The original High Light boards were 2.5 umol/j which was surprisingly high for their time and also considering they were CRI94-95. The Optisolis and Seoul Semiconductor UV Sunlike LEDs were not that efficient (around 2.1-2.2 umol/j), but the 2700K CRI90 Nichias are still up there (2.75 umol/j) which dragged the efficiency up – which is why I used them.
We managed to get 20% better efficiency from the High Light Light 420 boards, but that was only after doing a lot of individual LED testing with LED Teknik and mainly because we had access to some 3.94 umol/j 660 monos. We now have access to a new batch of 660s that are testing 4.1-4.2 umol/j, so the Gen 3 (or whatever it is we come up with next) will be even more efficient.
Overall, however, it is really hard to get that kind of efficiency from ANY LED that contains UV and Far Red. The 405nm Nichias we use now are 71-72% efficient but still only around 2.3 umol/j because they have a very low QER of 3.36 (the theoretical maximum umol/j for a spectrum if the diode is 100% efficient).
720nm Far Red diodes, on the other hand, are still only around 55% efficient – even, though they have a high QER of 5.99