Afterburner and Dxtory video recording tests
SlowLefty
Join Date: 2011-02-13 Member: 81653Members, Reinforced - Shadow
Join Date: 2011-02-13 Member: 81653Members, Reinforced - Shadow
Just updated to MSI Afterburner 2.3.0 and noticed it now supports codecs and recording your mic.
So did some performance experiments and thought I would share the results.
<b>The system:</b>
PC is quad core at 3.3ghz, but 0ms waiting for GPU, so CPU is the bottleneck and recording affects the FPS.
<b>The test:</b>
Start server, cheats on, join marines, start in a corner facing the wall, then run around a bit.
<b>Recording settings:</b>
1/2 frame of 1920x1080 at 30 FPS, 30 seconds of footage.
<b>Afterburner results:</b>
NO RECORDING - 100 FPS
UNCOMPRESSED - 65 FPS, 1.2GB. Large SSD needed for this option.
RTV1 90% - 65 FPS, 310MB. Grainy speckles, not supported by VLC 2.0.4.
MJPG 80% - 65 FPS, 74MB. Dark even with gamma correction ticked.
X264VFW CODEC - 61 FPS, 36MB. Best quality, smallest size, but less FPS.
<b>Dxtory trial results:</b>
X264VFW CODEC - 79 FPS, 36MB. Lots of options, but at a £28/$44 price.
<b>Codec settings:</b>
X264VFW
Preset - Ultrafast
Tuning - None
Profile - Auto
Level - Auto
Fast Decode - Ticked
Zero Latency - Ticked
Rate control - Single pass (CRF)
Ratefactor - 23
Output mode - VFW
VFW FourCC - X264
<b>Codec notes:</b>
Changing 'Multithreaded optimization' in Afterburner made no FPS difference.
Changing 'Ratefactor' to highest or lowest quality made no FPS difference.
Changing 'Rate control' from 'CRF' to 'CQP' made the quality worse.
<b>Ratefactor notes:</b>
This setting is a quality VS file size trade off.
The default of 23 seemed to be optimum, higher and blocks start to be noticable.
23 = 72MB per minute.
21 = 90MB per minute.
17 = 134MB per minute.
<b>Thoughts:</b>
As NS2 is so CPU intensive, even with Dxtory, recording had a noticable hit.
To record all the time I would first replace my aging Q6600 with an i5 or i7.
Dxtory definitely gives you something for your money.
If you have CPU power to spare, Afterburner could be a great free alternative.
So did some performance experiments and thought I would share the results.
<b>The system:</b>
PC is quad core at 3.3ghz, but 0ms waiting for GPU, so CPU is the bottleneck and recording affects the FPS.
<b>The test:</b>
Start server, cheats on, join marines, start in a corner facing the wall, then run around a bit.
<b>Recording settings:</b>
1/2 frame of 1920x1080 at 30 FPS, 30 seconds of footage.
<b>Afterburner results:</b>
NO RECORDING - 100 FPS
UNCOMPRESSED - 65 FPS, 1.2GB. Large SSD needed for this option.
RTV1 90% - 65 FPS, 310MB. Grainy speckles, not supported by VLC 2.0.4.
MJPG 80% - 65 FPS, 74MB. Dark even with gamma correction ticked.
X264VFW CODEC - 61 FPS, 36MB. Best quality, smallest size, but less FPS.
<b>Dxtory trial results:</b>
X264VFW CODEC - 79 FPS, 36MB. Lots of options, but at a £28/$44 price.
<b>Codec settings:</b>
X264VFW
Preset - Ultrafast
Tuning - None
Profile - Auto
Level - Auto
Fast Decode - Ticked
Zero Latency - Ticked
Rate control - Single pass (CRF)
Ratefactor - 23
Output mode - VFW
VFW FourCC - X264
<b>Codec notes:</b>
Changing 'Multithreaded optimization' in Afterburner made no FPS difference.
Changing 'Ratefactor' to highest or lowest quality made no FPS difference.
Changing 'Rate control' from 'CRF' to 'CQP' made the quality worse.
<b>Ratefactor notes:</b>
This setting is a quality VS file size trade off.
The default of 23 seemed to be optimum, higher and blocks start to be noticable.
23 = 72MB per minute.
21 = 90MB per minute.
17 = 134MB per minute.
<b>Thoughts:</b>
As NS2 is so CPU intensive, even with Dxtory, recording had a noticable hit.
To record all the time I would first replace my aging Q6600 with an i5 or i7.
Dxtory definitely gives you something for your money.
If you have CPU power to spare, Afterburner could be a great free alternative.
Comments