i did a QB test in the sphere (which was weird trying to compare a qb to a cob due to the different format factor -with the size of the sphere the cob tests are usually single chip outside of the sphere shining into it- i ended up mounting a clu048 ideal holder on the back of the QB, using the QB as the heatsink, and flipping it over to test- in other words the board was still inside the sphere absorbing photons).
i did some low current tests and with the old S5 boards it took an estimated 3 cxm22 or 5 gen5 1212s to match it, as a caveat the first graph is real data the second is calculated by multiplying the cob curve wattage by 3 and 5 respectively to represent multiple chips :
View attachment 3910240
on chip cost performance cobs win but theres cost of heatsinks to consider as well which can tip the scale in QB favor. this was as far as i got when i heard the S6 in QBs were coming out i gave up on testing. a fair test would be to use the same slate heatsink and mount 5 or 6 1212 gen 6 on it and map out a tent test vs S6 QB (and any other board comers). cobs were def at a thermal disadvantage in this test using the thin QB board itself as a heatsink... both cobs and QBs would perform better on a slate but how much better for each remains to be seen