The AMD Execution Thread [2007 - 2017]

[firstminion]

The most recent link I've seen was in this thread from last year:
https://forum.beyond3d.com/posts/1863877/

The agreement there has no requirement for renewal, as its termination is baked-in. After November 2014, patents filed by either party are no longer covered.
Long-term the agreement will end once all the patents it covers expire. Realistically, if there's not an additional agreement to cover newer tech, it would become less likely that compatibility past what has already been shared will continue.


The joint venture is composed of two entities. One of them is controlled by AMD and falls under the umbrella of the parent company. If there is some kind of gymnastics involved in getting past AMD's obligations, it would seemingly go through there. Manufacturing via foundries was freed up with the most recent agreement, I think. Who controls the actual end result might be where the question is asked about licensing.

Thanks, Dirk Meyer renewed it in 2009. I'm gonna leave the rest to patent lawyers.

 

[Razor1]

http://download.intel.com/pressroom/legal/AMD_settlement_agreement.pdf This is what was the settlement between AMD and Intel that created the changes to the initial contract.

 

[Deleted member 2197]

AMD Announces Joint Venture with NFME

Nantong Fujitsu Microelectronics Co., Ltd. (NFME) is a Chinese company that packages and tests integrated circuits. Recently, AMD has been working with China to reach that large market, especially given their ongoing cash concerns. This time, AMD sold 85% of its stake in two locations, AMD Penang, Malaysia and AMD Suzhou, Jiangsu, China, to NFME and formed a joint venture with them, called TF-AMD Microelectronics Sdn Bhd.

First, AMD gets about $320 million USD in this transaction, after taxes and fees, and it also retains 15% of this venture. I am curious whether this will lead to a long-term source of income for AMD, even though the press release claims that this structure will be “cost neutral”. Either way, clearing a third of a billion dollars should help AMD to some extent. That equates to about two-to-three quarters of net-loss for the company, so it gives them about six-to-nine extra months of life on its own. That's not too bad if the transaction doesn't have any lasting consequences.

Second, NFME now has access to some interesting packaging and testing technologies. NFME's website claims that this allows them to handle dies up to 800mm2, substrates with up to 18 layers, and package sizes up to 75mm. These specifications sound like it pulls from their GPU experience, which could bring all of that effort and knowledge to completely different fields.

The press release states that 1,700 employees will be moved from AMD to this venture. They do not state whether any jobs are affected over and above this amount, though.

http://www.pcper.com/news/General-Tech/AMD-Announces-Joint-Venture-NFME

 

[Grall]

AMD selling pieces of itself to gain breathing room capital obviously isn't a sustainable strategy; I truly hope they'll start doing better soon, because we desperately need competition in the GPU marketplace...!

 

[Razor1]

well what they sold off, I think they could have done this along time ago, don't know why they would keep stuff like this in house anyways, I could understand if they had the fabs still, but after that was gone, this should have been outsourced soon after.

 

[silent_guy]

AMD selling pieces of itself to gain breathing room capital obviously isn't a sustainable strategy; I truly hope they'll start doing better soon, because we desperately need competition in the GPU marketplace...!

The real question is: what took them so long? They had no business being in the packaging business.

 

[Grall]

No?

Who packages Intel's chips? Don't they do that themselves?

 

[silent_guy]

No?

Who packages Intel's chips? Don't they do that themselves?

Intel's business model is one based on filling their fabs with silicon. AMD's business model is one of a fabless company.

Also, the difference in scale between Intel and AMD is massive. Intel can spread its packaging ops NRE (assuming they do it themselves) across a much higher volume.

AMD needs to specialize in what they do best.

 

[Grall]

Also, the difference in scale between Intel and AMD is massive. Intel can spread its packaging ops NRE (assuming they do it themselves) across a much higher volume.

And AMDs packaging subsidiary did not take on third-party work also...?

Just because you don't own any fabs anymore doesn't mean you have to sell off anything chip-related, does it?

 

[CSI PC]

Although moving those staff-assets, what effect does that have on R&D and implementation regarding future multi-chip module (MCM)?
Intel moved this direction with Broadwell, and this is I think integral to Zen.
Also I think AMD also warn they now have more reliance upon GF - I assume with the movement of the team and their assets, which has implications if any delays-issues.
Cheers

 

[silent_guy]

And AMDs packaging subsidiary did not take on third-party work also...?

The question is whether or not that subsidiary had a volume that's similar to its competitors. If doesn't matter whether where that volume comes from. The fact that it was sold for a couple of hundred millions makes me believe that it was just a small bit player in the packaging market.

Just because you don't own any fabs anymore doesn't mean you have to sell off anything chip-related, does it?

Tech is a market that has a high tendency to be winner take all. If you're not number one or number two, you get the scraps. Better to sell it and use it for something where you have a chance to be at the top.

 

[silent_guy]

Although moving those staff-assets, what effect does that have on R&D and implementation regarding future multi-chip module (MCM)?
Intel moved this direction with Broadwell, and this is I think integral to Zen.
Also I think AMD also warn they now have more reliance upon GF - I assume with the movement of the team and their assets, which has implications if any delays-issues.
Cheers

Companies that don't have this technology in-house seem to be able to make MCMs just fine.

 

[3dilettante]

The question is whether or not that subsidiary had a volume that's similar to its competitors. If doesn't matter whether where that volume comes from. The fact that it was sold for a couple of hundred millions makes me believe that it was just a small bit player in the packaging market.

This may have contributed to the long time it took to spin off the facilities, if the established market was generally fine with or without it. Would this go into a desire to create more national tech resources for the Chinese company making the purchase?

If so, it might be a long-term play that made the division interesting enough to buy. Somewhat like how ATIC bought AMD's fabs as part of a long-term project, despite the fact that the market could have gone on without them.

 

[CSI PC]

Companies that don't have this technology in-house seem to be able to make MCMs just fine.

Regarding large die-architecture CPUs/GPUs?
I thought some such as Xilinx still keep this in-house.
Anyone know if NVIDIA outsource this fully themselves to TSMC?
Cheers

 

[silent_guy]

Most big fabless companies have process experts who work closely with the fab partner. I expect this to be no different for advanced packaging.
How could Nvidia not be outsourcing this? What's the alternative? That they receive individual dies and mount them on an interposer themselves when there are companies like Amkor that build a whole business around this?

 

[CSI PC]

Most big fabless companies have process experts who work closely with the fab partner. I expect this to be no different for advanced packaging.
How could Nvidia not be outsourcing this? What's the alternative? That they receive individual dies and mount them on an interposer themselves when there are companies like Amkor that build a whole business around this?

Well my original post was exactly this point and AMD moving 1700 staff to the new joint venture and how it may impact future R&D-project implementations, where you responded they no longer need the staff.
Same way I am sure NVIDIA has the in-house experts in using the TSMC libraries (different subject but same context).

Maybe I am missing something but isn't AMD fabless since they sold off GF?
Just wondering how many staff involved in aspects of CPU/GPU design-project implementation are moved to the joint venture; to me this seems very pertinent considering AMD used to combine expertise of both teams and only recently looked to split the teams apart with one lead by Raja.
It does sound like AMD are losing some because as I mentioned they gave a warning they would be more dependant upon GF in their statement.

Cheers

 

[Kaotik]

Although moving those staff-assets, what effect does that have on R&D and implementation regarding future multi-chip module (MCM)?
Intel moved this direction with Broadwell, and this is I think integral to Zen.
Also I think AMD also warn they now have more reliance upon GF - I assume with the movement of the team and their assets, which has implications if any delays-issues.
Cheers

I don't see why it would affect MCMs (and they still do hold some stakes on there even though majority was sold)
And as things get more complex, packaging isn't that simple anymore anyway, just look at Fiji, it took originally 3 companies to put the pieces together, interposer with some work done from UMC, interposer finalizing and actually assembling Fiji & HBMs on it by Amkor and ASE

 

[silent_guy]

Well my original post was exactly this point and AMD moving 1700 staff to the new joint venture and how it may impact future R&D-project implementations, where you responded they no longer need the staff.

When I say that they still have some people in-house, I mean one or two. They don't do any real R&D work, but they write requirement specs and closely follow up with the partner to make sure AMD will get what they requested.

I don't see that as a contradiction between earlier statements about them not needing a packaging department.

 

[CSI PC]

When I say that they still have some people in-house, I mean one or two. They don't do any real R&D work, but they write requirement specs and closely follow up with the partner to make sure AMD will get what they requested.

I don't see that as a contradiction between earlier statements about them not needing a packaging department.

OK but why did they need a packaging department even before this announcement when they work closely with Global Foundry?
What have they been doing that could not be done since they sold off GF?
AMD has been fabless for a long time or was it only partial?

Sorry for another post,
just to emphasis my focus was specifically Multi-Chip Modules and stacked dies, rather than the whole lot in terms of fabless packaging department.
So we assume the new joint venture is doing all of this as well for their current operations and business/done with GF/mix of both?
Cheers

 

[3dilettante]

I recall some earlier statements that AMD contracted business there several years out, which should cover the projects and products already in the pipeline. Presumably it can renegotiate or shop for other business at the end of that, but it's in a notable end of decade period where a lot of things are coming due or ending for AMD.