I think is different to compare openlayers in SVG vs. Canvas to graph visualization with physics like in the directed graph sample. In the last case there is a lot of extra work that you need to do in Canvas to came up with the same effects.

Is that test taking into account different GPU accelerations for SVG like in this "old" study: http://blogs.msdn.com/b/ie/archive/2011/03/08/comparing-hard... ?