2006-12-15, 03:59 PM
I've done bit more testing with GB-PVR and the NVidia Decoder to figure out why playback uses so much CPU and is generally unfavorable. When I opened a MPG file in WMP or GraphEdt (letting it create a default graph) playback was good and had very low CPU usage, and I wanted the same within GB-PVR.
The conclusion? NVidia PureVideo decoders perform terrible with VMR enabled and supposed harware accelleration, worse than the Nero decoders w/o hardware accelleration. The good performance I was seeing in WMP and GraphEdt were due to Overlay mode (which has Pros and Cons) - VMR mode stinks.
Results might be different for you with your hardware.
My System:
AMD 2500
768 Meg RAM
NVidia 6200 w/ 256 Meg
NVidia 93.71 Forceware Drivers
Latest PureVideo Decoders (as of today - 12/15)
Testing: I tested by selecting my video mode and decoder in GB-PVR's config, playing the same segment of a video each time. Within GB-PVR I would play the video briefly, stop it and then open LastGraph.grf in the GB-PVR directory with GraphEdt. This allowed me to see the filters present, check their options and play the file for a couple minutes, watching CPU usage. The Overlay mode graph I couldn't open with GraphEdt (I got an unregistered filter error) so instead I did playback in GB-PVR and graphed it's CPU usage.
In VMR mode I did some tests removing GB-PVR specific filters from the graphs (GB-PVR Flow Status and GB-PVR Parser) to see what impact there was. As sub predicted, there was no impact whatsoever. The GB-PVR filters are not introducing performance penalties.
The only area that I have some uncertainty about is Overlay mode. GB-PVR's use of the NVidia Decoders doesn't perform as well as GraphEdt or WMP in Overlay mode - the difference is significant. I can't tell what's going on because I can't open the Overlay mode graph. I have a little more to investigate that might shed some light on it, but at this point I've proven GB-PVR isn't at fault for the VMR mode problems anyway.
Take a look at the graphs below:
The conclusion? NVidia PureVideo decoders perform terrible with VMR enabled and supposed harware accelleration, worse than the Nero decoders w/o hardware accelleration. The good performance I was seeing in WMP and GraphEdt were due to Overlay mode (which has Pros and Cons) - VMR mode stinks.
Results might be different for you with your hardware.
My System:
AMD 2500
768 Meg RAM
NVidia 6200 w/ 256 Meg
NVidia 93.71 Forceware Drivers
Latest PureVideo Decoders (as of today - 12/15)
Testing: I tested by selecting my video mode and decoder in GB-PVR's config, playing the same segment of a video each time. Within GB-PVR I would play the video briefly, stop it and then open LastGraph.grf in the GB-PVR directory with GraphEdt. This allowed me to see the filters present, check their options and play the file for a couple minutes, watching CPU usage. The Overlay mode graph I couldn't open with GraphEdt (I got an unregistered filter error) so instead I did playback in GB-PVR and graphed it's CPU usage.
In VMR mode I did some tests removing GB-PVR specific filters from the graphs (GB-PVR Flow Status and GB-PVR Parser) to see what impact there was. As sub predicted, there was no impact whatsoever. The GB-PVR filters are not introducing performance penalties.
The only area that I have some uncertainty about is Overlay mode. GB-PVR's use of the NVidia Decoders doesn't perform as well as GraphEdt or WMP in Overlay mode - the difference is significant. I can't tell what's going on because I can't open the Overlay mode graph. I have a little more to investigate that might shed some light on it, but at this point I've proven GB-PVR isn't at fault for the VMR mode problems anyway.
Take a look at the graphs below: