It's Come Back Up That Intel Is Reportedly Licensing Radeon Graphics IP

[Zucca]

No no no no no... This is all about patents. As messed up as such a fact is, in order to be in the GPU business at all, you must first get a patent license from one of the big GPU companies. If you don't then you're going to be sued and lose, it's as simple as that. Unless you can magically come up with a radical new GPU design that just happens to avoid all the patents and be performant/capable at the same time, good luck with that. This does not mean you're getting expertise or designs from the company beyond what exists in the patent filing, this does not mean that Intel had Nvidia GPUs onboard and now will have AMD ones. This is simply not the case. That is not to say that Intel couldn't get a license to have Geforce or Radeon gpus on their APUs but that's not what this is about.

No, you are completely wrong.
Intel can't choose to license the patents from different parties, it doesn't work that way. Intel's GPU design is their own, everyone knows it has no relation to Nvidia's own design, even after many years of licensing. The license deal is about implementing the GPU features required for API support, that's it.

No. But thanks for trying.

Now I'm actually interested how the reality is.

We'd need someone (preferably from intel) who was actually making the deal or soemone who is designing the GPU side on intel chips to confirm this.
Since there seems to be no one here in that position, next best thing would be to show some trusted sources... Maybe some article/blog post from an intel employee...

It's fruitless to continue the "No. You're the one who's wrong." -talk.

[droidhacker]

No, you are completely wrong.
Intel can't choose to license the patents from different parties, it doesn't work that way. Intel's GPU design is their own, everyone knows it has no relation to Nvidia's own design, even after many years of licensing. The license deal is about implementing the GPU features required for API support, that's it.

No. But thanks for trying.

[droidhacker]

<i>I would be very surprised if Intel engineers don't get "inspired" by some of AMD's tech now that they have full access to the patent portfolio.</i>

if that's true, why can't nvidia sue them for previous inspirations from nvidia patents? The fact that it is possible to switch patent protection providers implies that patents don't actually say anything in particular.

Assuming that they WERE inspired by nvidia patents (that's what you are contending), then they were fully authorized to be so inspired under the terms of their agreement. The fact that their agreement is terminated doesn't mean that they are no longer protected from NVidia's lawyers. The agreement will certainly have a termination clause written to the effect of any IP obtained from nvidia under the terms of the agreement, is protected in perpetuity.

But I will tell you how the real deal ACTUALLY works;
If, for example, ARM (think mali) has some agreement in place with nvidia, or Qualcomm (adreno) has some agreement in place with AMD (after all, qualcomm obtained adreno FROM AMD), then that agreement is extended to cover Intel. This protects Intel against ARM or Qualcomm or whatever, because they have a portfolio of patents that they can use to fight back.

The REAL STORY is that they ALL are infringing on EACH OTHER. This is one of the reasons why they are so scared of going open source -- not to protect their IP, but to HIDE it.

So if Nvidia goes after Qualcomm, Qualcomm goes right back at Nvidia with equal and opposite force, and the result is a stalemate.
If Intel isn't protected, then ARM goes after Intel, and Intel CAN'T FIGHT BACK, so ARM wins and Intel is out millions.


As far as inspiration goes, the patents aren't secrets. They're all filed at the USPTO. If they want to be inspired, they just need to send some nerds in there for a few days to read up on everyone else's IP. If they see something they like, they reach out to their competitor and make a deal to license that patent. Or they steal it and keep their drivers closed source.

[artivision]

Nvidia GPU 16nm HMP cut (2GHZp) =20Wp/Tflop. AMD GPU not HMP (2GHZp) or LP (1.6GHZp), but 14nm ULP cut (1.25GHZp) inside APU and HBM2 =10Wp/Tflop. AMD CPU 14nm HMP cut (3.5GHZ+p) =8-9Wp/core. So 4core 10Tflops and at least 16GB HBM2 at 135Wp standalone card with internal and external slots like SATA, fat single cooler for desktop and slim dual for laptop, no motherboard 400$p. 3core 5Tflops and 12GB HMB2 at 75Wp the same as before 230$p. 2core 2.5Tflops and 8GB HMB2 at 40Wp single slim only, 130$p. A server APU 8core 16GB HMB2 1Tflop at 80Wp dual slim only, 230$p. Also it is time to start thinking a Softcore upon a GPU. 10nm MIMD GPUs will hit 2.5GHZ and most things can be dual issued inside. So follow paradigms like EDGE and WAVESCALAR. In HPC market, scalar is a software driver that uses unified mini cores for it's calculations inside custom chips, cache management is also an app, you and Intel have those still hardcoded.

[vein]

Horrible why? they work better than AMD APU less power hungry better 2d, it's simply for desktop usage ,web/office, light light gaming, and decoding/encoding. With a laptop I only use the nvidia card when I need more 3d power (some game for example).

hmm...it was a while ago you used an AMD APU right? Comparing the one I have (A10-8700P) and the intel cpu I have (i7-6600U), the AMD APU is way better in 2d. Everything just flows better on it...especially firefox seems a lot more fluent.

[efikkan]

No no no no no... This is all about patents. As messed up as such a fact is, in order to be in the GPU business at all, you must first get a patent license from one of the big GPU companies. If you don't then you're going to be sued and lose, it's as simple as that. Unless you can magically come up with a radical new GPU design that just happens to avoid all the patents and be performant/capable at the same time, good luck with that. This does not mean you're getting expertise or designs from the company beyond what exists in the patent filing, this does not mean that Intel had Nvidia GPUs onboard and now will have AMD ones. This is simply not the case. That is not to say that Intel couldn't get a license to have Geforce or Radeon gpus on their APUs but that's not what this is about.

No, you are completely wrong.
Intel can't choose to license the patents from different parties, it doesn't work that way. Intel's GPU design is their own, everyone knows it has no relation to Nvidia's own design, even after many years of licensing. The license deal is about implementing the GPU features required for API support, that's it.

[Zucca]

...by some of AMD's tech now that they have full access to the patent portfolio.

”now that they have full access”

1) Are we already sure they already have made the deal?
2) Is it really a full access?

[Luke_Wolf]

It will mean they get AMD's assistance and access to some GPU features. They do this in order to save resources implementing the hardware required for new Direct3D/OpenGL/Vulkan features.


This deal is not about licensing patents, it's not like Intel can choose to license the same patents from two different parties.


Intel had the same deal with Nvidia for many years, and even though Nvidia have superior GPU architectures and are the largest experts in GPU technology, it still didn't help Intel's horrible GPUs. And now that they have the same deal with AMD, Intel's GPUs are not suddenly going to improve radically.

No no no no no... This is all about patents. As messed up as such a fact is, in order to be in the GPU business at all, you must first get a patent license from one of the big GPU companies. If you don't then you're going to be sued and lose, it's as simple as that. Unless you can magically come up with a radical new GPU design that just happens to avoid all the patents and be performant/capable at the same time, good luck with that. This does not mean you're getting expertise or designs from the company beyond what exists in the patent filing, this does not mean that Intel had Nvidia GPUs onboard and now will have AMD ones. This is simply not the case. That is not to say that Intel couldn't get a license to have Geforce or Radeon gpus on their APUs but that's not what this is about.

[efikkan]

Does it mean Intel will use AMD's architecture for the GPU, or just some of their hardware features?

It will mean they get AMD's assistance and access to some GPU features. They do this in order to save resources implementing the hardware required for new Direct3D/OpenGL/Vulkan features.

<i>I would be very surprised if Intel engineers don't get "inspired" by some of AMD's tech now that they have full access to the patent portfolio.</i>

if that's true, why can't nvidia sue them for previous inspirations from nvidia patents? The fact that it is possible to switch patent protection providers implies that patents don't actually say anything in particular.

This deal is not about licensing patents, it's not like Intel can choose to license the same patents from two different parties.

If there is one thing everyone can agree on is that Intel graphics are horrible. I could see a deal allowing Intel to use Radeon graphics in their chips as long as it specifies that GPU performance cannot exceed a specified % compared to AMD APU's. 60% performance of Raven ridge graphics would be a huge step up from Intel's iGPU performance, and it wouldn't damage Raven Ridge's sales.

Intel had the same deal with Nvidia for many years, and even though Nvidia have superior GPU architectures and are the largest experts in GPU technology, it still didn't help Intel's horrible GPUs. And now that they have the same deal with AMD, Intel's GPUs are not suddenly going to improve radically.

[andre30correia]

If there is one thing everyone can agree on is that Intel graphics are horrible. I could see a deal allowing Intel to use Radeon graphics in their chips as long as it specifies that GPU performance cannot exceed a specified % compared to AMD APU's. 60% performance of Raven ridge graphics would be a huge step up from Intel's iGPU performance, and it wouldn't damage Raven Ridge's sales.

AMD gets a bit of $ in return, Intel gets to cut it's iGPU division. It makes sense when you consider the fact that AMD will have an APU with Ryzen/Vega. Intel's APU sales will be pretty much gutted.

That probably isn't what the deal is about, but I wouldn't be shocked if it is. Intel is smart enough to realize what AMD getting the next 2 major consoles as well means. AMD's driver optimizations will be very good in the future, which is another area Intel could cut out. The deal might save Intel $, give AMD $, and result in a better ecosystem all around while not costing AMD anything. In fact it would benefit both AMD and Intel.

Horrible why? they work better than AMD APU less power hungry better 2d, it's simply for desktop usage ,web/office, light light gaming, and decoding/encoding. With a laptop I only use the nvidia card when I need more 3d power (some game for example).