AMD relaunches GPUOpen

[shankly1985]

Nicolas Thibieroz


A little over four years ago, I remarked that it was “time to open up the GPU”. In that time, GPUOpen has done just that, passing some amazing milestones.



We’ve launched fantastic new developer tools such as Radeon GPU Profiler – a ground-breaking graphics profiler that provides unprecedented in-depth access to GPU execution information that developers can harness to optimize their titles. In the same vein we are releasing this week Radeon Memory Visualizer, a tool that reveals the memory mapping of your game in truly novel ways for the PC platform.

We’ve seen the introduction -and today the expansion- of the FidelityFX family: a collection of highly-optimized, truly open-source effects that developers can easily integrate into their titles, or inspire themselves from when writing their own implementations.

The Radeon™ GPU Profiler

Truly open source shader-based effects
We published a variety of SDKs aimed at making developers’ lives easier by providing code samples or libraries educating them about new API features, or enabling convenient access to useful functionalities such as our widely-integrated Vulkan and D3D12 Memory Allocator libraries.

Our focus on providing documentation has not faltered, with ISA and programming guides for a variety of Radeon GPU architectures released over the last four years. Combined with over 50 tutorials, including guest posts from renowned game developers, these form the basis of the GPUOpen philosophy: opening up the GPU for developers to access the knowledge and functionality needed to foster innovation and constantly improve graphics performance.



All this and much, much more.



But the game industry – and the wider rendering community – has changed a lot in the last four years. Explicit APIs such as DirectX®12 and Vulkan® are now widely supported in games, and with them come a new set of exciting programming paradigms, language features and optimization opportunities. Change and innovation is what keeps the games industry – and the wider tech industry – so exciting. With so much change, it’s only right that we change too.



That is why today, I am delighted to announce that GPUOpen is reborn. We’re retaining the same core commitment to put you – the developer – at the heart of everything we do, but we’re doubling down on everything that made the original site so great. Fantastic tools, cutting-edge open-source effects and high-quality software components to allow you to get the most out of the hardware: GPU of course, but also CPU content to help you drive graphics rendering in the most efficient way possible. We’ve overhauled the site from top to tail. Streamlining it, to make it easier to find what you need to build legendary games.

A symbol of our commitment to openness
This week we’ll be celebrating the rebirth of GPUOpen by releasing new content, and a new official @GPUOpen Twitter handle that you may want to follow for regular updates and developer topics. We’ll conclude relaunch celebrations on Friday, with our online “Let’s build…” event where several AMD speakers will present a variety of topics relevant to graphics development and optimizations.

I’d like to thank all developers for the huge levels of support you have given us in the past four years. The GPUOpen team is delighted to be able to continue to serve the development community, and we’re all looking forward to welcoming you back to the site.

Nicolas Thibieroz
Nicolas Thibieroz is the Director of Worldwide Game Engineering at AMD. Over the years, he and his team have helped countless numbers of PC game developers tame the GPU to make faster and better-looking games. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.


https://gpuopen.com/lets-build-the-future/

https://twitter.com/GPUOpen

 

[shankly1985]

Create wonder. No black boxes
FidelityFX Contrast Adaptive Sharpening (CAS) provides a mixed ability to sharpen and optionally scale an image.
Contrast Adaptive Sharpening (CAS) provides a mixed ability to sharpen and optionally scale an image. The algorithm adjusts the amount of sharpening per pixel to target an even level of sharpness across the image. Areas of the input image that are already sharp are sharpened less, while areas that lack detail are sharpened more. This allows for higher overall natural visual sharpness with fewer artifacts.

CAS was designed to help increase the quality of existing Temporal Anti-Aliasing (TAA) solutions. TAA often introduces a variable amount of blur due to temporal feedback. The adaptive sharpening provided by CAS is ideal to restore detail in images produced after TAA .

CAS’ optional scaling capability is designed to support Dynamic Resolution Scaling (DRS). DRS changes render resolution every frame, which requires scaling prior to compositing the fixed-resolution User Interface (UI). CAS supports both up-sampling and down-sampling in the same single pass that applies sharpening.

COMPARISON

High-fidelity reflections in your scene, without costing the earth. SSSR uses your rendered frame to create brilliant reflections.
The term “Screen Space Reflections” describes an effect that can create realistic looking reflections purely based on information already present in the rendered image. The underlying algorithm shoots reflection rays from a depth imprint of the rendered scene (a so called “depth buffer”) and follows them in constant steps through the image until these rays intersect with the depth buffer again. Instead of taking constant sized steps, FidelityFX SSSR builds on an industry-leading algorithm that searches the rendered image in a hierarchical manner. This allows for larger and fewer steps on average, increasing performance of the search and quality of the final image.

To support glossy reflections, this FidelityFX effect jitters the reflection rays to create the sense of surfaces with varying roughness. However, that approach inherently introduces noise. Thus, FidelityFX SSSR comes with a high quality denoiser specifically optimized for the RDNA architecture. The denoiser combines the results from multiple frames to create a noise-free image. Furthermore, it allows to decrease the ray count based on surface roughness thus further speeding up screen space traversal.

COMPARISON

FidelityFX LPM provides an open source library to easily integrate HDR and wide gamut tone and gamut mapping into your game.

COMPARISON

FidelityFX Single Pass Downsampler (SPD) provides an RDNA-optimized solution for generating up to 12 MIP levels of a texture.
MIP levels have versions of the same texture but in smaller resolutions. They are used when the high-resolution texture is not necessarily needed (such as when objects are far from the camera, covering only a few pixels) or might introduce aliasing artefacts. MIP levels are also commonly used in effects like Bloom, Screen Space Reflections, and many more.

Use FidelityFX SPD as a building block to accelerate your post processing pipeline or texture creation.

More key features:

• 
  
  
  • Generates up to 12 mip levels (maximum source texture size is 4096x4096).
  • Single function call.
  • User defined 2x2 reduction function.
  • User controlled border handling.
  • Supports various image formats.
  • HLSL and GLSL versions available.
  • Rapid Packed Math support.
  • Uses optionally subgroup operations / SM6+ wave operations, which can provide faster performance.

FidelityFX Combined Adaptive Compute Ambient Occlusion (CACAO) is an RDNA-optimized implementation of ambient occlusion.

In 3D graphics, one of the biggest challenges is creating high quality lighting. Calculating physically accurate lighting is too computationally expensive to be done in real time, so instead multiple approximate models are used to create believable lighting.

One of these models is ambient lighting, which models indirect light in a scene. The most naïve ambient light model is to have ambient lighting constant across an entire scene. However, this can be greatly improved by ambient occlusion, in which there is less ambient light in areas with occluding geometry – for example in corners, where objects meet, nooks and crannies etc. This vastly improves believability of a scene and makes a scene easier to visually parse.

Computing ambient occlusion can be done in multiple ways. In static scenes ambient occlusion may be computed ahead of time. When this is done, the methods used are typically very expensive and take a long time to compute. However, in dynamic scenes in which content is not predictable ahead of time this is not an option. In this case the state of the art is to use screen space ambient occlusion, which computes the ambient occlusion each frame based on geometry rendered. When this approach is taken an implementation must be chosen which is a good balance of speed and quality.

The CACAO ambient occlusion implementation is a highly optimized implementation of ambient occlusion. It may be run at multiple different quality settings, allowing it to meet multiple different requirements for quality performance trade-offs, and be run across a wide range of hardware.

COMPARISON

 

[Wrinkly]

NM!

 

[Poneros]

The four new open-source FidelityFX features available today are:

• FidelityFX Screen Space Reflections: Our implementation of Stochastic Screen Space Reflections (SSSR) delivers high-quality reflections with minimal overhead, via an optimized compute shader and RDNA architecture-optimized denoising.
• FidelityFX Ambient Occlusion: Our RDNA architecture-optimized Combined Adaptive Compute Ambient Occlusion (CACAO) effect dynamically and efficiently helps improve the appearance of objects based their exposure to ambient light.
• FidelityFX HDR Mapper: Optimized for use with AMD FreeSync Premium Pro1 displays, AMD’s Luminance Preserving Mapper (LPM) delivers superior HDR and wide color gamut content for games.
• FidelityFX Downsampler: Single-pass and optimized for RDNA architecture, AMD’s compute shader-based Single Pass Downsampler (SPD) generates texture MIP levels using asynchronous compute for optimal performance.

Next Level Physics and Hair Effects Demos

FEMFX is our open-source CPU library for deformable material physics, using the Finite Element Method (FEM) and is available as source code and as a plug-in for Unreal Engine.
Highly optimized to take advantage of the processing power today’s modern multithreaded CPUs like AMD Ryzen series processors, FEMFX enables developers to add a whole new level of physics realism to their games.
TressFX is our GPU-accelerated realistic hair and fur rendering and simulation technology and the new version 4.1 offers improved performance and new features, including Unreal Engine integration.

To make it easier for developers to see FEMFX and TressFX in action, and for gamers to be able to check out some of the advanced effects they can expect to see in future game releases, we’ve put together some standalone demos that can be downloaded today.

For FEMFX, we have two demos – the “FEMFX Demo” is an interactive experience that shows off different FEM physics effects and the “Alien Pods” demo is an example of the gameplay puzzle mechanics that FEMFX could enable.

For TressFX, the “Old King” demo is a non-interactive real-time example of the kind of high-quality extremely realistic hair and fur effects our technology enables.

“Let’s Build…”, Our Developer Event You Can Attend from Home

To finish up the GPUOpen relaunch week, we are bringing everything together into our first-ever developer event you can attend virtually. The “Let’s build…” event includes many of the presentations we had planned for the now-postponed Game Developers Conference (GDC) 2020, giving everyone a chance to watch them.

Here is a preview of the video presentations you can watch starting 4 PM BST, 11 AM EDT this Friday:

• AMD Ryzen Processor Software Optimization: An introduction to the AMD Ryzen family of processors followed by advanced optimization topics.
• Optimizing for the Radeon RDNA Architecture: Join this session to learn about the details of RDNA and how it differs from GCN architecture.
• From Source to ISA: A Trip Down the Shader Compiler Pipeline: Uncover the secrets of the shader compiler on a journey from the source code to the RDNA ISA.
• A Review of GPUOpen Effects: This presentation will provide an overview of existing and new graphics effects available on GPUOpen, including FidelityFX.
• Curing Amnesia and Other GPU Maladies with AMD Developer Tools: Meet the newest tool in AMD's Radeon tool suite - the Radeon Memory Visualizer.
• Radeon ProRender Full Spectrum Rendering 2.0: The Universal Rendering API: An introduction to the upcoming Radeon ProRender 2.0, the universal rendering API.

https://community.amd.com/community/gaming/blog/2020/05/11/build-great-things-with-the-new-gpuopen

 

[shankly1985]

Posted here, but I actually like your thread starter more so I'll remove my thread.
https://forums.overclockers.co.uk/t...the-future-resigned-and-ready-to-go.18886699/

 

[Chuk_Chuk]

Hopefully AMD has the resources to get this into games. As far as i am aware, AMDs software tends to play nice with all graphics cards? compared to Nvidia software.

 

[Poneros]

Posted here, but I actually like your thread starter more so I'll remove my thread.
https://forums.overclockers.co.uk/t...the-future-resigned-and-ready-to-go.18886699/

Oh, my bad, for some reason I didn't see it. Mods can feel free to lock this one.

 

[varkanoid]

From all the blurb dont really get what this is. For instance the chess piece picture looks the same to me , and the helmet has slightly different brightness on the neck.
In one sentence what is this and what does it bring to the average PC gamer ?

 

[Stanners]

Threads merged, seemed a shame to lock one over the other!

 

[Ayahuasca]

From all the blurb dont really get what this is. For instance the chess piece picture looks the same to me , and the helmet has slightly different brightness on the neck.
In one sentence what is this and what does it bring to the average PC gamer ?

What screen are you viewing the images on? The chess picture has a clear difference in reflections if you look at the tiles in front of the middle 2 pawns and the knight on the left.

 

[Poneros]

Threads merged, seemed a shame to lock one over the other!

Thanks mate! Fast on the trigger

From all the blurb dont really get what this is. For instance the chess piece picture looks the same to me , and the helmet has slightly different brightness on the neck.
In one sentence what is this and what does it bring to the average PC gamer ?

They go into greater detail on the site. Essentially, it will be faster & more accurate reflections that any dev can implement with no strings attached.

https://gpuopen.com/fidelityfx-sssr/ It has a nice comparison slider but can't directly link it.

The term “Screen Space Reflections” describes an effect that can create realistic looking reflections purely based on information already present in the rendered image. The underlying algorithm shoots reflection rays from a depth imprint of the rendered scene (a so called “depth buffer”) and follows them in constant steps through the image until these rays intersect with the depth buffer again. Instead of taking constant sized steps, FidelityFX SSSR builds on an industry-leading algorithm that searches the rendered image in a hierarchical manner. This allows for larger and fewer steps on average, increasing performance of the search and quality of the final image.

To support glossy reflections, this FidelityFX effect jitters the reflection rays to create the sense of surfaces with varying roughness. However, that approach inherently introduces noise. Thus, FidelityFX SSSR comes with a high quality denoiser specifically optimized for the RDNA architecture. The denoiser combines the results from multiple frames to create a noise-free image. Furthermore, it allows to decrease the ray count based on surface roughness thus further speeding up screen space traversal.

 

[varkanoid]

What screen are you viewing the images on? The chess picture has a clear difference in reflections if you look at the tiles in front of the middle 2 pawns and the knight on the left.

Oh reflections, wasnt looking at reflections just the actual chess pieces themselves. Isnt that a bit of a con though it just a picture with no reflection against a picture with a reflection. Doesnt need this software or whatever it is just to add reflection on the floor ?

 

[Ayahuasca]

Oh reflections, wasnt looking at reflections just the actual chess pieces themselves. Isnt that a bit of a con though it just a picture with no reflection against a picture with a reflection. Doesnt need this software or whatever it is just to add reflection on the floor ?

Why wouldn't you be looking for reflections when the chess comparison is based on that feature Anyway, the first image does have reflections, they aren't turned off, it's just that the new method used in the second image does them much better.

 

[varkanoid]

I couldnt be bothered to read the marketing blurb just looked at the pictures for a comparison !

 

[shankly1985]

By the way guys you can download the effects samples and play around with them.

 

[Chuk_Chuk]

I wonder if these tools are usable on the consoles, since they run RDNA2. If they we could see this be widely adopted.

 

[motiv]

They should really concentrate on getting a good upscaler like Nvidia have now.

 

[EsaT]

They should really concentrate on getting a good upscaler like Nvidia have now.

Would expect one to be available for next-gen consoles.
Even GPU of XSX just won't be that good for high framerates at 4K with raytracing.

 

[shankly1985]

They should really concentrate on getting a good upscaler like Nvidia have now.

You mean DLSS? If so I think AMD will just use Microsoft version

 

[Poneros]

I wonder if these tools are usable on the consoles, since they run RDNA2. If they we could see this be widely adopted.

They absolutely are. It's just a matter of the devs choosing to implement them. Sometimes the support for them is not great, but that's on AMD to fix and push. Generally tho devs prefer to come up with their own solutions, so that way they can deploy it to all platforms equally.