There should not be much variance in the geometry of impact test bars, assuming you are testing to a known standard. Are you doing Izod or Gardner testing?
Length shouldn't matter but thickness will. As the the specimen gets thicker, it will transition from plane stress to plain strain, which is to say from a bi-axial to tri-axial stress state. This will facilitate craze formation near the crack tip and there will be a significant drop in energy per unit thickness.
Rick Fischer
Principal Engineer
Argonne National Laboratory
if you are using CVN test, the lenght also is important, since duration of test could be shorter if the test piece is shorter too, hence the energy could be lower than real. At the opposite side, if the test specimen is larger, more friction over the anvil, more energy to fracture