gives Apple quite a bit of negotiating leverage. So my guess is that Apple is going to get a really good deal from Samsung.
Especially if you have enormous volumes like Apple you'd think there there would be a cost advantage going with Samsung/GF 14nm (dual sourced, maybe not so many products) instead of TSMC 16FF+ (single sourced, everyone and their dog wants to use it). This would fit with the A9 using Samsung 14LP (cost more important) and the A9x using TSMC 16FF+ (performance more important).
hist78
Well-known member
Back to the basic. Does any one of you know the true 14/16nm capacity among these three players, TSMC, Samsung, and GF? Is GF's capacity on 14nm really meaningful or significant for the next two years? And considering the possible different yield rates between them, who is really cheaper or expensive? Or if any history can tell? Sorry, my question probably only CIA can answer.
Arthur Hanson
Well-known member
Remember when both Apple and Qualcom offered TSM one billion to cut to the head of the line?
Yes, we all laughed at that one. Funny stuff! I think that one was from NotsoAccurate.com
user_2013101
Banned
In 4Q14, 20nm, mostly for A8 and A8x, counts for 20% of TSMC revenue. To lose 60-80% of A9 production to Samsung next year may lead to 12-16% revenue decline. Comparing to this year’s 25% growth, the potential drop next year will be a catastrophic blow to TSMC.
However, the TSM prices have hit a new high this week. To believe Samsung’s propaganda, we must assume that the market had done a drastic miscalculation on TSM.
Samsung’s powerful PR proclaims that its 14nm is superior and thus won the A9 orders. However, curiously, at ARM Tech Conference in early Oct, instead of providing details and showing off its technological prowess, Samsung was deafening quiet on the process.
"Samsung showed a working 14nm FinFET device at the ARM Tech Con here. However the Korean giant would not provide details of the chip, any of its customers for the process, or the status of the process."
Apple’s orders are massive. If Samsung is to start A9 production in 4Q14 or 1Q15, then we should have seen at least some preparation, if not full ramping, at Samsung’s fab’s in Austin, TX, and at GF New York fab’s, too. To the best of my knowledge, there is no such development, so far. Well-connected people can inquire into contacts and see if any such ramping is ongoing.
At best, Samsung may start low volume 14nm production for its own mobile SoCs.
Other than Samsung’s PR, no circumstantial evidences to confirm Samsung will produce most of A9’s.
Last week, TSMC board approved US$5.5 billions of capex, just in time for the A9 production to commence around end of 1Q or early 2Q.
2014/11/11
TSMC Board of Directors Meeting Resolutions
However, the TSM prices have hit a new high this week. To believe Samsung’s propaganda, we must assume that the market had done a drastic miscalculation on TSM.
Samsung’s powerful PR proclaims that its 14nm is superior and thus won the A9 orders. However, curiously, at ARM Tech Conference in early Oct, instead of providing details and showing off its technological prowess, Samsung was deafening quiet on the process.
"Samsung showed a working 14nm FinFET device at the ARM Tech Con here. However the Korean giant would not provide details of the chip, any of its customers for the process, or the status of the process."
Apple’s orders are massive. If Samsung is to start A9 production in 4Q14 or 1Q15, then we should have seen at least some preparation, if not full ramping, at Samsung’s fab’s in Austin, TX, and at GF New York fab’s, too. To the best of my knowledge, there is no such development, so far. Well-connected people can inquire into contacts and see if any such ramping is ongoing.
At best, Samsung may start low volume 14nm production for its own mobile SoCs.
Other than Samsung’s PR, no circumstantial evidences to confirm Samsung will produce most of A9’s.
Last week, TSMC board approved US$5.5 billions of capex, just in time for the A9 production to commence around end of 1Q or early 2Q.
2014/11/11
TSMC Board of Directors Meeting Resolutions
Last edited:
Arthur Hanson
Well-known member
Dan, I think this is where the one billion came from
Apple, Qualcomm Bids Spurned for Exclusive TSMC Supplies - Bloomberg
Bloomberg is generally accurate and the investment would be in cheap foriegn earnings for both Apple and Qualcom since they have a ton of money they can't repatriate except at great tax expense. I think this report is accurate, but if not, please let me know. I know TSM had to stretch to build the latest giga fab that cost just shy of 17 billion.
Apple, Qualcomm Bids Spurned for Exclusive TSMC Supplies - Bloomberg
Bloomberg is generally accurate and the investment would be in cheap foriegn earnings for both Apple and Qualcom since they have a ton of money they can't repatriate except at great tax expense. I think this report is accurate, but if not, please let me know. I know TSM had to stretch to build the latest giga fab that cost just shy of 17 billion.
Last edited:
user_2013101
Banned
On 16FF+ volume production
If the 90% yield achieved at the 16FF+ risk production is true, the threshold of volume production has been surpassed. It’s just a matter of time to install and fine-tune additional equipments before cranking out 50K-60K wafers per month, perhaps beginning in July.
In other words, the volume production has started for all intents and purposes, except the nametag has not yet been changed from “risk” to “volume.”
Since 90% of equipments are shared between 20 and 16 nm productions, some 20nm capacity can be readily converted to 16FF+ production, to bump up the volume quickly, if necessary.
If the 90% yield achieved at the 16FF+ risk production is true, the threshold of volume production has been surpassed. It’s just a matter of time to install and fine-tune additional equipments before cranking out 50K-60K wafers per month, perhaps beginning in July.
In other words, the volume production has started for all intents and purposes, except the nametag has not yet been changed from “risk” to “volume.”
Since 90% of equipments are shared between 20 and 16 nm productions, some 20nm capacity can be readily converted to 16FF+ production, to bump up the volume quickly, if necessary.
Last edited:
Correct me if I'm wrong, but I would say it's almost impossible to start with 90% yield in a new node. Even given that it has the same back end as 20nm, it's still no cake walk. The manufacture of fins that are both thin and tall is quite difficult.
As well as that TSMC has no plans to convert any 20nm in 2015. They said that 20nm will be slightly less than 20% of revenue in 2015, which is the same as it is at end of 2014.
user_2013101
Banned
On Nov 12, Credit Suisse analyst Randy Abrams wrote: "We believe 20nm yields are approaching 80%, and 16 nanometer FF+ SRAM yields exceed 90%"
Since TSMC, or foundries in general, does not publish specific yield numbers, we have no way to know for certain. Besides, the SRAM yield may not be the same as that from A9 test chips. Therefore, I started my post with an “IF.”
I think you and Lefty and many others perhaps have this question in mind:
Intel suffers multiple delays during the rollout of 14nm FinFET, how can TSMC deliver the 16FF+ with ease?
I am no expert in the process technologies. I try to relay what I heard as follows:
Intel’s 22nm is really 28nm with FinFET. Intel jumps from the so-called 22nm directly into 14nm.
TSMC approaches the FinFET more programmatically. After achieving high yields at the 28nm planar, it went first to 20nm planar, where it perfected the double patterning. Then, it added FinFET onto 20nm BEOL with minor improvements, and the new process is 16FF. After the FinFET stabilized, the BEOL was further improved and the process becomes 16FF+.
That’s why TSMC stated that, leveraging the 20nm experience, the 16nm has achieved faster yield improvements than any prior nodes. At conference calls, sometimes TSMC lumps together 20nm and 16nm as a big node.
At end, it appears that TSMC’s step-by-step approach turns out better.
Last edited: