Q™
IPCT Contributor
"Intel®" vs. "Intel®+VideoPostProc?"
- Thread starter Q™
- Start date
fenderman
Staff member
- Joined
- Mar 9, 2014
- Messages
- 36,901
- Reaction score
- 21,270
4.3.2 - April 12, 2016
- The Intel hardware decoding on Options/Cameras now has become a drop list was an additional choice, either No, Yes, or Yes with Video Post Processing (VPP). The new setting additionally moves the post-decode color space conversion from NV12 to RGB32 into hardware as possible. In a benchmark test of 12 cameras (a mix of 2MP, 720p and several SD) processing 1.9 MB/s of video, the average 25% CPU usage becomes 20% with just the hardware decoding, but is further lowered to 16% when VPP is added, for a total reduction on CPU demand of 36%. With fewer cameras and bandwidth, the gains are even more dramatic (50% or more). With higher demands however, perhaps with more than 20 cameras and 3MB/s, the performance advantage of adding VPP becomes negligible, and may actually being to worsen performance as the CPU's ability to perform video tasks becomes the bottleneck over CPU cycles. For this reason, you may choose to use hardware decoding with or without VPP on a per camera basis, perhaps to assist with your most demanding resolution and frame rate cameras.
- The camera specific Intel hardware decoding selection has moved to the Video tab in camera properties. You may choose to use the default as set on Options/Cameras or any specific setting.
- If you are using hardware decoding with a particular camera, hardware decoding is now also used by the BVR playback engine whether in the local viewer or remote client.
- The preparation for the addition of hardware VPP required extensive revision of how video frames are formatted internally. If you notice an instance where a video frame is upside down where it should not be, my apologies--please notify the developer of this.