AMD and GlobalFoundries Wafer Supply Agreement Now Non-Exclusive, Paves Way for 7nm sIOD

[btarunr]

AMD in a filing with the U.S. Securities and Exchange Commission (SEC), revealed that its wafer supply agreement with GlobalFoundries has been amended. Under the new terms, AMD places orders for wafers from GlobalFoundries up to 2024, with purchase targets set for each year leading up to 2024. Beyond meeting these targets, AMD is free from all other exclusivity commitments. The agreement was previously amended in January 2019, setting annual purchase targets for 2019, 2020, and 2021, while beginning a de-coupling between AMD and GlobalFoundries. This enabled the company to source 7 nm (or smaller) chips, such as CCDs and GPUs, from other foundries, such as TSMC, while keeping GlobalFoundries exclusive for 12 nm (or larger) nodes.

The updated wafer supply agreement unlocks many possibilities for AMD. For starters, it can finally build a next-generation sIOD (server I/O die) on a more efficient node than GlobalFoundries 12LP, such as TSMC 7 nm. This transition to 7 nm will be needed as the next-gen "Genoa" EPYC processor could feature future I/O standards such as DDR5 memory and PCI-Express Gen 5, and the switching fabric for these could be too power-hungry on 12 nm. The "Zen 4" CPU core complex dies (CCDs) of "Genoa" are expected to be built on TSMC 5 nm.



View at TechPowerUp Main Site

 

[thesmokingman]

Finally!

 

[olymind1]

Question: Does the SoC needs to be on smaller manufacturing process for desktop / server processors? For laptops and small devices i can see the need for even lower power consumption.

 

[bencrutz]

Question: Does the SoC needs to be on smaller manufacturing process for desktop / server processors? For laptops and small devices i can see the need for even lower power consumption.

well, renoir have excellent memory controller, fabbed in TSMC's 7nm - no separate IOD, it's a monolithic design.
I would expect the IOD perform much better, moving to 7nm.

 

[nBagasW]

In current shortage of Zen 3 CPU, sometimes I wonder what if AMD keep producing Zen+ based CPU like 1600AF, 1700AF and the likes with Glofo 12nm process. I think it will still sell well if priced reasonably.

 

[Mathragh]

Question: Does the SoC needs to be on smaller manufacturing process for desktop / server processors? For laptops and small devices i can see the need for even lower power consumption.

The SOC would definitely gain from being produced on a smaller and more frugal node. Current AMD server hardware is actually suffering quite a bit from a very high base line power use because of the IO die.
A lot of interesting info on this topic can be found in this anandtech review: Power & Efficiency: IOD Power Overhead? - AMD 3rd Gen EPYC Milan Review: A Peak vs Per Core Performance Balance (anandtech.com)

 

[Wirko]

Beyond meeting these targets, AMD is free from all other exclusivity commitments.

Did any exlusivity commitments exist before this last amendment? This is what AnandTech reported two years ago:

The big news with the latest WSA update is that it allows AMD to contract any foundry to produce chips at 7 nm and thinner nodes without any one-time payments or royalties to GlobalFoundries.

 

[Chrispy_]

The real question is why the hell isn't GF making a better process?

I know there's still a healthy market for their 12/14nm process but they can't just stick their head in the sand and ignore newer, faster, denser, more efficient nodes; That way lies a dead end and the end of GloFo.

 

[Voluman]

The real question is why the hell isn't GF making a better process?

I know there's still a healthy market for their 12/14nm process but they can't just stick their head in the sand and ignore newer, faster, denser, more efficient nodes; That way lies a dead end and the end of GloFo.

Well it is simple, money.

 

[Wirko]

The real question is why the hell isn't GF making a better process?

I know there's still a healthy market for their 12/14nm process but they can't just stick their head in the sand and ignore newer, faster, denser, more efficient nodes; That way lies a dead end and the end of GloFo.

They may license 7nm without EUV in 2024-2025, for example, and then buy used and new equipment plus the knowhow to set it up. Setting up a production line and achieving good yields should cost far less and take much less time when the technology is mature. 7nm will be far, far from obsolete at that time, and there will be no shortage of potential customers looking to move their 45nm or 28 nm designs to something newer.
It's a strategy can keep GF profitable in the long term, even it they always choose to follow rather than lead; but they must at least follow, not just stick their head in silica sand forever.

 

[quadibloc]

Since Intel seems to be coming back - their latest 11th generation chips don't seem too shabby, even though they're still on 14nm - it's high time that AMD no longer is fighting with one hand tied behind its back!

 

[TheinsanegamerN]

The real question is why the hell isn't GF making a better process?

I know there's still a healthy market for their 12/14nm process but they can't just stick their head in the sand and ignore newer, faster, denser, more efficient nodes; That way lies a dead end and the end of GloFo.

GloFo dumped tons of money into their sub 28nm and was getting nowhere. Rather then invest entire country's worth of GDP into it like intel did, GloFo sought instead to license nodes from other sources. Their 14nm tech is licensed from Samsung, for instance, and it stands to reason that GloFo will do the same to samsung 8nm when samsung moves up again

Foundry Futures: TSMC, Samsung, GlobalFoundries, and Intel Gear Up For 7nm and Beyond

Samsung, TSMC, GlobalFoundries, and Intel all have major technology ramps scheduled for the next four years. Here's what is coming down the pipe.

www.extremetech.com

There are plenty of customers that dont need a cutting edge node, and GloFo is more then happy to fulfill those quotas. Doesnt do much for us enthusiasts but it's a huge help for other sectors that just need processors for various things.

 

[mechtech]

If only the GF had a good 7nm node, it would be a win for everyone.

 

[seronx]

The real question is why the hell isn't GF making a better process?

I know there's still a healthy market for their 12/14nm process but they can't just stick their head in the sand and ignore newer, faster, denser, more efficient nodes; That way lies a dead end and the end of GloFo.

GlobalFoundries has a UMC-like problem;

Other than AMD/IBM, GlobalFoundries has a normal amount of zero customers on 14nm/12nm. No customers to support a push to 7nm, means no 7nm and lower nodes. AMD/IBM ditched GlobalFoundries when they would have had to give $3B to get 7LP to 12,000 wpm. Since, no else bothered thus it died and it will remain dead.

Now compare UMC to TSMC:

GlobalFoundries canning 7nm makes sense. GlobalFoundries doing whatever happened to 12FDX, makes none. Unlike, the FinFET nodes, it actually had reserved customers in 2017.

Dream Chip: A Vision for Your Car

Dream Chip is a company based in Germany just outside Hannover. Martin Zeller presented at CDNLive Silicon Valley on A New Computer Vision Processor Chip Design...

community.cadence.com

"Migrate to 12FDX (12nm)"

Dream Chips then and now Goodix: Veyron to Chiron

There is also a case that the Rockchip 3588 potentially was going to be on 12FDX as well.

Basically, anything with 8LPP could have been 12FDX instead. Especially, if there was a 22nm node that is synonymous with 22FDX.

 

[Wirko]

GloFo dumped tons of money into their sub 28nm and was getting nowhere. Rather then invest entire country's worth of GDP into it like intel did, GloFo sought instead to license nodes from other sources. Their 14nm tech is licensed from Samsung, for instance, and it stands to reason that GloFo will do the same to samsung 8nm when samsung moves up again

Foundry Futures: TSMC, Samsung, GlobalFoundries, and Intel Gear Up For 7nm and Beyond

Samsung, TSMC, GlobalFoundries, and Intel all have major technology ramps scheduled for the next four years. Here's what is coming down the pipe.

www.extremetech.com

There are plenty of customers that dont need a cutting edge node, and GloFo is more then happy to fulfill those quotas. Doesnt do much for us enthusiasts but it's a huge help for other sectors that just need processors for various things.

Yes. GF is a part of the alliance that seems to still be alive.

https://www.samsungfoundry.com/foun....do?_mainLayOut=homepageLayout&menuIndex=0201

 

[Chrispy_]

GloFo dumped tons of money into their sub 28nm and was getting nowhere. Rather then invest entire country's worth of GDP into it like intel did, GloFo sought instead to license nodes from other sources. Their 14nm tech is licensed from Samsung, for instance, and it stands to reason that GloFo will do the same to samsung 8nm when samsung moves up again

Foundry Futures: TSMC, Samsung, GlobalFoundries, and Intel Gear Up For 7nm and Beyond

Samsung, TSMC, GlobalFoundries, and Intel all have major technology ramps scheduled for the next four years. Here's what is coming down the pipe.

www.extremetech.com

There are plenty of customers that dont need a cutting edge node, and GloFo is more then happy to fulfill those quotas. Doesnt do much for us enthusiasts but it's a huge help for other sectors that just need processors for various things.

Interesting.

I wonder what the delay is this time? GloFo were there with 14nm at the beginning when AMD, Apple, & Nvidia finally moved off TSMC 28nm. If they're licensing from Samsung, why give up the unstoppably profitable chance to license Samsung 8nm? TSMC has been raking in ungodly amounts of money since 7FF and I figured GloFo would love to compete with that, even if it's only a licensed process and not 100% their own.

As of the last 15 months, anything made on 7nm or 8nm will sell for top dollar, no matter what.

 

[Calmmo]

If only the GF had a good 7nm node, it would be a win for everyone.

I mean if you were willing to give them the billions required for that i'm sure they wouldn't object.

 

[mechtech]

I mean if you were willing to give them the billions required for that i'm sure they wouldn't object.

Of course they wouldn't, but it's up to the companies to invest in themselves to get more and better nodes to get more customers. I mean it seems to be there is a big market for 7/8nm, but I could be wrong.