November 17, 20196 yr I've build many a PC in the past and every time I do that I research what the latest tech is. I haven't build a PC in some years but am obviously considering an upgrade once MSFS is released. This has led me to the following: The little I've read so far on the GeForce RTX 2080 is that it has the capacity for "deep learning super sampling" (DLSS). This appears to be an alternative approach to AA that is less resource intensive and so greatly increases frame rates. The demo videos I've seen on YouTube look equivalent in image quality but with 10 - 30 fps increases. My question to the experts is, can DLSS work on any game or does it have to be specifically coded for it? Is it a possible under DX11 or does it require DX12 or is it DX-independent? If it's generic and DX11, I wonder if it's part of the mix that's giving the MSFS alpha such a high frame rate with outstanding image quality (and the reason why their demo rigs are using 2080 cards). If it's not currently possible, I wonder if an Asobosoft staff member would like to drop in here briefly and tell us if they have plans to implement it in the future (oh and TrackIR 😎).
November 17, 20196 yr DLSS needs to be tailored for specific games by training the AI behind it. I've tried it in a couple of games that support it; Metro Exodus and Control. I found that the image quality wasn't noticeably better than simply running the game at a lower resolution with a fast AA solution like FXAA. The YouTube compression algorithm tends to hide the detail, so you can't really spot the lack of detail in videos. Edited November 17, 20196 yr by Ikaro
Archived
This topic is now archived and is closed to further replies.