To get an idea about the possible savings of header compression schemes we compute the optimum and therefore theoretical bounds for voice, audio, and video traffic. For this first and simple calculation, we assume that no IP overhead (including the RTP and UDP headers) is used at all.

The potenial saving for voice and audio services are displayed. For voice we can use the Linear Predictive Coding (LPC) with 5.6 kbps, a GSM codec with 13.2 kbps, and a codec following the ITU-T standard G.711 with 60.0 kbps. For audio LPC leads to poor performance values. Thus, GSM or G.711 will be used for audio. A packet generation rate of 50Hz (a sample every 20ms) is assumed.

	mean bitrate	IPv4 savings	IPv6 savings
codec	[kbps]	[%]	[%]
LPC	5.6	74	81
GSM	13.2	55	65
G.711	60.0	21	29

In comparison to voice streams with fixed frame sizes, the frame sizes of a video stream vary over time. The size of the video frames depend on the content of the video sequence and the used encoder settings. We encoded the generally accepted video reference streams formen, costguard, akiyo, and silent in the QCIF format (176x144 pixel) with the H.26L encoder using the IBBPBBPBBPBB group of picture (GoP) structure.
 

	mean bitrate	IPv4	IPv6
movie	[kbps]	[%]	[%]
foremen	34.5	22	30
costguard	19.1	33	43
akiyo	46.8	18	24
silent	23.8	29	38