PowerVR ray-tracing hardware

[Timmah!]

https://imgtec.com/news/press-release/otoy-imagination-unveil-breakthrough-powervr-ray-tracing-hardware-platform-cinematic-real-time-rendering/

So what is this and how do you think this is going to affect the GPU market and current Nvidia dominance, once its released?

 

[dark zero]

Nothing, but in Mobile they will get finally an edge, something that nVIDIA failed HARD.

 

[ThatBuzzkiller]

It's hardware accelerated ray tracing and it'll mean nothing in the GPU market when NO GAMES support it ...

There's more to rendering than a bunch of ray intersection tests too like building your acceleration structures for dynamic scenes, BSDF sampling, multiple bounces of indirect illumination, anti-aliasing and simulating pathological materials along with other extensions including participating media ...

I doubt whatever PowerVR has are as good as full fledged discrete GPUs at half of the above which are important for physically accurate rendering and they only advertise efficiency rather than claiming the performance crown in it's own special case ...

 

[Timmah!]

It's hardware accelerated ray tracing and it'll mean nothing in the GPU market when NO GAMES support it ...

There's more to rendering than a bunch of ray intersection tests too like building your acceleration structures for dynamic scenes, BSDF sampling, multiple bounces of indirect illumination, anti-aliasing and simulating pathological materials along with other extensions including participating media ...

I doubt whatever PowerVR has are as good as full fledged discrete GPUs at half of the above which are important for physically accurate rendering and they only advertise efficiency rather than claiming the performance crown in it's own special case ...

The article says this>

“It is a major reason why OTOY is merging Brigade and OctaneRender 4 next year, which will bring real time cinematic path tracing to both Unity and Unreal Engine. We feel this is a huge milestone in helping the games industry move beyond legacy rasterization frameworks built on OpenGL, Vulkan or DirectX."

I dont know, but i read this as Octane is being built into Unreal/Unity in near future and PowerVR has hardware, which can run it 10x more efficient than current Nvidia offering, which Octane 3 runs on. So i would not dismiss it so easily. You never know, perhaps thanks to this they could become big player in the GPU industry in the future.

 

[ThatBuzzkiller]

The article says this>

“It is a major reason why OTOY is merging Brigade and OctaneRender 4 next year, which will bring real time cinematic path tracing to both Unity and Unreal Engine. We feel this is a huge milestone in helping the games industry move beyond legacy rasterization frameworks built on OpenGL, Vulkan or DirectX."

I dont know, but i read this as Octane is being built into Unreal/Unity in near future and PowerVR has hardware, which can run it 10x more efficient than current Nvidia offering, which Octane 3 runs on. So i would not dismiss it so easily. You never know, perhaps thanks to this they could become big player in the GPU industry in the future.

We can already do real time path tracing, the question is can we reduce the noise or error to an acceptable level ? And Imagination Technologies has yet to demonstrate their hardware with indirect DIFFUSE lighting otherwise they wouldn't be advertising the stripped down ray tracing instead of fully physically based global illumination ...

Imagination Technologies becoming a big player ? LOL, I wonder what major features they've standardized in gfx APIs in the last decade ...

Even if they do have hardware extensions for accelerating ray intersection tests they don't even have a say in the Direct3D spec ...

 

[SolMiester]

Nothing, but in Mobile they will get finally an edge, something that nVIDIA failed HARD.

Eh?, NV own's the mobile segment?, where have you been?

 

[Timmah!]

We can already do real time path tracing, the question is can we reduce the noise or error to an acceptable level ? And Imagination Technologies has yet to demonstrate their hardware with indirect DIFFUSE lighting otherwise they wouldn't be advertising the stripped down ray tracing instead of fully physically based global illumination ...

Imagination Technologies becoming a big player ? LOL, I wonder what major features they've standardized in gfx APIs in the last decade ...

Even if they do have hardware extensions for accelerating ray intersection tests they don't even have a say in the Direct3D spec ...

Where do you get this "stripped down raytracing" thing? I may have failed to quite understand what they say, but once again, as i read it, IMO they just report they developed hardware capable of running Octane Render at 10x higher speed at same wattage levels compared to how it runs currently.

Meanwhile, as an Octane Render user myself since cca 2009, i am kinda positive fully physically based global illumination is one of the its key features. At least i think thats what its "pathtracing" and "PMC" kernels are all about?

 

[ThatBuzzkiller]

Where do you get this "stripped down raytracing" thing? I may have failed to quite understand what they say, but once again, as i read it, IMO they just report they developed hardware capable of running Octane Render at 10x higher speed at same wattage levels compared to how it runs currently.

Meanwhile, as an Octane Render user myself since cca 2009, i am kinda positive fully physically based global illumination is one of the its key features. At least i think thats what its "pathtracing" and "PMC" kernels are all about?

They may have accelerated ray intersection test in the Octane Render but you should ask yourself this ...

Why are they only demoing indirect specular and transparency when a good renderer accounts for many types of materials like diffuse surfaces, anisotropy, or other exotic materials that exhibit some sort of inscattering like snow, skin, and even soap ...

Life is far more than just indirect specular, transparency, and analytically correct hard shadows ...

The reason I have yet to believe that Imagination Technologies holds the true panacea for physically accurate rendering is that their trying to boast their IPs as best as they can by avoiding other very complex light transport scenarios ...

Hence why Imagination is advertising "ray-tracing" (inferior light transport method) compared to "approximating the rendering equation" or "photo-realistic global illumination" or "it can match a path tracer with low noise in real time", OK I'm bad at making up taglines but you get the idea ...

 

[antihelten]

Even if they do have hardware extensions for accelerating ray intersection tests they don't even have a say in the Direct3D spec ...

I don't think they care about D3D, their target here is almost certainly Vulkan (Imagination is also a member of Khronos).

Imagination Tech is still a mobile GPU firm first and foremost and in that segment it's very much OpenGL ES and Vulkan.

Their previous ray tracing implementation was done with OpenGL ES:
http://www.anandtech.com/show/10139...trates-powervr-vulkan-sdk-powervr-ray-tracing

 

[ThatBuzzkiller]

I don't think they care about D3D, their target here is almost certainly Vulkan (Imagination is also a member of Khronos).

Imagination Tech is still a mobile GPU firm first and foremost and in that segment it's very much OpenGL ES and Vulkan.

Their previous ray tracing implementation was done with OpenGL ES:
http://www.anandtech.com/show/10139...trates-powervr-vulkan-sdk-powervr-ray-tracing

Then that only means they have no effect on on the direction of AAA development or standardized hardware features which means they'll never become the so called "big player" ...

Khronos Group and the architecture review board has pretty much always lagged behind DirectX in terms of support and in some respects API design ...

What Microsoft demands about the D3D spec means that Khronos Group must follow suit or they risk alienating the entire industry like they once did with OpenGL ...

 

[3DVagabond]

Then that only means they have no effect on on the direction of AAA development or standardized hardware features which means they'll never become the so called "big player" ...

Khronos Group and the architecture review board has pretty much always lagged behind DirectX in terms of support and in some respects API design ...

What Microsoft demands about the D3D spec means that Khronos Group must follow suit or they risk alienating the entire industry like they once did with OpenGL ...

Why are you so negative about this? You act like there's no potential at all. Just embrace the tech.

 

[antihelten]

Then that only means they have no effect on on the direction of AAA development or standardized hardware features which means they'll never become the so called "big player" ...

Khronos Group and the architecture review board has pretty much always lagged behind DirectX in terms of support and in some respects API design ...

What Microsoft demands about the D3D spec means that Khronos Group must follow suit or they risk alienating the entire industry like they once did with OpenGL ...

I honestly don't think Imagination is trying to become a "big player" in the first place though.

To me this just looks like Imagination Tech trying to stand out from the crowd in the mobile graphics market. With Intel dropping out of the mobile market, that leaves just Apple and Mediatek as notable Imagination customers. Apple taking their GPU internal in the future is probably not a completely implausible scenario, so Imagination Tech will want to have some way of at least appearing to offer something unique, to avoid becoming completely irrelevant in this market (I doubt they will be able to maintain the revenue required to develop competitive GPU designs with just Mediatek as a customer).

With that being said though I do think that hardware accelerated ray tracing will become the standard at some point, but it will most likely be driven by AMD and Nvidia.

Edit: Actually, now that I think about it, this could be a play to get out of the mobile market, but not to get into the gaming market (at least not primarily so), but rather to get into the professional graphics market. This market would be willing to pay a couple of thousand dollars for a ray tracing co-processor in a volume that could possibly make sense for a company the size of Imagination Tech. This would however mean that Imagination Tech would have to start selling hardware instead of just licensing designs.

 

[ThatBuzzkiller]

Why are you so negative about this? You act like there's no potential at all. Just embrace the tech.

There's only potential to be found in ubiquitous hardware support and I maybe playing devil's advocate but in no way do I deny the inherent advantage Imagination has with their IP ...

It's just the matter of the reality that hardware without ISV support will go the way of the dodo ...

I don't think there are enough advantages for other IHVs to adopt this IP when you still have to build your acceleration structure and it doesn't solve the shading cost ...

I honestly don't think Imagination is trying to become a "big player" in the first place though.

To me this just looks like Imagination Tech trying to stand out from the crowd in the mobile graphics market. With Intel dropping out of the mobile market, that leaves just Apple and Mediatek as notable Imagination customers. Apple taking their GPU internal in the future is probably not a completely implausible scenario, so Imagination Tech will want to have some way of at least appearing to offer something unique, to avoid becoming completely irrelevant in this market (I doubt they will be able to maintain the revenue required to develop competitive GPU designs with just Mediatek as a customer).

With that being said though I do think that hardware accelerated ray tracing will become the standard at some point, but it will most likely be driven by AMD and Nvidia.

Edit: Actually, now that I think about it, this could be a play to get out of the mobile market, but not to get into the gaming market (at least not primarily so), but rather to get into the professional graphics market. This market would be willing to pay a couple of thousand dollars for a ray tracing co-processor in a volume that could possibly make sense for a company the size of Imagination Tech. This would however mean that Imagination Tech would have to start selling hardware instead of just licensing designs.

AMD or Nvidia could drive hardware accelerated ray tracing once we've reached the end of transistor technology for a while but let's talk about now and the immediate future instead since that's a long way off ...

The problem of them breaking into professional graphics is that their software is not mature enough and I don't think their software team could deliver fast enough when we take a look at AMD who is much more resourceful but struggles as it is when competing with Nvidia in that segment but what's more is that they don't have any recent experience in building big dies when that is going to make it harder to capture the performance crown thus they'll be forced lower their prices leading to a deflationary spiral ...

 

[antihelten]

The problem of them breaking into professional graphics is that their software is not mature enough and I don't think their software team could deliver fast enough when we take a look at AMD who is much more resourceful but struggles as it is when competing with Nvidia in that segment but what's more is that they don't have any recent experience in building big dies when that is going to make it harder to capture the performance crown thus they'll be forced lower their prices leading to a deflationary spiral ...

I doubt Imagination could handle the software side either. They would basically have to do it through partnerships with the various developers of renderers out there. Whether or not they could convince any of the big ones to do this is anyone's guess (although I doubt they could do it with what they have shown so far, they would likely need a more robust hardware solution before anyone would even consider jumping aboard).