AMD and Samsung Announce Strategic Partnership in Mobile IP

[entity279]

I know @Nebuchadnezzar knows a whole lot more than us mortals regarding this market, but it really doesn't look like this is a case where Samsung negotiated with AMD a patenting deal 3 years ago and it's announcing it only now.

Well one plausible explanation for the cloak and dagger wording in the anouncement would be trying to not get sued by QC . (as @Gubbi helpfully hinted at)

If as per Anand piece, Samsung has been working on a GPU for 7 years.. It's very hard to believe that Samsung would have thrown all that work away (unless ofc they are going through a heavy corporate restructuring).

 

[Bondrewd]

It's very hard to believe that Samsung would have thrown all that work away (unless ofc they are going through a heavy corporate restructuring).

The current one was like their 3rd try.

 

[Nebuchadnezzar]

In 2018 alone Samsung SLSI launched one SoC with a Mali G71 MP1, one with a Mali G71-MP2, one with a Mali G72 MP3, one with a Mali G72 MP18 and a supposed die-shrink with a Mali T720.
In their ARMv7 days, they'd release SoCs with Mali and PowerVR GPUs in parallel.

The team adapting AMD's Radeon IP GPU to their SoC could be a different team than the one working on their in-house GPU.

That's a stupid comparison. You're comparing delivered RTL to actual architecture/microarchitecture development. AMD isn't delivering RTL to Samsung, it's an architecture license.

 

[anexanhume]

Samsung is pushing hard in the Automtive industry. One of the more common complaints however was the GPU on their SoCs is a bit behind when compared to the NVIDIA Xavier and Parker.
You have to consider that in Automotive the power envelope while generally not so high as desktop is much higher than smartphone and tablets.
One of the things that immediately comes to my mind is the integration of compute oriented architecture of AMD's Radeon in Samsung's Exynos Auto.
The requirement for compute power in Automotive is growing exponentially due to ADAS and Autonomous vehicles running a wide variety of inferencing and DNN algorithms and NVIDIA is quite strong in this area with their Tegra SoCs.
And it would not make sense for Samsung to license it from the competitor NVIDIA who is making Tegra SoCs for the GPU.
Couple this to the fact that AMD has a strong open source drivers and a budding compute stack it could be something to watch out.
You can guess who is the only closed source GPU vendor. Intel has massive open source components and DNN/compute stack as well.

They could make a play for future Nintendo hardware too.

 

[Deleted member 13524]

That's a stupid comparison. You're comparing delivered RTL to actual architecture/microarchitecture development. AMD isn't delivering RTL to Samsung, it's an architecture license.

You're 100% sure AMD isn't selling a semi-custom GPU block for Samsung to use in a future SoC, akin to ARM selling Mali blocks?

 

[Nebuchadnezzar]

You're 100% sure AMD isn't selling a semi-custom GPU block for Samsung to use in a future SoC, akin to ARM selling Mali blocks?

Yes. Seems I have to speak more bluntly: I've had confirmation it's an actual architecture license, hence why I said it looks to be like how Apple is using IMG's license.

AMD wouldn't have had the resources to develop their own mobile targeted GPU on their own. It's Samsung's responsibility to implement it into a working GPU microarchitecture, and it appears that's what they've been doing for 2-3 years.

 

[Frenetic Pony]

Yes. Seems I have to speak more bluntly: I've had confirmation it's an actual architecture license, hence why I said it looks to be like how Apple is using IMG's license.

AMD wouldn't have had the resources to develop their own mobile targeted GPU on their own. It's Samsung's responsibility to implement it into a working GPU microarchitecture, and it appears that's what they've been doing for 2-3 years.

Nor would AMD have had the desire to do so, Imagination was the only big, independent GPU licensor in the mobile industry and look what happened to them. Everything else is ARM or vertically integrated.

Still... I mean good for AMD one supposes. More revenue, without a huge amount of risk. I wonder what this GPU will look like, and whether it might even fall under Navi arch at all.

 

[Jawed]

I suppose this means there should be a mobile RDNA thread.

If it's a tiler and leans heavily on primitive shading that makes it good for mobile?...

 

[BRiT]

Well at least we have confirmation of who knows stuff and who just wants to argue to argue...

 

[itsmydamnation]

Yes. Seems I have to speak more bluntly: I've had confirmation it's an actual architecture license, hence why I said it looks to be like how Apple is using IMG's license.

AMD wouldn't have had the resources to develop their own mobile targeted GPU on their own. It's Samsung's responsibility to implement it into a working GPU microarchitecture, and it appears that's what they've been doing for 2-3 years.

Do you know much much of a mobile architecture is delivered by AMD for Samsung to modify/implement or is it a case of here are all the IP blocks as AMD have designed them for AMD targets and do with them what you wish?

 

[Nebuchadnezzar]

Do you know much much of a mobile architecture is delivered by AMD for Samsung to modify/implement or is it a case of here are all the IP blocks as AMD have designed them for AMD targets and do with them what you wish?

It's architectural license to modify/implement for mobile.

It increasingly looks like this is indeed a new deal and SGPU might have been halted, some are saying patent issues with Nvidia.

 

[Silent_Buddha]

One other thing to consider is Microsoft's increased interest in Windows on ARM with Samsung as a key partner. Having GPU IP that's already used in Windows might make driver development easier and help with the work of Windows on ARM as well as making sure the GPU feature levels are up to date.

While the wording is specifically for mobile, I believe some of the chips that Microsoft is using for Windows on ARM are Samsung mobile chips.

Regards,
SB

 

[Entropy]

One other thing to consider is Microsoft's increased interest in Windows on ARM with Samsung as a key partner. Having GPU IP that's already used in Windows might make driver development easier and help with the work of Windows on ARM as well as making sure the GPU feature levels are up to date.

While the wording is specifically for mobile, I believe some of the chips that Microsoft is using for Windows on ARM are Samsung mobile chips.

This will be interesting to see play out in mobile space alone, but if Windows on ARM gains a stronger foothold, comparisons against both Qualcomm and x86 offerings will be intriguing. Implementation is everything of course, but Samsung has some serious engineering and foundry muscle, if they choose to apply it.

 

[CarstenS]

It's architectural license to modify/implement for mobile.

It increasingly looks like this is indeed a new deal and SGPU might have been halted, some are saying patent issues with Nvidia.

Wouldn't an architecture license not also cover at least some patent issues, Samsung had in the past with Nvidia? Maybe their mini-cross license was about to expire and Samsung wanted to cover their own GPU-IP?

 

[Nebuchadnezzar]

Wouldn't an architecture license not also cover at least some patent issues, Samsung had in the past with Nvidia? Maybe their mini-cross license was about to expire and Samsung wanted to cover their own GPU-IP?

Yes it would. Essentially this is AMD giving access to their patents via an architectural license.

Some Koreans are saying Nvidia didn't want Samsung to make their own GPU because they wanted Samsung to license from them.

The thing is that this indeed looks like going forward with AMD instead of their SGPU. Samsung's GPU VP seems to have quit following the events, so the thing stinks a bit.

 

[Kaotik]

There were rumors in 2016 that Samsung was looking to license AMD gfx IP. The old Polaris/Vega/Navi-roadmap already had "Scalability" marked for Navi. How far reaching would it be to think that the rumors were in fact real and things were already happening in the background back then and that "Scalability" was referring to Samsung making mobile GPU based on Navi?

 

[Nebuchadnezzar]

There's some pointers from Samsung's existing GPU people that what they were working on wasn't AMD related. The whole thing is thoroughly confusing especially since they had a very large GPU hiring spree in the first half of 2018.

 

[Deleted member 13524]

Yes. Seems I have to speak more bluntly: I've had confirmation it's an actual architecture license, hence why I said it looks to be like how Apple is using IMG's license.

Ok thanks for the clarification.

It's architectural license to modify/implement for mobile.

Do you know if the deal includes AMD being able to use Samsung's research on ULP GPUs for their own advantage, say for a 5-15W APU down the road to tackle Intel's Y series?

Some Koreans are saying Nvidia didn't want Samsung to make their own GPU because they wanted Samsung to license from them.

Makes sense...

The thing is that this indeed looks like going forward with AMD instead of their SGPU. Samsung's GPU VP seems to have quit following the events, so the thing stinks a bit.

So isn't that different from what apple is currently doing with IMG?
At least from a marketing perspective, it looks like Samsung will have to use AMD's branding (RDNA, Radeon) whereas Apple dropped PowerVR altogether.

 

[Nebuchadnezzar]

Do you know if the deal includes AMD being able to use Samsung's research on ULP GPUs for their own advantage, say for a 5-15W APU down the road to tackle Intel's Y series?

That's the big question, absolutely no idea but I don't see why they wouldn't be able to.

So isn't that different from what apple is currently doing with IMG?
At least from a marketing perspective, it looks like Samsung will have to use AMD's branding (RDNA, Radeon) whereas Apple dropped PowerVR altogether.

Apple didn't drop PowerVR, it's still very much a TBDR with PVRTC.

 

[Deleted member 13524]

Apple didn't drop PowerVR, it's still very much a TBDR with PVRTC.

I said from a marketing perspective. Apple dropped the PowerVR brand name, Samsung will seemingly have to call their iGPU a Radeon that uses RDNA.

This definitely makes a huge difference in brand value boost between IMG's and AMD's case, for what could be essentially a similar deal in IP licensing.
AMD's stock soared after they announced the deal. IMG's stock collapsed by 60% when apple announced they weren't "using" PowerVR anymore.