wccftech - NVIDIA Invites Developers to Test Experimental DLSS Models Directly From Company’s Supercomputer
 
“NVIDIA recently began inviting developers to test the newest build for DLSS (Deep Learning Super Sampling) and submit their experiences and findings to the developer forum on NVIDIA's site.
 
NVIDIA is enabling developers to explore and evaluate experimental AI models for Deep Learning Super Sampling (DLSS). Developers can download experimental Dynamic-link libraries (DLLs), test how the latest DLSS research enhances their games, and provide feedback for future improvements.
 
What is interesting about the process is they are allowing developers to train and test using NVIDIA's supercomputer. Through this utilization, they are hoping to continue to improve visual quality on more games and applications.
 
One of the key advantages of a deep learning approach to super sampling is that the AI model can continuously improve through ongoing training on NVIDIA’s supercomputer.
 
There are currently two new experimental DLSS Models for download. The findings that NVIDIA receives may or may not be on the final DLSS releases to the public. NVIDIA states that the two downloads ‘are not fully validated and may contain regressions.’
 
The first test will show the developer ‘improved object detail in motion,’ as well as ‘particle visibility.’ NVIDIA strived to improve ‘handling of moderate native instability,’ while also minimizing ghosting on displays. The second offers an alternative improvement when it comes to ghosting in applications and games.”
 
My thoughts:  Getting developers involved is probably not totally altruistic on NVIDIA’s part.  Improving imagery is great and I hope that they can keep lowering the overhead required.