Speculation Nvidia AI upscaling. - Psych forums

Our partner

Jump to:

Thoughts from transgressing dimension
Here you can see some of my wild thoughts and you may find some good worldly ideas on here. I just love thinking and thought I should let my thinking be read.

highdimensionman: Consumer 2; Posts: 55; Joined: Tue Jul 08, 2014 5:38 pm; Blog: View Blog (1012)

Archives

November 2022

China should know they can change their fate and the worlds fate: Mon Nov 28, 2022 8:08 pm

September 2022
August 2022
July 2022
June 2022
May 2022
April 2022
March 2022
February 2022
January 2022
December 2021
November 2021
October 2021
September 2021
August 2021
July 2021
June 2021
May 2021
April 2021
January 2021
December 2020
November 2020
October 2020
September 2020
August 2020
July 2020
June 2020
May 2020
April 2020
February 2020
November 2019
October 2019
September 2019
May 2018
April 2018
March 2018
February 2018
December 2017
August 2017
May 2017
April 2017
March 2017
February 2017
December 2016
November 2016
October 2016
September 2015
August 2015
April 2015
March 2015
February 2015
January 2015
December 2014
November 2014
September 2014
August 2014

Search Blogs

Advanced search

Speculation Nvidia AI upscaling.

by highdimensionman on Tue Apr 12, 2022 9:48 pm

Imagine you take a 580p lossless video. use red green and blue 3 bit dithering, lower some resolution, take out small triangular pieces, tweak the accuracy down, triangulate the remain data into a video file. Then train a double data type augmented dataset the first data type is for data regarding general estimation based morphic upscaling and piece analysis general estimation for filling in the blanks. So this data upscales by morphing in a higher resolution. The second data type is for handing bias. So rather than turning a face into a more generalist estimated upscale the video is tailored relative to a deeper data set.
Nvidia could make an architecture extension for ARM that inferred/rendered efficiently from this 2 data type augmented dataset.
This Technology could be licensed as an ARM extension for a few million dollars.
In Gaming on the next nintendo to come out in a few years time hopfully, It's key to remember that 4k could be done more efficiently if you used concentration point field 3d surfacing. by having a cache of point concentration graph information in 2d you could represent 3d surfacing in point concentration data meaning that you could lower the number of triangles needed working with per frame and lower the data accuracy optimally in sync with touch up data for 400p <-> 800p upscaled to 1k<->2k then a lower quality upscaler is used without touch up data to take it to 4k.
Tweak the surfacing for lumen cross Hardware raytracing lowering accuracy relative to what can later be upscaled.
Also having an engine which dynamically tweaks accuracy
for a higher frame rate.
This could mean that you could have some next gen nintendo 35watt system play any level of PS5 game looking roughly as good.

Diagnosis - Paranoid Schizophrenia although challenged.
Medication - Clopixol injection 4 weekly.
Personal diagnosis - I battle with the chemistry and biological effects of Clopixol every day trying to improve on the heath and range of my approach.
I try to improve my thinking. Hopefully some good ideas shine through.

0 Comments Viewed 3321 times

Who is online

Registered users: Bing [Bot], Google [Bot], Google Adsense [Bot], Google Feedfetcher, Yahoo [Bot]