Explain to me how they are doing anything if the temperatures are the same before and after a radiator? I'd love to explain it to my old thermodynamics teacher, I'm sure he'd get a good laugh at it too.
I'll give you a little lesson. Energy. The life blood of the world...
...I'll skip all the fluff and get to it. Assuming each GPU is dissipating 200W of power, the water collects that 200W of power into its stream. 4x 200W = 800W. I'm not sure what flow you have, I'll use 1.5GPM (341LPH). Since heat capacity of water is given in J/kg/K, we need to convert 341LPH into kg/s (J=W*s) = .0947kg/s. I'll add here that since the density of water is 1kg/L, all you need to do is divide LPH by 3600. Heat capacity of water = 4182J/kg/K = 4128W*s/kg/K. (4128W*s/kg/K)(.0947kg/s) = 391W/K. So we'll simply use 391W/°C. I assumed ~2GPM earlier, because that is what we typically use at work. I'm not sure if PC water cooling usually flows more or less than that, but that is quite a bit of liquid.
Since we have 800W total, 800W/391W/°C = 2°C increase in temperature.
Who knows, maybe your thermocouples aren't that accurate. If you want to post your GPM, I can get a more accurate answer.