Are CCDs approaching the sensitivity of photomultiplier tubes? I still see photomultiplier tubes (PMT) used for even benchtop spectroscopy experiments.  How much more sensitive are they to CCDs?  
I found a post which claims about 1.5 x more sensitivity for PMTs, but any references or more complete understanding of the problem would be really appreciated.  
I'm also wondering if they need to be cooled or specialized chips or if commodity imaging chips like from a camera has reasonable S/N characteristics for detection?  
 A: Photo-multiplier tubes, specially those with III-V semiconductor photo-cathodes, are able to detect single photons, in the main part of the visible spectrum. (blue to red)
CCDs are not capable of doing that.     
Gallium Arsenide, and Gallium Phosphide are the principal III-V semi-conductor photo-cathode types.  The highest quantum efficiency is obtained with "Dormer Window" types, where the light is introduced at grazing angles to the photo-cathode, so the photons see a thicker layer of absorbing material than the resulting emitted photo-electrons do, as they are extracted from the face of the cathode film.     
Cooling of the photo-multiplier tubes is often used to reduce thermionic emission from both the photo-cathode and the multiplier dynodes.    Dry ice, and liquid nitrogen being two available methods.   Some photo-multiplier tubes also include accessory magnets, often ceramic, to steer the emitted photo-electrons, to the multiplier structure, while generally deflecting cathode thermionic emission electrons, which come from a wider area of the cathode.   This results in lower noise.
