100 corsika files from simulation run 1540 were processed by ucr+mmc+ppc+fat-reader (fat-reader was used to apply the LC condition and create triggered events) in 2 hours 2 minutes on cudatest (using all 6 GPUs). This is about 73 seconds per 4.0138 second run (translates to spending about 18.2 hours of calculation time for simulation of 1 hour of detector lifetime); compare with the average of 3 hours 19 minutes per run spent by photonics-based production. A part of the difference is due to re-using corsika files for the ppc production.

Figures below are for both SPICE model and nominal AHA ice, first shown for the 90% efficiency scaling (usually applied in standard production as a 10% cable shadowing correction), and then for 100% scaling. This link has plots for the ice with the tilt parametrized from the dust logger data (by Ryan Bay).

<--- SPICE model --->
Efficiency: x 0.9 no rate adjustment	hole ice	data: black ppc: blue
<--- nominal AHA ice --->
Efficiency: x 0.9 no rate adjustment	hole ice	data: black ppc: blue
<--- SPICE model --->
Efficiency: x 1.0 no rate adjustment	hole ice	data: black ppc: blue
<--- nominal AHA ice --->
Efficiency: x 1.0 no rate adjustment	hole ice	data: black ppc: blue