Purpose We compared gross characterization of intracranial dural arteriovenous fistulas (DAVFs) between unenhanced 3-tesla 3-dimensional (3D) time-of-flight (TOF) magnetic resonance angiography (MRA) and digital subtraction angiography (DSA). Methods We subjected 26 consecutive patients with intracranial DAVF to unenhanced 3T 3D TOF MRA and to DSA. Two independent sets of observers inspected the main arterial feeders, fistula site, and venous drainage pattern on MRA and DSA images. Interobserver and intermodality agreements were assessed by k statistics. Results Interobserver agreement was excellent for fistula site (κ = 0.919; 95% confidence interval [CI], 0.805 to 1.000), good for main arterial feeders (κ = 0.711; 95% CI, 0.483 to 0.984), and very good for venous drainage (κ = 0.900; 95% CI, 0.766 to 1.000). Intermodality agreement was excellent for fistula site (κ = 0.968; 95% CI, 0.906 to 1.000) and good for main arterial feeder (κ = 0.809; 95% CI, 0.598 to 1.000) and venous drainage (κ = 0.837; 95% CI, 0.660 to 1.000). Conclusion Gross characterization of intracranial DAVF was similar for both imaging modalities, but unenhanced 3T 3D TOF MRA cannot replace DSA.