Purpose: To report retro-prospective, long-term data on the prevalence of implant failures related to maxillary and mandibular arches and to different implant surfaces in a large number of edentulous patients. Materials and Methods: Altogether, 3,493 and 1,092 edentulous arches were consecutively treated with implants with turned (1986-2002) or moderately rough (2003-2015) surfaces, respectively, during two time periods at one referral clinic. All implant failures were consecutively identified during routine follow-up, and a multivariate logistic regression analysis was performed to analyze implant failure related to arch and implant surface. Results: Overall cumulative survival rates (CSR) for arches treated with turned surface implants were 75.7% and 94.6% for the maxilla and mandible, respectively. The corresponding 10-year CSRs for arches treated with implants with a moderately rough surface were 91.9% and 96.1%, respectively. The strongest significant association (P .05) with risk for implant failure was the maxilla, and this was more pronounced for implants with a turned surface. Age at surgery, implant surgeon, calendar year of surgery, and time of follow-up also had significant associations with risk of implant failure (P .05). Conclusion: Risk for implant failure was significantly higher for treatment in the maxilla, but this risk was decreased significantly when using implants with a moderately rough surface. The impact of surface was not so obvious for treatment in the mandible. Risk for late implant failures after the first year was lower for implants with a moderately rough surface in the maxilla, but this risk seemed to be comparable for the different surfaces in the mandible.