Purpose: This study compared three-dimensional (3D) accuracy of conventional and digital workflows in the fabrication of implant FAFDP frameworks for an edentulous mandible. Materials and Methods: A heat-polymerized polymethyl-methacrylate master model simulated an edentulous mandible to be restored with a fixed prosthesis supported by five implants (A-E). A conventional open-tray, splinted-coping polyether impression was poured with Type IV dental stone to produce a stone model, which was scanned with a dental laboratory scanner to produce a virtual model. Full-contour (ZFC, n=5) and cutback (ZCB, n=5) zirconia frameworks were fabricated with CAD/CAM. The stone model was used as a working model to fabricate conventional noble metal frameworks (CNB, n=5). Each test  framework was attached to five implants with 35Ncm torque application. Test models were fabricated with Type IV dental stone and allowed to set completely before removing the test frameworks. Centroid positions and central axes of the implants in master and test models were measured with a coordinate measuring machine and compared. The local coordinate system comprised of implant A centroid as origin; implants A, C, E as XY-plane; implants A, E as X-axis. Four linear distortion parameters (dx, dy, dz, dR), two angular distortion parameters (dθx, dθy), and 3D distance distortion values (ΔR, %ΔR) were assessed. Results: Magnitude of mean dx, dy, dz, and dR ranged from 39.9±26.4μm (CNB-B) to 263.3±94.6μm (CNB-E), from 56.9±5.4μm (ZFC-B) to 124.5±37.6μm (ZCB-C), from -8.3±27.6μm (ZFCD) to -24.6±17.3μm (ZCB-D), and from 80.6±29.4μm (CNB-B) to 263.3±94.6μm (CNB-E) respectively. Magnitude of mean dθx and dθy ranged from 0.025±0.486 degree (ZFC-C) to 1.490±0.383 degree (CNB-C), and from -0.050±0.171 degree (ZCB-B) to 1.263±0.501 degree (ZFC-C) respectively. One-way ANOVA found differences among groups for dx, dy, dz, dθx, and dθy at some implants, but there was no clear pattern on the poorest group. For global linear distortion, CNB-D fared the worst. CNB exhibited the worst accuracy in ΔR and %ΔR for all reference distances except A-E. Conclusions: There were no differences between ZFC and ZCB for all parameters. CNB fared poorest for ΔR and %ΔR for three of four reference distances, and for dR for one of four implant positions.