You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!

AMD Officially Diversifies 14nm Manufacturing With Samsung

I don’t normally jump on the quick news of the day as I’m an industry analyst and not a journalist, but I came across a tidbit on Advanced Micro Devices that I thought would be important to my audience. News about where a chip company manufactures their silicon is important as it provides insights into the bounding box, capabilities and risk profile of the company over the next 1-3 years.

View attachment 17976

Following Advanced Micro Devices positive second quarterly earnings release, many industry analysts were peppering Advanced Micro Devices with questions on exactly where they were manufacturing their upcoming 14nm products. For years, AMD has been competitive with NVIDIA on graphics but behind Intel microprocessor and SoC nodes and process technology, and if AMD can successfully move to 14nm, it has the potential to narrow the gap. In my opinion, not AMD’s opinion, GlobalFoundries has been a major disappointment for years as they struggled with AMD silicon, often being their decelerator, so any diversification from GF is a positive one.

AMD sent me some answers on their 14nm plans and this is a really big deal:

“AMD has strong foundry partnerships and our primary manufacturing partners are GLOBALFOUNDRIES and TSMC. We have run some product at Samsung and we have the option of enabling production with Samsung if needed as part of the strategic collaboration agreement they have with GLOBALFOUNDRIES to deliver 14nm FinFET process technology capacity.”

Essentially, AMD has run real products, not just test chips through Samsung’s fab and have the ability now, based on an ambiguous “if needed” trigger point, to fab there. The quote above creates many questions, I understand, but just the notion that AMD has the ability to fab their products at Samsung is a positive, as Samsung has been so successful the past few years. Also note that this wouldn’t be the same exact process as Samsung fabs Apple products on as APUs and SoCs need a higher performance process, not the ones smartphone chips are done with.

All in all, this lowers the risk to AMD and their products and puts them in a much better position on CPUs, APUs and SoCs then they have been in years.
 
Last edited by a moderator:

Daniel Nenni

Admin
Staff member
There are two reasons for this:

  1. Samsung and GlobalFoundries are GDS compatible at 14nm so it shouldn't matter which foundry is used (GF licensed a copy exact version of 14nm from Samsung).
  2. GF is skipping 10nm so AMD has no choice but to move to a different foundry. The choice is Samsung or TSMC. From what I have heard TSMC has a pretty big entry fee ti limit the number of customers at 10nm, similar to what they did at 20nm. This will allow TSMC to focus on a quick transition to 7nm, same as they did from 20nm to 16nm. Remember, 10nm and 7nm use the same fabs like 20 and 16nm did so it is deja vu all over again.
Unlike TSCM, Samsung will not do a fast transfer to 7nm and will build out the 10nm process with different versions for low power and high performance etc....
 

Fred Chen

Member
No diversification at 10nm

There are two reasons for this:

  1. Samsung and GlobalFoundries are GDS compatible at 14nm so it shouldn't matter which foundry is used (GF licensed a copy exact version of 14nm from Samsung).
  2. GF is skipping 10nm so AMD has no choice but to move to a different foundry. The choice is Samsung or TSMC. From what I have heard TSMC has a pretty big entry fee ti limit the number of customers at 10nm, similar to what they did at 20nm. This will allow TSMC to focus on a quick transition to 7nm, same as they did from 20nm to 16nm. Remember, 10nm and 7nm use the same fabs like 20 and 16nm did so it is deja vu all over again.
Unlike TSCM, Samsung will not do a fast transfer to 7nm and will build out the 10nm process with different versions for low power and high performance etc....
So it seems like no diversity at 10nm (only Samsung besides Intel, which is expected off-limits to AMD).
 

Li Yisuo

Member
IBM 7nm was the old news last year. With all those IBM veterans from high performance server field inside GF now, I believe GF should be 1st choice for AMD. High performance process optimization is a bit different from low power. TSMC is obviously the leader in low power applications.
 

lefty

Member
There is a wafer supply agreement between AMD and Globalfoundries. They have to manufacture most of their stuff at Globalfoundries.
 

benb

Member
I'm pretty curious about how 7nm is coming almost at the same time as 10nm. And skeptical.

As we know, 20nm became "14nm" or "16nm" with the addition of Finfets to a 20nm process flow. So the enabler of a fast transition was the simplification of just adding Finfets to a 20nm flow.

At 10nm/7nm, there is no comparable simplification. An alternative explanation: If you are losing the race to 10nm, throw dust in the air by talking up 7nm vaporware.
 

astilo

New member
I'm pretty curious about how 7nm is coming almost at the same time as 10nm. And skeptical.

As we know, 20nm became "14nm" or "16nm" with the addition of Finfets to a 20nm process flow. So the enabler of a fast transition was the simplification of just adding Finfets to a 20nm flow.

At 10nm/7nm, there is no comparable simplification. An alternative explanation: If you are losing the race to 10nm, throw dust in the air by talking up 7nm vaporware.
From GF
"We would like to confirm that GF is transitioning directly from 14nm to 7nm. We consider 10nm as more a half node in scaling, due to its limited performance adder over 14nm for most applications. For most customers in most of the markets, 7nm appears to be a more favorable financial equation. It offers a much larger economic benefit, as well as performance and power advantages, that in most cases balances the design cost a customer would have to spend to move to the next node.
...we will be leveraging our presence at SUNY Polytechnic in Albany, the talent and know-how gained from the acquisition of IBM Microelectronics, and the world-class R&D pipeline from the IBM Research Alliance—which last year produced the industry’s first 7nm test chip with working transistors."
 

ownerofsouls

New member
Very interesting Astilo... History tells two stories about the fabled IBM microelectronics group - one whom is excellent at publishing R&D proof of concept papers, yet challenged at transitioning to a yielding manufacturing process. The foundry world of today has a different tolerance than the IDM world of yesterday in which the IBM folks experience originates. I hope the know-how from IBM has been updated to bring yielding technologies to market on time, as that was never a strength of theirs at the leading edge (90nm G5 Apple processor, Cell Processor, crumbling failure of IBM process development club for not delivering yielding tech).
 

mbello

Member
It is a well known fact that GF has a copy exact license from Samsung for 14nm, no surprise they could use Samsung.
About the opinion that GF has decelerated AMD, it is hard to put the blame on GF knowing that 14nm is old news and somehow AMD still has no 14nm silicon in the marker and won't have one for another 4-5 months. Maybe we will see 10nm parts shipping by the time the first Zen CPUs will be coming out.
Zen may still be a success, but it is definitely very late.
 

hist78

Member
For all this AMD has greed to write a cheque for $100 million to GlobalFoundries, spread across 2016 and 2017. It will also make unspecified quarterly payments to GlobalFoundries for every wafer it buys from the company's rivals.
In total, the new deal will see AMD hit with a $335 million one-off charge in the third quarter of this year.
So the never ending alimony payment is not cheap. The original AMD/GF spun off agreement seems to create a long lasting negative burden on AMD.
 

astilo

New member
So the never ending alimony payment is not cheap. The original AMD/GF spun off agreement seems to create a long lasting negative burden on AMD.
That's over. This is just a new (and different) deal to me, convenient for both the parties. There are few big customers out there for the foundries. If we leave Intel out of the equation, since they both do design and manufacturing in house, we have basically Apple, Qualcomm, AMD and NVidia. TSMC is currently developing all the new nodes to serve Apple at the very beginning. AMD has not the power (read money) to change that. By striking such a deal with GloFo, they can have instead the possibility to customize and optimize the new technology and on the other hand, GF can share part of the risk of developing the new node with a customer.
Samsung is also in a different position as foundry, since they can manufacture their own SoCs.
 
Top