What follows are graphs of the mean encode time over 5 successive encodes for 4 relatively short videos using v8.0.9.0. In addition the same encodes were run against v8.0.8.2, less the
FE2_VP6_CONCURRENCY parameter, as some modest gains were made outside the scope of this parameter. Note that exceeding the per-processor input buffer maximum (60) will reduce the overall gain as the encode process will become serialized until all frames are buffered.
The graphs are split into 2 main sections grouped by the processor on the host machine:
System Specifications: Slackware 10.2 (i686), Pentium 4 540 (3.2GHz HT)
System Specifications: openSUSE 10.2 (x86_64), Intel Xeon X5355 (Quad-Core 2@2.66GHz)