Bugs vs Stan for areal data

Hi everyone,
I’m implementing a CAR proper model (http://mc-stan.org/users/documentation/case-studies/mbjoseph-CARStan.html like this) in STAN, i have a 3700 areal data, so at the start the command multi_normal_prec was too slower (3 days for a simulation) but with the sparse implementation (in the link) the simulation is much faster.I’m new in bayesian analysis,i started with STAN, and i’m asking, where STAN could be considered better than Bugs in this environment? ( i think, but i didn’t tried for problem of time, that BUGS in this context could be much slower than STAN)

For specifically this type of model, BUGS actually works ok, but we did some tests (there’s a very long thread here somewhere) and if memory serves, the implementation in that case study was faster than the one in BUGS. So in that sense it’s better. But BUGS is “sufficient”.

Can you tell me where is this thread? I searched but i’m not able to find it;thank’s a lot

I think it was this one Case study on spatial models for areal data - Poisson CAR/IAR