Performance Comparison - Versions and profiles

Coordinator
Apr 30, 2014 at 2:17 AM
Edited May 23, 2014 at 5:11 PM
Latest benchmark details here.
A quick comparison between 2.3.13 v/s 2.3.15 v/s 2.3.15 with Hardware Encoding. The time taken to complete conversion is measured in hours.

Source WTV video is:
Video codec: H.264
Audio codec : AAC LATM
Duration : 1hr 5min
Resolution: 1080p (1920 x 1080)
Computer/Conversion profile:
Processor: Intel Core i7 3540M (3.0Ghz)
Video card: HD Graphics 4000
RAM: 6GB
OS: Windows 8.1 Professional 64bit
Comskip: Donator
Max Resolution: 1080p (1920 x 1080)
Automatic Quality Detection: Enabled (default)
Quality: 0% (default)
Volume: 1 (default)
Cropping: Enabled (default)
.
As you can see there is about 25% increase in performance between 2.3.13 and 2.3.15 and about 1800% increase in performance between 2.3.13 and 2.3.15 with H/W encoding

Image
Marked as answer by rboy1 on 4/29/2014 at 7:36 PM
Coordinator
Apr 30, 2014 at 2:20 AM
See this thread for more comparisons on 2.3.12 and 2.3.14

https://mcebuddy2x.codeplex.com/discussions/395633
Coordinator
Apr 30, 2014 at 2:21 AM
See this thread for more comparisons on 2.3.15 profiles (Software Encoding)

https://mcebuddy2x.codeplex.com/discussions/509545
Coordinator
Apr 30, 2014 at 2:37 AM
A quick comparison between 2.3.13 v/s 2.3.15 v/s 2.3.15 with Hardware Encoding. The time taken to complete conversion is measured in hours.

Source WTV video is:
Video codec: H.264
Audio codec : AAC LATM
Duration : 1hr 5min
Computer profile:
Processor: Intel Core i7 3540M (3.0Ghz)
Video card: HD Graphics 4000
RAM: 6GB
OS: Windows 8.1 Professional 64bit
Comskip: Donator
.
As you can see there is about 25% increase in performance between 2.3.13 and 2.3.15 and about 1800% increase in performance between 2.3.13 and 2.3.15 with H/W encoding

Image
Jul 25, 2014 at 3:04 AM
Is it possible to post some comparison w/ OpenCL? Folks w/ AMD CPU are left w/ OpenCL for faster encoding.

My current card HD4600 doesn't support OpenCL. I am looking to upgrade to one that supports, but wanted to know how much benefit I can get if I were to put a newer card. I hope to take some benefit of upgrading to 2.3.15.
Jul 26, 2014 at 1:18 AM
When I enabled Quicksync, it cut my MP4 Normal in half, but when I tried MP4 High Quality my times didn't improve. The log file states that hardware encoding only works on first pass, not on 2 pass. Is this correct?
Coordinator
Jul 26, 2014 at 3:09 AM
Correct. When using quick sync it defaults to single pass

Coordinator
Jul 26, 2014 at 12:51 PM
That's a hardware limitation, quicksync does not support 2 pass yet.


Jul 26, 2014 at 2:41 PM
Understood. So on your comparison chart above, how are you getting such a drastic time improvement on the MP4 High Quality setting when the Quicksync only speeds up the first pass, but not at all on the second?
Coordinator
Jul 26, 2014 at 7:57 PM
It still comparing to non accelerated

Jul 31, 2014 at 6:21 PM
I'm going to be running a comparison tonight/tomorrow for myself. My hardware isn't nearly that top-notch, but I have an Intel i3-4340 CPU and HD5770 GPU in the same box, so I think I can get a pretty solid comparison of relative OpenCL vs. QuickSync performance. I'll post the results as soon as I have them.
Aug 1, 2014 at 2:12 AM
Look forward for the comparison @belatukadro. Thanks for running them.
Aug 14, 2014 at 5:15 AM
belatukadro wrote:
I'm going to be running a comparison tonight/tomorrow for myself. My hardware isn't nearly that top-notch, but I have an Intel i3-4340 CPU and HD5770 GPU in the same box, so I think I can get a pretty solid comparison of relative OpenCL vs. QuickSync performance. I'll post the results as soon as I have them.
i'd be very interested in those results too.

by the way, i currently use WTV-Unprocessed, because i want to keep the original quality (size isn't an issue for my storage) but i'm curious... is there a way to use the GPU to process the file (originally MPEG2/H.264) instead of the CPU processing the entire file just to remove Ads ?
Oct 21, 2016 at 4:06 PM
Edited Oct 21, 2016 at 5:38 PM
p3ngwin wrote:
belatukadro wrote:
I'm going to be running a comparison tonight/tomorrow for myself. My hardware isn't nearly that top-notch, but I have an Intel i3-4340 CPU and HD5770 GPU in the same box, so I think I can get a pretty solid comparison of relative OpenCL vs. QuickSync performance. I'll post the results as soon as I have them.
i'd be very interested in those results too.

by the way, i currently use WTV-Unprocessed, because i want to keep the original quality (size isn't an issue for my storage) but i'm curious... is there a way to use the GPU to process the file (originally MPEG2/H.264) instead of the CPU processing the entire file just to remove Ads ?
I have a i5-4590K Haswell CPU (overclocked to 4.4ghz) , a ATI R9 200 GPU, and a seperate machine with an i5-4400t Skylake.

Typically I use the MKV Normal Quality profile with a +20% quality enhancement. Via QuickSync conversion, I get between 100-200 fps depending on the video - I can convert a 30min WTV in about 8-12 minutes and a 60min WTV in about 15-20 minutes.

If I use OpenCL on my graphics card, I typically get around 30-40 fps - meaning OpenCL is about 3 times slower.
Oct 21, 2016 at 4:08 PM
Edited Oct 21, 2016 at 6:17 PM
p3ngwin wrote:
belatukadro wrote:
I'm going to be running a comparison tonight/tomorrow for myself. My hardware isn't nearly that top-notch, but I have an Intel i3-4340 CPU and HD5770 GPU in the same box, so I think I can get a pretty solid comparison of relative OpenCL vs. QuickSync performance. I'll post the results as soon as I have them.
i'd be very interested in those results too.

by the way, i currently use WTV-Unprocessed, because i want to keep the original quality (size isn't an issue for my storage) but i'm curious... is there a way to use the GPU to process the file (originally MPEG2/H.264) instead of the CPU processing the entire file just to remove Ads ?
This would require comskip to use OpenCL via FFMPEG. In order to turn that on, go to your comskip.ini file and change hardware_decode=0 to hardware_decode=1

I find that GPU (OpenCL) encoding for comskip is much slower than using the CPU (hardware_decode=0 and thread_count=4). CPU encoding is over 1000fps and GPU is around 100fps