Indie game storeFree gamesFun gamesHorror games
Game developmentAssetsComics
SalesBundles
Jobs
Tags

barry1900

2
Posts
1
Topics
A member registered May 25, 2020

Recent community posts

(2 edits)

Hey there!

I suggest sharing the experience of Dain-app tests in this topic to identify the optimal settings. So that beginners do not to sort through all random combinations of settings through long experiments, and don't use obviously ineffective examples.

I found out about this wonderful application 3 days ago. And starting the selection of optimal settings for fast rendering, I proceeded from the recommendations of users from web sites.
Dain-app has recommended default settings for Section Size:480px and Section Padding:150. I also saw custom examples on the forum, for example 400x100. And when I tried different settings, I basically changed only the value of Section Size, but i don't changes Section Padding, i left it by default.
 
As it turned out, I underestimated this setting, today's experiments have made me understand that rendering can happen much faster if you set the Section Padding value to 0.
*At least in the configuration of my computer, interpolation has become significantly faster.

My PC specs: Win 10 / RTX 2080 Ti Strix OC 11Gb (EK water cooled) / Ryzen 3970x / 128 RAM / m.2 970 Evo Plus SSD

For the test, several short video clips with different resolutions were choice. Each video is 0.5 seconds long and consists of 15 frames (30fps).

Below I will give the results of my tests.
// test video resolution // 
/ @ Section Size x Padding in px / total rendering time / (mb VRAM usage in interpolation process) / CPU load / GPU working frequency /
***I was interested to see how changing the split frame into sections settings will affect the load and distribution of system resources.

1280х720 - interpolate х8
 @ off     - 4m:21s
+@ 720x0   - 4:39        (3887-5893mb VRAM used)
+@ 800x0   - 4:41        (4100-6617mb)
 @ 600x0   - 5:00        (3000-4571mb)
 @ 840x150 - 5:37        (4600-7370mb)
 @ 960x150 - 5:39        (5000-8096mb)
 @ 600x50  - 7:23
 @ 600x100 - 8:17        (3800-7713mb)
 @ 600x150 - 9:09        (3000-7800mb)
 @ 450x150 - 9:15        (4600-5900mb)
 @ 600x600 - 13:04       (4000-9800mb)
 @ 150x150 - 23:01       (4600-5900mb)
\

1920х1080 - interpolate х4
 @ 850x150 -  6:57       (3200-9647mb)
//
1920х1080 - interpolate х2
+@ 480x0   -  1:42       (3469mb) CPU 14-45%, GPU 1635-1950MHz
+@ 500x0   -  1:42       (3463mb) CPU 14-55%, GPU 1635-1950MHz
 @ 500x0   -  1:42 (repeat, 2 try)
 @ 540x0   -  1:58       (4571mb)
 @ 850x0   -  1:58/1:59  (4300-7461mb) CPU 3-27%, GPU 1635-1950MHz (2 try)
 @ 950x0   -  2:07
 @ 400x0   -  2:24       (2500-3469mb) CPU 14-34%, GPU 1620-1950MHz
 @ 270x0   -  2:39       (2687mb) CPU 15-45%, GPU 1620-1950MHz
 @ 850x150 -  3:03       (9597mb)
 @ 700x150 -  3:26       (4700-9900mb)
 @ 480x150 -  3:56       (3100-8611mb) CPU 3-25%, GPU 1605-1950MHz
 @ 108x0   -  4:49       (1659mb) CPU 58% and GPU 1950MHz - without the slightest change
 @ 135x0   -  5:14       (2011mb) CPU 58%, GPU 1680-1950MHz
 @ 50x0    -  7:45
\

3840x1620 - interpolate х4
+@ 500x0   - 11:27       (3469mb) CPU 19-58% (~25), GPU 1635-1950MHz
 @ 840x150 - 13:48       (9646mb)
 @ 850x150 - 13:48       (9646mb)
 @ 800x150 - 18:42       (9646mb)
 @ 600x150 - 24:48       (3800-8600mb)
 @ 800x200 -   :         out of mem
 @ 860x150 -   :         out of mem
//
3840x1620 - interpolate х2
+@ 500x0   -  4:15       (2500-3469mb) CPU 19-58%, GPU 1635-1950MHz
 @ 850x0   -  4:20       (7463mb) CPU 3-20%, GPU 1635-1950MHz
 @ 400x0   -  6:03       (3471mb) CPU 20-58% (~25), GPU 1635-1950MHz
\

3840x2160 - interpolate х4
+@ 500x0   - 16:16       (3469-3471mb) CPU 19-42%, GPU 1635-1950MHz
 @ 600x150 - 23:45       (8682-9694mb)
 @ 800x150 -   :         out of mem 
\

(1 edit)

Win 10 /// RTX 2080 Ti Strix OC 11Gb (water cooled) / Ryzen 3970x / 128 RAM / m.2 970 Evo Plus SSD

Test file 0,5sec 1080p 30fps x2 interpolation (15 frames into 29). 15 frames original from the video. This means that only 14 new frames was generated.
Split frame into section setting:
 @ 850x150 - 3m:03s ~ 12 frames rendered about 2m:37s
 @ 700x150 - 3m:26s ~ 12 frames rendered about 2m:56s

Even if calculate the processing of all 29 frames during interpolation, your PC configuration completed this task faster.
Maybe it depends on the content in the image? Otherwise, I don’t understand why my computer rendered frames much slower than yours ...