|Microsoft DirectX 9.0|
In a Microsoft?DirectPlay?Voice session using a peer-to-peer topology, each voice-session client streams its voice audio data directly to every other voice-session client. Each client receives all individual incoming voice audio streams, mixes the received streams, and plays the resulting mixed signal on the client's computer.
The advantage of using a peer-to-peer topology is that no computer in the voice session requires high bandwidth or processor power. However, the bandwidth and processor usage on each client's computer varies according to the number of incoming and outgoing audio streams. The number of outgoing voice audio streams is equal to the number of targets participating in the voice session, unless the network provider is capable of true multicasting, as noted below. The number of incoming voice audio streams depends on how many other voice-session clients are targeting the client in question and also on how many of the other clients are speaking.
As a game design consideration, it is not useful for a voice-session client to be the target of more than about six to eight other clients. If all six to eight clients are speaking at once, the conversation can become confusing, and communication between clients can be difficult.
If the DirectPlay network session supports true multicasting, the number of outgoing voice audio streams can be reduced considerably. If all clients are part of a multicast network and the target of the voice stream is a DirectPlay group, there is only one outgoing stream.
A DirectPlay voice session using a peer-to-peer voice topology supports 3-D spatialization of the voice data using the IDirectPlayVoiceClient::Create3DSoundBuffer method.
It is important to note that a voice session using a peer-to-peer voice topology cannot be used if the network transport is a client/server session.