We compare two approaches for the accurate calculation of the energy levels of weakly bound boson trimers. The first approach is based on correlation function Monte Carlo employing optimized trial functions, while the second approach is based on the discrete variable representation. A trimer with atoms of half the mass of neon is used as a test problem for benchmark calculations. The two approaches yield identical results, within error bars, for all the J = 0 energy levels below the dissociation threshold. The relative merits of the two techniques are discussed, and a perspective is given for extension to larger clusters.