The sky is falling!!! 
Cree has done this to us before (XM-L2, anyone?). Cree will do it to us again. We are enthusiasts who are running these LEDs far outside of their ratings, so anything could change for us at any time. Cree rates the XP-G2 at maximum of 1.5A. That's all the guarantee, so when we can only get 3.5A instead of 4.5A on a single battery well, that's our problem. Same goes for the slight die changes and dedomed emitter performance / tint shift.
I’m pretty sure those are not genuine cree LEDs, changing the die size without notice would be PIA for light designers because existing optics for XP-G2 would not give desired beam angles anymore,which means product redesign; it’s hard to believe that any LED manufacturer would do that without special notice on website and in datasheet.
There’s a press release dated 8th March 2016 specifically saying that the XP-G2 and XP-L are now up to 9% brighter.
led4power
Its a common issue with those big companies, they simply dont give a damn, they can affort it, they are huge
I usualy work with Phillips, Osram est in my work and when we just manage to apply a product in a proiect they just change it without a notice, yet the customer do care
In that case, the emitting angle isnt fixed, the flux isnt fixed too, everything has its” confort for the producer” margins, so i am quite sure its a genuine product
I received the leds from LI. I hope to test them on monday (no guarantees), but already I did a visual comparison with an old batch XP-G2 S4 2B from intl-outdoor with the stereo-microscope tonight, left old XP-G2 S4 2B, right new XP-G2 S4 2B (when mouse-over works again I will insert it):
At least it is clear that the production proces has undergone a change, as LI already showed, and led4power already reported a year ago: the new die seems slightly bigger (dedoming will give exact numbers), the number of dots on the die have increased, the phosfor is rougher (not clear on the picture, very apparent in real life), the tint of the phosfor is a bit more orange with small red bits in it which are absent in the old die. This suggests a warmer tint, and this suggestion gets stronger when the boards are inspected under 365nm light (left old, right new, btw the Noctigon boards are clearly another batch too, a different red paint is used):
But when the leds are lighted up (in series on a led-tester, picture is at ~1mA, but at 50mA the difference is the same), the tint difference is not dramatic, the old led a bit more green, the new led a bit more red, but I would place them at the same colour temperature. If both would be 2B I can not estimate by eye. I have a small colour tester coming in next month that can tell more about that.
I'm curious about any output differences. More about that to come!
Output results.
I tested the new XP-G2 S4 2B for voltage and output and plotted it against an earlier test of the XP-G2 S4 2B. I followed my usual method. Some remarks:
*these are tests of single leds, mind that results will vary among equally spec'ed leds within the voltage and output bin-classes
*since last XP-G2 S4 2B test my current measuring has changed a bit, the same current reads 0.1A lower now, this is not corrected in the graph
*since last test I cleaned the inside of my integrating sphere, it reads ~1% higher now, this is also not corrected in the graph.
What can be seen:
-If you correct the new numbers to 0.1A higher current and 1% lower output, the old and new led have virtually the same output up to 3A. Above 3A the old led seems to gradually perform better, so the new led seems to have a slightly worse thermal resistance. But it must be said that I did not reflow the leds myself (based on own experiments I use solder quite generously) , both were done by intl-outdoor: a thermal pad with somewhat sparse solder could already account for this difference.
-The voltage is a different story, once again in a new batch the voltage went up: by 0.1V at 1.5A, and 0.2V at 4A. This is of course killing for single li-ion use in flashlights, and this may be the single most important cause of the lower performance reported by LI. There is of course the uncertainty of measuring just one led, it could be a particularly high voltage one.
-At 1.5A, which is the highest current that these leds are officially rated for, the performance of the new batch XP-G2 S4 2B is great, a somewhat higher voltage but I doubt that it falls out of specs with that :-(
I'm not sure if I need to proceed with dedoming and further measurements, the results sofar explain IMO pretty well what is going on with this led...
I could just be that for general lighting this new led is an improvement: the colour reproduction seems to have gone up a bit 
Thanks Djozz,
Don’t be mad but please continue with de domed test when you’ll have time(you can try new 10 minutes hot gas method we talked about)…
I can not show the whole world how ugly and unusable (lower lux) new XP-G2 S4 2B is upon dedoming in combination with fet driver but you can
IO reflows their emitters well… They are all like this even in my redneck lux meter tests 
so I think we can remove “uncertain measuring of one led”…
Do all 12 of them and I am pretty convinced same result will appear…
Here’s is a photoshopped picture from above, the leds were in the same picture so with the same magnification, with dome on. The difference could be caused by the die size but the dome curvature could also be somewhat different.
This is all the time I have today for this, maybe next week…
In general I have more time for talking on BLF (I can do that on my tablet while leisurely sitting on the couch in a sort of social configuration with the family ) than time for doing mods and tests (for that I retreat to my ‘man corner’ of the living room sitting with my back towards the family and with my mind completely absent)
Nice working space… Mine looks like a living mess comparing to yours
Thank you very much but since I know you are thrower flashlight fan I am sure you’ll try new dedomed xp g2 s4 2b
All I can do now is start searching for better or old S4 emitters and thanks to N10Sivern for selling me some…
Big thanks to Djozz for testing this.
P.S.
Hope you’ll manage to test new xpg2s42b “Hulk” version in near future
What older versions are the closest to s4 2b when dedoming?
I mean good output and tint.
Thanks djozz, once again! Excellent info. From what’s been described, the Vf increase at high amps is the main problem, as you said. The tint is another issue related to the design change CREE apparently did.
Hope they don’t start killing the XP-L’s in the same way - I’m afraid this might be a new design change, a shift in manufacturing the dies. Least with the XM-L2’s, the highest U4 bins seem to come back in performance with a lower Vf than the U2’s of the 2nd generation, and U3 generation.
Can anyone provide old s4 2b ?
2pcs , if anyone is somewhere near me.
I am also in search for old s4 2b. I need at least 50 of them… So guys please take a look at your drawers please
So… the Vf is higher… so its less efficient… the die is bigger… so it throws less… it cant handle being “overdriven” as much… and it goes a horrible bright green (albeit a VERY intense hotspot of green) after being dedomed…
…and all this so that its ever so slightly brighter… Doesnt sound like an improvement to me…
I think I might just go with XP-L HI LEDs from now on… not such a tight beam as DD XP-G2’s, but at least the total lumen count is higher to compensate…
Sure,
Nothing wrong with XP-L ’s except throw potential…
We must see results of de domed 70CRI Nichia 219C or Osram Oslon 3rd Gen 6500k. They are both better performer(in dome on version) tha XP-G2 S4 2B as reported by Djozz…
XPL HI is a bad idea, at least at the current HI development state, an example
Those are impossible to be found, i actualy have never seen that listed anywhere
stock of emitters
Thought I would share this if you need older stock
List
2B S2 Reel date Sept 2014
1D S3 Reel date Sept 2015
3C S2 Reel date April 2015
1C S3 Reel date June 2015
1D S4 Reel date NOV 2015
2B S4 Reel date AUG 2015
2C S4 Reel date AUG 2015
1A S2 Reel date JULY 2015
Cheers
Mark
I see Hank’s reel is dated 25 Feb 2015? Are you guys looking for older than this? Do we know when these issues started?