New saturated critical heat flux (CHF) data have been obtained experimentally in two different multi-microchannel heat sinks made in copper with three low pressure refrigerants (R134a, R236fa, R245fa). One of the test sections had 20 parallel rectangular channels, 467 µm wide and 4052 µm deep while the second had 29 channels, 199 µm wide and 756 µm deep. The microchannels were 30 mm long in the flow direction where a 20 mm length in the middle was heated with an electrical resistance deposited on a silicon plate. Base CHF values were measured from 37 to 342 W/cm2 for mass velocities from 100 to 4000 kg/m2s. When increasing the mass velocity, CHF was observed to increase while the rate of increase was slower at high velocities. While CHF increased moderately with large inlet subcooling (e.g. 20 K) in the H = 4052 µm channels, inlet subcooling seemed to play less a role as the channel size decreased. CHF showed reversed tendency with increasing inlet saturation temperature (10 ≤ Tsat ≤ 50°C) for the two test sections. The experimental data were compared with existing prediction methods. The data demonstrated good agreement with several predictive methods using the heated equivalent diameter Dhe and the actual mass velocity Geq to implement the circular tube correlations. Flow visualization was conducted with and without an orifice at the inlet of each microchannel. Visualization confirmed the existence of flow instability, back flow and non-uniform distribution of flow among the channels when the orifices were removed. Flow patterns in the microchannels and their evolution with increasing heat flux were observed. The flow boiling curves suggest that in the case of omitting the orifice, the boiling incipient occurred at a higher heat flux resulting in a higher overshoot of wall temperature. Moreover, the flow was easily subjected to instability and caused CHF to occur at much lower values than occurred with the orifices in place.