Ray tracing gets most of the attention when one of NVIDIA's RTX cards is reviewed, and rightfully so but it is not the only new feature these cards bring to the table. Content Adaptive Shading is one type of Variable Rate Shading, or VRS, which allows a Turing card to divide a screen into groups of pixels and then focus the application of shading to those groups which require it the most, spending processing time on shading areas which do not.
The Tech Report delves into this topic in more depth, as well as showing off what it does to one of the few games which currently support it. See just how The New Colossus is improved by CAS in this article.
"While Nvidia's RTX ray-tracing stack may be getting all the press, the Turing architecture has plenty of other tricks up its sleeve. One of these is called variable-rate shading, and its capabilities lay the foundation for a technique called content-adaptive shading. We tested what this tech can do for the performance of Wolfenstein II: The New Colossus."
Here are some more Graphics Card articles from around the web:
- XFX RX 590 Fatboy @ OCC
- XFX Radeon RX 590 Fatboy 8G @ Guru of 3D
- The Red Devil RX 590 vs. the EVGA GTX 1060 SC 6GB Overclocking Showdown @ Babeltech Reviews
- 20-Way AMD / NVIDIA Linux Gaming Benchmarks For The 2018 Holidays @ Phoronix
Can’t seem to find the link
Can’t seem to find the link to the Tech Report article in the above. Did it go on American Thanksgiving holiday already?
Edit to add: Link
If I manage to forget to put
If I manage to forget to put it inline, it is always at the bottom of the article. Fixing that now.
And folks are mad at AMD for
And folks are mad at AMD for not geting Vega’s Hardware based Primitive Shaders fully enabled via drivers and/or an API. AMD had too many new hardware feaures in Vega that required new driver or API support in order for games developers to make use of that IP. Now AMD will only have the time to get the Explicit Primitive Shaders enabled in software/API/Drivers by Navi’s release so maybe that will be backportable to Vega in 2019-2020.
Nvidia is a bit farther along with its new RTX IP support but that’s only because the Tensor Core IP has been around since Volta and Only the Ray Tracing Cores are new with the Turing GPU Micro-Arch. But still gamers are likewise unhappy with Nvidia’s decision to go with Ray Tracing and not more ROPs for faster rester oriented gaming. Now Nvidia has to work even harder getting the Games Developer’s code optimized for all of Turing’s new features like VRS and Ray Tracing/BVH.
Hopefully by 2020 most games will be able to fully make use of all that new Nvidia and AMD IP that is in the GPUs’ hardware but can’t as of yet be fully utilized by games developers without tons of support from the GPU makers over a period on months. Hopefully AMD is already been working on Tensor Core IP for obvious Professional Compute/AI market reasons and some Form of DLSS Like IP will be available on AMD’s consumer GPU variants within the next few years. That Nvidia Ray Tracing/BVH acceleration IP will probably take AMD longer to design and get certified than any Tensor Core IP.
Maybe AMD can include some FPGA based Ray Tracing IP along with its HBM2 Stacks and get some of that IP onto its Navi SKUs in FPGA form while AMD works on getting Ray Tracin IP directly integrated on its GPUs in ASIC form later on.
AMD has a patent filing for some on the HBM2/Newer HBM DIE stacks FPGA compute included along with the HBM Memory DIEs. So maybe AMD could Implement any Ray Tracing/BVH acceleration logic in FPGA from and do the Ray Tracing work right on the HBM# stacks. Even if Navi’s design is already frozen FPGAs can be added in on the GPU/HBM2-stacks module or even the Zen-2 MCM along with the Zen-2 Die Chiplets and the Big I/O Die. Then the FPGA could be programmed for Ray Tracing workloads and even reprogrammed later on if any improved Ray Tracing methods became available.
I would not be supprised if Sony had AMD add some form of FPGA IP onto their next generation PS5 based APU’s design that could be at programmed for Tensor Core like compute and AI based upscalling for 4K console gaming. Because AI based upscaling is just what would help console gaming the most without overtaxing the already relatively weak GPU designs that are used on gaming consoles.