Noninvasive ventilation (NIV) during exercise has been suggested to sustain higher training intensity but the type of NIV interface, patient-ventilator asynchronies (PVA) or technological limitation of the ventilator may interfere with exercise. We assessed whether these parameters affect endurance exercise capacity in severe COPD patients. In total, 21 patients with severe COPD not eligible to home NIV performed three constant workload tests. The first test was carried out on spontaneous breathing (SB) and the following ones with NIV and a nasal or oronasal mask in a randomized order. PVA and indicators of ventilator performance were assessed through a comprehensive analysis of the flow pressure tracing raw data from the ventilator. The time limit was significantly reduced with both masks (406 s (197–666), 240 s (131–385) and 189 s (115–545), p < 0.01 for tests in SB, with oronasal and nasal mask, respectively). There were few PVA with an oronasal mask (median: 3.4% (1.7–5.2)) but the ventilator reached its maximal generating capacity (median flowmax: 208.0 L/s (189.5–224.8) while inspiratory pressure dropped throughout exercise (from 10.1 (9.4–11.4) to 8.8 cmH2O (8.6–10.8), p < 0.01). PVA were more frequent with nasal mask (median: 12.8% (3.2–31.6), p < 0.01). Particularly, the proportion of patients with ineffective efforts > 10% was significantly higher with nasal interface (0% versus 33.3%, p < 0.01). NIV did not effectively improve endurance capacity in COPD patients not acclimated to home NIV. This was due to a technological limitation of the ventilator for the oronasal mask and the consequence either of an insufficient pressure support or a technological limitation for the nasal mask.