Predictions of the Fractal Engine Simulation code were compared with SI engine data in a previous paper. These comparisons were extremely good except for the single data set available at a low engine speed. Because of uncertainty regarding whether the lack of agreement for this case resulted from some difficulty with the experimental data or was due to lack of proper speed dependence in the model, additional comparisons are made for a range of speeds from 300-1500 rpm. The fractal burning model is a turbulence driven model (i.e., driven primarily by the turbulence intensity) that divides the combustion process into four sequential phases: 1) kernel formation, 2) early flame growth, 3) fully developed turbulent flame propagation, and 4) end of combustion. The kernel formation process was not included in the previous version of this model, but was found to be required to predict engine speed effects. An empirical ignition (or wrinkling) delay was used, during which the flame propagates at the stretched laminar flame speed. The delay time decreases rapidly with increasing engine speed and no delay is needed at 1500 rpm. The kernel radii at the end of this delay agree well with the ∼2 mm found experimentally in a different engine. This delay is the only adjustable constant in the model. For those cases where it is required, the model predictions are not very sensitive to ±10% variations in the delay constant. Inclusion of this delay results in predictions that are in excellent agreement with the data for the u' boundary layer thickness, the fractal dimension, the location of peak pressure, the rate of pressure rise, the indicated work during combustion (taken to be from the ignition crank angle to 390 °CA), and the peak pressure. It is also found that accounting for the turbulence intensity boundary layer is critical to properly simulating the decreased burning rate near the end of combustion. It is also found that the model predictions are not overly sensitive to our present lack of knowledge regarding the integral length scale in engines, but the integral length scale may play an important role during the ignition delay. Finally, it is found that the predictions are not sensitive to which of the various strain rate models are used.