We evaluate a simple "cube"device for measuring Mean Radiant Temperature (MRT) as an alternative to the globe thermometer. The device uses 6 orthogonal non-contacting surface temperature sensors on the faces of a cube to evaluate the MRT at a point. Each sensor consists of a thermopile and IR filter and internal circuitry to output a temperature based on blackbody emission from surfaces, and are not sensitive to air movement, which a globe thermometer must be calibrated for. The 6 orthogonal measurements are averaged to estimate the Mean Radiant Temperature, and opposing sensors on the cube structure are used to produce radiant asymmetry data. We compare this data to a standard black globe under conditions where free and forced convection cause the globes to be very inaccurate as we have shown in previous work. This paper demonstrates the 6-sided cube can eliminate the error of globes due to sensitivity to changing air movement and convection. We experimentally demonstrate that the globes are limited by self-induced buoyancy flow, with the current ASHRAE standard method resulting in 50% lower separation of MRT from the air temperature (24 deg C air temp, 29 deg C globe MRT, and 35 deg C cube temp). A proposed mixed convection model for globe correction also produces a result that agrees better with the cube results. Finally we propose future work to better consider how overlapping fields of view of the 6 sensors affect results as evaluated by rotating the device 45 degrees.