Unified Cache Architecture + Adaptive Shading

Unified Cache Architecture

The NVIDIA TU116 also features a unified architecture for shared memory, L1 and texture caching. This allows the L1 cache to dynamically adjust its size for optimal performance.

Its L1 cache can grow as large as 64 KB in size, with 32 KB per SM shared memory allocation. Or it can shrink to 32 KB, which would allow 64 KB for shared memory.

In addition, combining the L1 data cache with shared memory cache reduces latency, and increases throughput. In fact, NVIDIA claims its hit throughout is quadrupled compared to Pascal.

NVIDIA showcased Call of Duty: Black Ops 4, which they claimed received a 40% performance boost due to the Turing unified cache architecture.

 

Adaptive Shading

The Turing architecture introduced a new shading technology called Variable Rate Shading (VRS)., with two key algorithms – Content Adaptive Sharing and Motion Adaptive Shading.

Irrespective of the algorithm used, the TU116 can adjust the shading rate for different regions of a scene, or even for specific objects. This reduces areas that need to be rendered in full detail, boosting performance.

Motion Adaptive Shading is used to adjust the shading rate based on the amount of motion present in a particular region. In a racing game, the player’s car would be rendered in full detail, while the road, lane markings and street signs could be rendered in lower detail with fewer samples. Since they’re moving rapidly – the human eye won’t perceive a slight reduction in pixel quality.

[adrotate group=”1″]

Content Adaptive Shading adjusts the shading rate according to spatial and temporal colour coherence across frames. In cases where color or spatial changes are minimal from frame to frame (for example, the sky in a scene, or a wall), the shading rate can be lowered in successive frames to improve performance without any noticeable reduction in image quality.

Wolfenstein II: The New Colossus was the first game to feature Turing’s Adaptive Shading technology. The GeForce GTX 1660 Ti receive a 50% boost in performance over the GeForce GTX 1060 when Adaptive Shading was enabled.

Next Page > Presentation Slides, Full Tech Briefing + Q&A Session

 

Support Tech ARP!

If you like our work, you can help support our work by visiting our sponsors, participating in the Tech ARP Forums, or even donating to our fund. Any help you can render is greatly appreciated!