It seems that the hacker’s attack on NVIDIA this time was “effective”. In the leaked 75GB confidential file , not only did we dig out the three future architectures of Ada, Hopper, and Blackwell and their core codes , even the DLSS source code!
DLSS is deep learning supersampling, a super-resolution image quality enhancement technology introduced by the NVIDIA Ampere architecture, which can greatly improve the game frame rate, especially with RTX light tracking technology, which can take into account high image quality and high frame rate.
AMD FSR and Intel XeSS can be said to be led by it.
The DLSS source code leaked this time is very rich in content, including various C++ files, header files, development resources, programming instructions, etc. , many of which are internal secrets, and of course some have been made public, which can be found on GitHub.
It is worth mentioning that the version corresponding to these leaked files is DLSS 2.2. Although the technology has been upgraded to DLSS 2.3 in November last year , the difference is not big, and the technical principles are the same.
At present, many experts are digging into these files to see what secrets they can find, and maybe find a way to make Linux open source drivers support DLSS.
If AMD and Intel take the opportunity to study…