Memory usage and anticipating it.

[pvanheus]

I am trying to predict the likely memory consumption of IQ-TREE jobs and noticed that memory is reported in two places. First (if -m TESTNEW or -m TEST is used) the memory required for ModelFinder (NOTE: ModelFinder requires XX MB RAM!) and then soon after the completion of ModelFinder (NOTE: YY MB RAM (ZZ GB) is required!). In all of the runs I have done, the amount required for ModelFinder is larger than that reported later. I have a few questions:

1. Is the second report of required memory how much would be needed for tree building? I.e. if ModelFinder is not used, then this is a reasonable estimate of memory required?

2. Is there a way to perform these estimates in advanced of submitting an IQ-TREE run? Alternatively, is there a database of examples of alignments, parameters and associated RAM requirements?

Thank you for any insights!
P.S. some example runs from my side:

1. 1828 sequences with 76531 columns, 57157 distinct patterns 38300 parsimony-informative, 38231 singleton sites, 0 constant sites - this is with a SNP alignment and --seqtype DNA --safe -m TESTNEW --msub nuclear --ufboot 1000 -fconst 749408,1420820,1415626,749147.
   ModelFinder: 32749 MB RAM, second estimate: 13772 MB RAM

2. 129 sequences with 14171 columns, 4051 distinct patterns 5840 parsimony-informative, 8331 singleton sites, 0 constant sites - another SNP alignment, --seqtype DNA --date dates.tsv --safe -m --msub nuclear --cmin 2 --cmax 10 --merit AIC --ninit 100 --ntop 20 --nbest 5 --nstop 100 --radius 6 --perturb 0.5 --sup-min 0.0 --ufboot 1000 --nmax 1000 --nstep 100 --bcor 0.99 --beps 0.5 -fconst 756176,1443569,1438045,756273
   ModelFinder: 185 MB RAM, second estimate: 117 MB RAM