Performance question

Support forum for DGDecNV
User avatar
Rocky
Posts: 3555
Joined: Fri Sep 06, 2019 12:57 pm

Performance question

Post by Rocky »

Tested a 4K source on RTX 3090:

AVSMeter 3.0.9.0 (x64), (c) Groucho2004, 2012-2021
AviSynth+ 3.7.1 (r3593, 3.7, x86_64) (3.7.1.0)

Number of frames: 3857
Length (hh:mm:ss.ms): 00:02:34.280
Frame width: 3840
Frame height: 2160
Framerate: 25.000 (25/1)
Colorspace: YUV420P10

Results for FFMS2:

Frames processed: 3857 (0 - 3856)
FPS (min | max | average): 54.24 | 515.0 | 118.6
Process memory usage (max): 849 MiB
Thread count: 36
CPU usage (average): 67.7%

Time (elapsed): 00:00:32.513

Results for DGSource:

Frames processed: 3857 (0 - 3856)
FPS (min | max | average): 113.4 | 277.4 | 265.2
Process memory usage (max): 390 MiB
Thread count: 24
CPU usage (average): 7.3%

Time (elapsed): 00:00:14.541

DGSource is more than twice as fast with about 1/10 of the CPU usage and about 1/2 the memory usage.

Now mark the thread resolved. ;)
DAE avatar
Guest 2
Posts: 903
Joined: Mon Sep 20, 2010 2:18 pm

Performance question

Post by Guest 2 »

Rocky wrote:
Tue Jan 04, 2022 9:30 am
Now mark the thread resolved. ;)
Please have some real x265 encoding numbers to show us, both with SW and HW decoder.

Just curious about performance differences on a newer and faster machine.

How much GPU memory do so many threads use? Just keep GPU-Z in background and use appropriate AVSMeter switch.
User avatar
Rocky
Posts: 3555
Joined: Fri Sep 06, 2019 12:57 pm

Performance question

Post by Rocky »

DGDecNV is a decoder so I give decoding tests. If you want to benchmark encoders or transcoding go for it.
DAE avatar
Guest 2
Posts: 903
Joined: Mon Sep 20, 2010 2:18 pm

Performance question

Post by Guest 2 »

Guest 2 wrote:
Mon Dec 27, 2021 4:34 am
Some years ago there was the possibility to choose between CUDA and CUVID decoding but it has disappeared.
Perhaps you forgot about this and today it just arose again to my mind.

P.S: I can't find anymore the thread about DGDecNV benchmarking. Now that a I have a "new" 1660 SUPER card, I'd like to test it again. Where has it gone? :oops:
User avatar
Rocky
Posts: 3555
Joined: Fri Sep 06, 2019 12:57 pm

Performance question

Post by Rocky »

That functionality was removed by nVidia.

Not sure what thread you are talking about. Jog my memory.
DAE avatar
Guest 2
Posts: 903
Joined: Mon Sep 20, 2010 2:18 pm

Performance question

Post by Guest 2 »

Rocky wrote:
Thu Mar 24, 2022 7:26 am
Not sure what thread you are talking about. Jog my memory.
There was a thread on Doom9 with instruction about how to measure decoding FPS with DGIndex (or DGDecNV?).

I remember that "Disable display" flag was involved too.

There were many results, so I was curious about comparing my "new" card.

That thread has been unfortunately deleted.

Can you please instruct us how to benchmark raw decoding performance (better if AVS/VS independent) and perhaps open a proper thread?
User avatar
Rocky
Posts: 3555
Joined: Fri Sep 06, 2019 12:57 pm

Performance question

Post by Rocky »

Guest 2 wrote:
Thu Mar 24, 2022 7:37 am
That thread has been unfortunately deleted.
Cancellation seems in vogue recently.
Can you please instruct us how to benchmark raw decoding performance (better if AVS/VS independent) and perhaps open a proper thread?
Sure, will do.
Post Reply