No- I think what I am trying to say is the opposite of this. There is a huge difference in radiant output power (light) between a true 1W power die and the collection of little chip dies that are placed on a metal core PCB and these LED sellers are claiming as '3W LED's'. They are not '3W LED's', thay are maybe '3W modules' or '3W arrays', but definitely NOT '3W LED's'- you can't take a bunch of inefficient chips, bunch them together, and have them get any more efficient. All, and I mean ALL, gains made in LED light output have come about as a result of increases in 'injection efficiency'- the amount of power in vs. the amount of usable light out, vs. the amount of heat produced. Underestimate the importance of thermal management of the LED at your own risk- nearly any and all problems you will ever have with an LED are thermally related.
At this stage of LED development, the approx. 1mm^2 1W power die is the at the locus of injection efficiency, cost/ wafer yield efficiency, and thermal manageability. While it is true that there are 'some' highly efficient, true 3W dies being produced by a few companies, these are terrifically expensive and not available in the wavelengths necessary to satisfy the photosynthetic demand of the plant. And while it is also true that the materials used to create the semiconductor junction (which actually creates the light), for example, InGaN, is similar for all manufacturers, the efficiency gains made by the leading manufacturers have come about as improvements in process and materials are are covered up by corporate IP. For the 'lesser' LED manufacturers, there is little they can do to improve their products without ending up in court- these LED guys are always at each others' throats suing each other. A company like Epistar, for example, has focused on AlInGap (and similar) process improvement. For a company like Everlight, they simply make all the oddball wavelengths no one else wants to touch and sell them cheap. What else can they do? Bridglux is feverishly working on a silicon process improvement- another story all together.
And yes, the difference in light output per Watt injected is huge- the peak radiant flux output from from a set of binned Cree dies (not the 'theoretical' output but the real output when you actually purchase them) is perhaps 2X to 3X that of an Epistar red die for every Watt injected. And Epistar is actually a semi-respectable maker, Everlight is a bit of a joke.
So my point would be that true grower gains in g/ Watt calculations starts with the LED die used in the system itself, and carries through to the thermal management scheme as well as the optical system employed by the designer. While it may be true that you can use and have success with an LED fixture which retailed for $2.50 or $3.00/ W, you will probably be consuming the same amount of power for similar results that you could get using HPS. The true potential of LED is not being well represented with these fixtures- and yes, I believe that a well designed LED light system will pay not only for it's high initial cost vs. HID, but also vs. cheaper LED versions. The last statement is, admittedly, opinion, and also, I am not shilling for Cree- there are other makers of LED dies with, for the most part, similar performance characteristics, I just use them as example because everyone in this market seems to be familiar with them.
