Purpose: To compare implants with machined vs roughened surfaces placed flapless in totally edentulous jaws and immediately restored with metal-resin screw-retained cross-arch prostheses. Mandibles were rehabilitated with two implants (Fixed-on-2 or Fo2) and maxillae with three implants (Fixed-on-3 or Fo3). Materials and methods: Forty edentulous or to be rendered edentulous patients (20 in the mandible and 20 in the maxilla) were randomised to the machined group (20 patients: 10 mandibles and 10 maxillae) and to the roughened group (20 patients: 10 mandibles and 10 maxillae) according to a parallel group design. To be immediately loaded implants had to be inserted with a minimum torque of 60 Ncm. Outcome measures were prosthesis and implant failures, complications and peri-implant marginal bone level changes evaluated up to 1 year post-loading. Results: Flaps were raised in four patients from the machined group. Four prostheses on machined implants and three on roughened implants were delayed for loading because a sufficient insertion torque was not obtained. There were no dropouts 1 year after loading. Two maxillary machined implants were lost in two patients (difference in proportions = 0.10; 95% CI = -0.03 to 0.23; P (Fisher's exact test) = 0.487); one maxillary Fo3 prosthesis on machined implants and one mandibular Fo2 prosthesis on roughened implants had to be remade (difference in proportions = 0; 95% CI = -0.14 to 0.14; P (Fisher's exact test) = 1.000). Five patients with machined implants had six complications vs seven patients who had eight complications at roughened implants (difference in proportions = -0.10; 95% CI = -0.38 to 0.18; P (Fisher's exact test) = 0.731). There were no statistically significant differences for implant failures, prosthetic failures or complications between groups. There were no statistically significant differences for marginal peri-implant bone levels between the two groups (estimate of the difference = -0.06 mm; 95% CI = -0.23 to 0.10; P (ANCOVA) = 0.445), with both groups losing marginal bone in a statistically significant way (0.35 ± 0.23 mm for machined and 0.42 ± 0.27 mm for roughened surface). Conclusions: These preliminary results suggest that immediately loaded cross-arch prostheses can be supported by only two mandibular or three maxillary dental implants at least up to 1 year post-loading, independently of the type of implant surface used. Longer follow-ups are needed to understand whether one of the two-implant surfaces is preferable.