LED test / review - New Cree XLamp XM-L2 U3 color kit 51 (≈ 6200 K, design change after PCN issued)

LED test / review


Cree XLamp XM-L2 (New design after PCN issued)


Update 12/26/2022

I have now completely dedomed this LED! Results of this process and also general color and luminance measurements can be found here!

For a number of years the XM-L2, like its predecessor the XM-L, was one of the most widely used LEDs for flashlights. Its good performance coupled with attractive pricing and relatively easy dedoming also made it a good choice for conversions and range records. Increasingly, however, it is mostly being replaced by Osram and Luminus emitters.

A few years ago, Cree introduced a new platform for LEDs, which was finally launched on the market with XP-G3 and XP-L2 in the course of 2016 and 2017. Contrary to a possibly expected production stop of the XM-L2, a PCN (Product Change Notification, link to Mouser) was published in May 2021, which announced a product change for the XM-L2. This concerned the electrical parameters as well as the design. The new design should be delivered from August 2021.

Due to the large deviation from the previously known design, I decided to test this LED and especially to investigate the suitability with classic dedoming in flashlights.

Technical data

Tj 85 °C, If 700 mA

Type: single die
Binning: U3, typ. 320 lm
Color kit: 51 (≈ 6200 K)
Rated voltage: typ. 2.72 V max. 3.15 V
Forward current: max. 3,000 mA
Peak current: ---
Viewing angle: typ. 120°
Thermal resistance: typ. 0.7 °C/W
Max. temperature Tj: max. 150 °C

Official datasheet: Cree XM-L2 (PDF, 1,6 MB)

Note: it is important to note that the current version of the data sheet on Cree's website is only valid for the new version of the XM-L2! For the old version with classic bonding, this data sheet hosted by Mouser (PDF, 1.9 MB) can be consulted. The year for the copyright in the footer of the datasheet can give good information about the release (new: 2022, old version: until 2019).

First appearance

The new version differs significantly from the previous XM-L2. The yellow luminous surface is covered by a clear silicone dome. Around the LED chip is a plain white surface. While the previous design had the contacting wires on the top side (lateral design), the new design is manufactured as a flip chip design, i.e. the contacting is on the bottom side of the LED chip. This design is already known from earlier LEDs, among others the XP-L2 and also from other manufacturers like Nichia with the 319A.

A light gray plus sign marks the anode.

Like the older XM-L2, the new version is 5.00 x 5.00 mm, which means it still has the XM footprint. It is still fully compatible with XM footprint accessories, such as centering aids and boards, allowing easy swapping with an XM LED.

The thermal pad (center contact area) is electrically neutral. The recess in the thermal pad is another marker for the anode.

LED chip and die

The luminous surface is similar to the XP-G3 and XP-L2. A fine granulation with small blue and yellow areas is visible at low current. The mixing of the light of these areas in secondary optics corresponds to the specified color temperature.

The luminous area is 7.90 mm² with dome. It is about 1 mm² smaller than the original XM-L2 minus the bonding, which should improve the luminance.

The lack of bonding wires on the top of the die increases efficiency due to the lack of shading and the LED should be more suitable for focus optics because it provides a cleaner light image without "cut corners".

Power and overcurrent capabilities

Within official parameters:

  • At 3,000 mA (official maximum current): 1,369 lm @ 3.18 V (dedomed: 1,171 lm).
  • Power at official maximum: 9.54 W
  • efficiency at 3,000 mA: 143.5 lm/W (dedomed: 122.7 lm/W)
  • at 700 mA (binning conditions, 25 °C Tj): 379 lm @ 2.80 V (dedomed: 325 lm) - converted to 85 °C according to data sheet and Cree PCT: 345 lm (dedomed: 298 lm)

The XM-L2 U3 tested here not only meets the specified binning (320 lm @ 85 °C/700 mA) but even reaches binning U4. This could indicate a very well running production with high yield and the therefore necessary relabeling to lower binnings, similar to what is sometimes done with computer processors and other semiconductor products.

The performance is simply impressive to say the least. Nearly 4000 lumens and over 15 amps of maximum current show that the past few years have also seen quite a bit of progress on the XM footprint. The XP-L2 is easily surpassed, not to mention the old XM-L2. It should be noted that the XM-L2 tested here was irreparably destroyed at about 5 A by burned-out bonding wires and therefore could not develop its full performance.

The thermal resistance of 0.7 °C/W stated in the data sheet should be plausible, since even the performance of the XP-L2 is clearly surpassed. However, since the maximum current is significantly below the XHP70 (0.9 °C/W), it can be assumed that the heat dissipation is limited by the area of the thermal pad in the XM footprint and thus limits the maximum current.


  • Maximum reached at 16.8 A, at this point 3,833 lm @ 4.48 V (dedomed: 16.2 A @ 3,231 lm / 4.44 V)
  • Power at maximum 75.3 W (dedomed: 71.9 W)
  • Sweet spot at about 9 A (3,036 lm @ 3.85 V), dedomed 8.6 A (2,510 lm @ 3.80 V)
  • Power in sweet spot 34.65 W (dedomed: 32.7 W)
  • Efficiency at maximum 50.9 lm/W (dedomed: 42.2 lm/W)
  • Efficiency in the sweet spot 87.6 lm/W (dedomed: 76.8 lm/W)

Luminance and dedoming

Werte at 25 °C Tsp, 85 °C Tsp values 6-14 % lower

Update 12/06/2022: Updated luminance chart due to insufficient measurement!

Due to inadequate measuring devices with low accuracy and measuring methods the values for the XM-L2 new design were way too high, in the table above are the values taken with the much better equipment and more accuracy in taking the measurements. Please taking this to consideration if you mod some lights with this LED! In the future the luminance measurements will be done with this new setup with seems to be much more accurate.

Now the luminance is not as high as before, more in the range of XP-L HD or XP-L2, which seems realistic considering the LED design. (The high luminance values I got in the first measurement was also the final reason to revamp the whole setup and equipment.) Even the older XP-L HI has a much higher luminance at lower currents. For far reaching lights this might not be the best LED to built in. Still it should be noted that even in direct drive on a potent 18650 should not be achieved more than 6 to 7 A, whereby other LEDs are at an big advantage in luminance, like the good old Oslon Black Flat or newer Oslon CSLNM1 / PM1 emitters.

Dedoming is possible, I shaved the LED. The silicone is harder and more difficult to cut than earlier XM-L2s, requiring a thorough approach. Dedoming changes the optical parameters, which affect luminous flux and luminance.

The luminous area here is 4 mm², just like the old XM-L2. However, the percentage difference in the luminous area after the dome is much smaller.

The stronger glow around the die is noticeable than with the existing dome. Presumably, the die is similar to those of the XP-L2 and -G3, where light is also emitted laterally and thus effects on optics and light image are to be expected. Due to the different light emitting area, the die does not correspond to that of the XP-L2, possibly a new (Cree-exclusive) chip platform is used here.

What is interesting here is the small difference in luminance after dedoming. The luminance increases by not even 10%, despite the reduction in luminous area. This was confirmed again in my second measurement, because the 2nd sample has almost the same performance like the first one. I suspect that the luminance and the luminous flux are increased so much by the bundling already present in the LED that dedoming simply no longer offers any real advantage here.

It is obvious that LEDs of the latest generation are increasingly being optimized for luminance with the highest possible efficiency at the same time. This should primarily serve to reduce the size of the package without loss of luminance while at the same time maximizing efficiency, which is particularly relevant in the automotive market. This also becomes clear in comparison with the dedomed LED - the dome obviously has a different refractive index than in the earlier XM-L2, because it focuses the luminous surface much more narrowly, which can be clearly seen in the strong color difference depending on the beam angle. After dedoming, the color difference is much less intense and the light color blends more. This corresponds in the broadest sense to the old XM-L2.

Light quality and use in optics

Despite the strong color differences in the radiation pattern, the light pattern is usable in a standard reflector. There are minimal yellow rings around the same colored spot, but these should be easily removed with a slightly modified reflector design or an OP reflector. So for use in reflectors, this LED is fine.

The photo above shows the light pattern in dedomed state, but except for a slightly narrower spot, nothing changes fundamentally compared to the LED with dome. It should be noted, however, that the light color in the spot is nevertheless colder than the spill, which is more pronounced in the LED with dome.

The light color is a bluish white. The specification of typ. 70 CRI seems realistic here, the color red is only reproduced strongly distorted.


A very interesting LED. Even though it is still marketed under the name "XM-L2" and even uses the same order codes as the previous version, it is basically a completely different model with entirely different properties and design than its predecessor. The efficiency is very high, the internal structure of the LED is completely different (flip chip instead of lateral chip) and also the optical properties are different from the previous XM-L2. The luminance is much higher, and the thermal resistance has been reduced so much that almost three times the power of its predecessor is possible.

Especially at the beginning of the introduction of this new version, the immediate use of the former order codes was a huge problem. Due to the mostly unknown date codes of the reels, it was very difficult or even impossible for the customer to find out whether an LED of the latest design or still the predecessor was delivered. In addition, the current data sheet on Cree's site is only valid for the variant tested here. I have not found a version history.

For me, this XM-L2 in the new design is more or less what the XM-L3 should have been. Significantly improved electrical characteristics, higher luminance and full compatibility with existing designs and accessories make this LED a clear recommendation if a clean light image is not essential, but a replacement of an XM-L or XM-L2 is necessary.


  • Impressive performance
  • low Vf and very low thermal resistance
  • fully compatible with XM accessories


  • luminance in the range of older emitters like XP-L(2) or SST-40-W
  • dedoming is hardly worthwhile
  • light quality in reflectors needs some improvement


  • Due to the continued use of the previous order codes unclear whether LEDs of this type are already shipped as XM-L2
  • Current data sheet is only valid for the new version of the XM-L2

Thank you for reading the test. Hope this review will be helpful for some of your modding projects :)

Greetings, Dominik

3 Thanks

< reserved >


Thanks for this. It looks impressive.

Step aside SFT40

The STF40 is a completly different LED, without dome and with higher luminance at lower currents. I don't tested this emitter yet (is on my to-do list) but this seems obvious since the domeless variants are quite superior in case of high luminance at relatively low currents.

This new XM-L2 delivers high luminance only at extremely high currents (10 Amps and above), which is simply not usable in reallife flashlight use.

1 Thank

Lots of people already enjoy the 30-60 seconds of turbo with their SBT90.2 lights and multi emitter lights running 20+ amps.

koef3, thanks so much for this! This is the BLF! :slight_smile:

Very interesting emitter. I wonder if they had any target markets for it. Thanks for going to the trouble of dedoming, too…that was one of the things that first popped into my head as I was beginning to read.

Curious how it looks in plastic diffusion optics and TIRs.

2 Thanks

Shame about that Cree Rainbow™

I’d be interested in high-CRI versions.

  1. Thank you for cranking out these tests lately--they are tremendous resources and answer burning questions for the community.
  2. As-is, the dedomed version could not beat the SFT40 in luminance.
  3. Would it be possible to also "dice" the LED post-dedome, the same way people would an XHP50.2? Removing some material surrounding the die should help reduce light leakage (the glow around the die as you pointed out) and hopefully increase luminance. Note that all the top-performing thrower LEDs (Osrams and SFT40) have completely opaque material surrounding the die, which reduces leakage.
  4. I think the roughness of the phosphor (very grainy) could be a reason for the less-than-expected intensity gain post-dedome.
  5. This provides a neutral/warm alternative to the SFT40, which is a huge development in my book!

Thanks for this test and other information koef3. This is a surprising result.

thanks for the info!

Thanks for the tests @koef3 :+1: . Nice to know CREE is updating old emitters.

Glad to see some more LED tests coming out

If you happen to have the time, would like to see a break down of XM-L Gen 2 Color HI

I just looked into the datasheet.

As I thought this LED needs a special board, because of the four independent LED chips. I don't have this type of board with DTP thermal heat dissipation, I don't even have any type of board for this type of LED :D

So I have currently no plans to test this special LED, sorry.

Thanks :TH :+1: UMBS-UP:

Update 12/06/2022: Updated luminance chart due to insufficient measurement!

Due to inadequate measuring devices with low accuracy and measuring methods the values for the XM-L2 new design were way too high, in the table above are the values taken with the much better equipment and more accuracy in taking the measurements. Please taking this to consideration if you mod some lights with this LED! In the future the luminance measurements will be done with this new setup with seems to be much more accurate.

Now the luminance is not as high as before, more in the range of XP-L HD or XP-L2, which seems realistic considering the LED design.

The high luminance values I got in the first measurement was also the final reason to revamp the whole setup and equipment. But detecting and eliminating possible measurement errors is much more important than measuring anything at all and hoping that the values are correct. One is always evolving... :D

The lumens measurements seem to be at least 14% higher than they should be, and the luminance data is inconsistent.

I do appreciate your effort though.

If you're referring to Cree's specs - I'm relatively sure that the luminous flux of the XM-L2 new design is actually higher than specified. My last Cree LED of another type I tested was a little less efficient than I expected. Occasionally Cree emitters are binned higher than noted in the order code. I have found emitters with 83 CRI (measured in a professional lab) with XP-E2 with typ. 75 CRI, and an XP-G2 that was even more than 30% (!) worse than expected. That just happens sometimes...

The chart shows the older XM-L2 U2 putting out 2050 lumens at 5A, for example. That seems high, but I haven’t tested that LED myself.

I’ve been meaning to do some LED tests. I’ll order some of the new XM-L2 to test with a Maukka calibrated lumen tube and see what I get.

Then it is important to get the same order code from the same supplier (and in best case from the same batch/reel).

The "XM-L2 U2" was in fact a XM-L2 U4 (the manufacturer stated U2 bin but it was much higher maybe a silent change of bin which seems unlikely but still possible), and it is important to know that I test at 25 deg C, not 85 deg C.

In the foreseeable future nevertheless I will upgrade to a integrating sphere with proper geometry and also (in best case) with light sources calibrated by federal agencies / universities or something like that.