The aim of this study was to assess the accuracy of a surface-guided radiotherapy (SGRT) system for setup and intra-fraction motion control in frameless non-coplanar stereotactic radiosurgery (fSRS) using actual patient data immobilized with two different types of open-faced masks and employing a novel SGRT systems settings. Forty-four SRS patients were immobilized with two types of open-faced masks. Sixty lesions were treated, involving the analysis of 68 cone-beam scans (CBCT), 157 megavoltage (MV) images, and 521 SGRT monitoring sessions. The average SGRT translations/rotations and 3D vectors (MAG-Trasl and MAG-Rot) were compared with CBCT or antero-posterior MV images for 0° table or non-coplanar beams, respectively. The intrafraction control was evaluated based on the average shifts obtained from each monitoring session. To assess the association between the SGRT system and the CBCT, the two types of masks and the 3D vectors, a generalized estimating equations (GEE) regression analysis was performed. The Wilcoxon singed-rank test for paired samples was performed to detect differences in couch rotation with longitudinal (LNG) and lateral (LAT) translations and/or yaw. The average SGRT corrections were smaller than those detected by CBCT (≤0.5mm and 0.1°), with largest differences in LNG and yaw. The GEE analysis indicated that the average MAG-Trasl, obtained by the SGRT system, was not statistically different (p=0.09) for both mask types, while, the MAG-Rot was different (p=0.01). For non-coplanar beams, the Wilcoxon singed-rank test demonstrated no significantly differences for the corrections (LNG, LAT, and yaw) for any table rotation except for LNG corrections at 65° (p=0.04) and 75° (p=0.03) table angle position; LAT shifts at 65° (p=0.03) and 270° (p<0.001) table angle position, and yaw rotation at 30° (p=0.02) table angle position. The average intrafraction motion was<0.1mm and 0.1° for any table angle. The SGRT system used, along with the novel workflow performed, can achieve the setup and intra-fraction motion control accuracy required to perform non-coplanar fSRS treatments. Both masks ensure the accuracy required for fSRS while providing a suitable surface for monitoring.