Understand throughput drop between old simulation and 2024 min. bias
A throughput test on the 2024 MB sample has been recently added, using the hlt1_pp_forward_then_matching_no_ut sequence:
The throughput test reports 68 kHz on the new sample, vs 80 kHz when using the old MB sample (upgrade_mc_minbias_scifi_v5_retinacluster_000_v1_newLHCbID_new_UT_geometry).
Looking at the breakdowns, it seems like the scifi decoding is now a significant component, while in the past it had a small contribution. The two samples have different decoding versions (v5 for the old one, vs v8 for the new one).
Any ideas?