Sixstring and Toaster both are good people to listen to on this subject. I think if it were me in your situation.......wait, I am!
So my 2 cents.......there are a bunch of different metrics to consider like PFF, PAR watts, umols, and a hole bunch of scientific theory to explain why, in theory, x amount of watts is needed vs what you would use in HPS or other conventional lighting systems. In my humble opinion, this information can only take you so far. But I think you know this as you are already here asking questions like me. SOOO, what does matter? The only thing I really think matters at the end of the day is real world results, replicate-able and repeatable results. There are number of individuals doing this with varying degrees of success using a wide variety of LED technology. As far as cxb3590's are concerned, I would look at growmau5 on youtube if you want some really good real world results using this technology. What sixstring and toaster are saying is right in line with what Growmau5 recommends in his videos.....little surprise since it was sixstring that introduced me to growmau5.
To that effect my only critique is of size and density of the produced herb I am seeing. As such, seeing as everyone else is going with the 1 COB per square foot theory, I think I will do the american thing and push for more power. So, here in another couple weeks I will be starting a side by side pitting 16 cxb3590's (800 watts) vs 1000 watt digital HPS. I very well may find that past 1 COB per square foot there are diminishing returns though.....very very possible.
If I was more concerned with budget and trying to find the least amount of COB's to get the job done, I would probably build modular. Make a 4 COB setup and see how it goes. If you need more, get more. The 185h drivers are less expensive per COB than the 240h drivers....so it may be better to build in 4's than 5's....but whatever floats your boat! Just please post as I know I and probably several others would be interested to see your results with whatever you decide on.