The blackbody radiation limit has traditionally been set forth as the ultimate performance limit of thermal detectors. However, this fundamental limit assumes that the detector absorbs uniformly throughout the thermal spectrum. In much the same way as photon detectors can achieve very high D* because they do not absorb photon energies below their bandgap, so too can thermal detectors except that thermal detectors are not limited to cryogenic operation. In both cases, the enhanced theoretical D* is achieved because the radiation noise is reduced in a device that does not absorb at a uniform high level throughout the thermal emission band. There are multiple ways to achieve high D* in thermal detectors. One is to use materials that absorb only in a certain spectral range, just as in photon detectors. For example a detector made from PbSe, with proper optical coupling, absorbs only photons with wavelengths shorter than 4.9μm. The radiation limited detectivity of such a device can theoretically exceed 9 × 1010cmHz1/2/W in the MWIR. Even with Johnson and 1/f noise estimates included, it can still exceed 2.5×1010cmHz1/2/W in the MWIR. Another technique, applicable for narrowband thermal detectors, is probably even more powerful. Consider a thermal detector that is almost completely transparent. Here, the radiation noise has been reduced but the signal has been reduced even more. However, if the device is now placed inside an optical cavity, then at one wavelength and in one direction, the nearly transparent detector couples to the cavity resonance to absorb at 100%. Radiation from all other wavelengths and directions are rejected by the cavity or are absorbed only weakly by the detector. It is shown that theoretically, the D* of these devices are roughly proportional to the inverse square root of the spectral resonant width under certain conditions. It is also shown that even including Johnson noise and 1/f noise, the practically achievable D* approaches or exceeds 10 11 cmHz1/2/W.