OBJECTIVE Three-dimensional echocardiography (3DE) is becoming increasingly common in clinical environments. However, the quality of the images depends on the acoustic window, and it can be difficult to identify the endocardial borders. The objective of this study was to evaluate the performance of 3DE in determining the volumes and left ventricular ejection fraction in unselected patients against the gold standard, cardiac magnetic resonance (CMR). METHODS In 47 unselected patients who underwent CMR, we performed 3DE using a real-time acquisition technique and semiautomatic border detection. RESULTS We excluded 4 patients (8.5%) because they had an extremely deficient acoustic window. In the remaining 43 patients, including those with a suboptimal acoustic window, we obtained the following correlations between 3DE and CMR: end-diastolic volume, 0.71; end-systolic volume, 0.77; ejection fraction, 0.74. Although the end-diastolic volume was systematically underestimated, no significant differences were observed in the ejection fraction. When the 11 patients with suboptimal acoustic windows were excluded, we observed a systematic underestimation of the end-diastolic and end-systolic volumes, which paradoxically gave rise to improved correlation coefficients (0.79, 0.92, and 0.84, respectively) and a more accurate ejection fraction. CONCLUSIONS Compared to CMR, 3DE systematically underestimates the ventricular volumes but enables adequate determination of the left ventricular ejection fraction regardless of the quality of the acoustic window.