This paper describes a summary of final prediction results by system-scale safety analysis codes during the OECD/NEA/CSNI ISP-50 exercise, targeting a 50% Direct Vessel Injection (DVI) line break integral effect test performed with the Advanced Thermal-Hydraulic Test Loop for Accident Simulation (ATLAS). This ISP-50 exercise has been performed in two consecutive phases: "blind" and "open" phases. Quantitative comparisons were performed using the Fast Fourier Transform Based Method (FFTBM) to compare the overall accuracy of the collected calculations. Great user effects resulting from the combination of the possible reasons were found in the blind phase, confirming that user effect is still one of the major issues in connection with the system thermal-hydraulic code application. Open calculations showed better prediction accuracy than the blind calculations in terms of average amplitude (AA) value. A total of nineteen organizations from eleven countries participated in this ISP-50 program and eight leading thermal-hydraulic system analysis codes were used: APROS, ATHLET, CATHARE, KORSAR, MARS-KS, RELAP5/MOD3.3, TECH-M-97, and TRACE.