With the driver dimmed all the way down, driver was pulling 24W (4W per cob) from the wall. Took lumen measurement @ 12". Increased power until I got double the lumens, recorded power output from the killawatt and it was 51W (8.5W per cob). That's really not bad (48W would have been perfectly linear increase). Increased power to get a lumen reading that was 3X of the original (24W reading), and it took 83W (13.8W per cob) of power to achieve. So at that point its falling off a bit in efficiency.
This supports my original thought when I started, cheap no names can put out good efficiency ~10W. The real question, as someone else noted, is about light quality (CRI).
