History of DLSS – Nvidia Made the Technology in Two Weeks?

Technology DLSS (Deep Learning Super Sampling) became one of key innovations in graphics over the past 20 years, changing the balance of power in the graphics card market. A new look at its origins is offered by the book The Nvidia Way: Jensen Huang and the Making of a Tech Giant, the author of which claims that DLSS was born in literally two weeks before its first announcement.

According to the publication, the idea artificial image scaling with the help of neural networks appeared in 2018, when the CEO Nvidia Jensen Huang was looking for a spectacular demonstration for a presentation at SIGGRAPHOne of the employees suggested DLAA (Deep Learning Anti-Aliasing), but Huang did not consider smoothing critical: "A better picture won't sell much GPU», - the book quotes the CEO of Nvidia as saying.

However, it was this conversation that led him to a key idea: "What if instead of improving an already good picture, we force the AI ​​to increase the resolution, making weaker cards more powerful?" This is how the concept was born DLSS - usage Tensor cores for upscaling, which allows video cards to render an image at a lower resolution and then upscale it to 4K without significant loss of quality.

After all week After the idea, Huang received the first prototype from the team Nvidia Research, and in two weeks DLSS was already featured in the SIGGRAPH 2018 presentation. The book emphasizes that the developers initially saw DLSS differently, but Huang immediately realized that this technology would help sell GPU expensive.

According to the author of the book, DLSS Allows Nvidia to Increase Value of Entire Line of Graphics Cards - now even mid-range and budget models could provide picture quality at the level of top solutions. This gave the company the opportunity raise prices and promote your technology as key argument in the fight against AMD and Intel.

Was DLSS really invented in a couple of weeks? The story sounds impressive, but too perfect to be true. However, even if the technology has been in development longer, one thing is clear - Nvidia saw it not just as a tool to improve graphics, but as a way to monetize GPU.