I compared the sleep current of the newly released nRF54L15 with the nRF52840 and MG24, known for their low power consumption. I hope this proves useful.
Measurements were taken using ppk2 while applying 3.8V to the battery pads on the device’s back. Since sleep current is highly sensitive to temperature, room temperature was maintained as close to 25°C as possible.
For the BSP, I used “Seeed nRF52 Boards 1.1.10” for the nRF52840 and “Silicon Labs 3.0.0” for the MG24, working within the capabilities of the Arduino IDE. The nRF54L15 was worked on in a VsCode + PlatformIO environment. No tricky operations were performed to lower the current values.
Comparison results:
- Deep sleep current is sufficiently low at around 2uA, showing no significant difference.
- Light sleep(idle) current is significantly worse for MG24 at 60uA.
- As an example of operating current, I compared the current during execution of
while(mills()-timestamp);. The nRF54L15 shows significantly lower current consumption.
Overall, the nRF54L15 appears to be the most power-efficient device.
Results may vary depending on device-specific tuning. Please share any additional information you have.
The code used for the experiment is attached below.
XIAO_SleepCurrent.zip (6.5 KB)
4 Likes
Hi there,
So I feel like some PLUG is needed here, and so much more to be discovered. If seeed would take full advantage more could be gained.
Also Some of the users having issues may not be aware of the Power domains and turning things on when needed, I’m just starting to Scratch the silicon surface 
Thank you for the efforts and enlightenment 
PLUG
The nRF54L Series is among the first publicly available low-power wireless SoCs fabricated using 22 nm process technology. This technology offers high transistor density, lower average power consumption, and improved process efficiency, enabling the creation of compact, efficient, and feature-rich chips for advanced applications.
To optimize energy consumption, the nRF54L Series architecture is divided into four power domains that are independently powered, depending on what features are active at any point in time. This architecture enables ultra-low-power operation by allowing domains to be powered independently, allowing large sections of the chip to be turned off when not actively needed.
HTH
GL 
PJ 
When discussing XIAO’s sleep current, the current consumption of onboard devices other than the MPU must be considered. The most dominant factor is the SBD leakage current, which exhibits significant variation and is temperature-sensitive.
[typ. uA] nRF54L15 nRF52840 MG24
MPU I_sleep 0.6 0.4 0.25
SBD Ir 5 Unknown 5
3.3V Reg Iq 0.3 1 0.3
LevelShifter IccA 2 - -
Measured Value Ioff 1.7 2.4 1.6
1 Like
Interesting discovery 
So, with BLE, Zigbee, Matter support out of the box, a stable development and complete environment (including Micropython), the EFR32MG24 is looking to be quite a “favourable” option? (Yes, in Australia and other Commonwealth countries, there is a “u” in “favourable” 
)
The MG24 has excellent sleep current values, comparable to the nRF, but its operating current consumption when the radio is active seems less favorable. I suspect the transmit stage amplifier design is fundamentally different.
‘Comparison of Advertising Current for XIAO nRF54L15, nRF52840, and MG24’
1 Like
Hi there,
So , “Better” depends on the job. MG24 absolutely has a very nice ‘out-of-the-box’ story for Zigbee/Thread/Matter (and even CircuitPython).
But on raw efficiency, nRF54L15 wins the common cases: ~3.3–3.4 mA RX and ~4.8–5.2 mA @0 dBm TX, with ~0.8 µA System-OFF + RTC. MG24’s own datasheet lists ~4.4 mA RX and ~5.0 mA @0 dBm TX. Also, 54L15 does include 802.15.4, so Thread/Matter aren’t off the table. Pick the part for your constraints—ecosystem speed vs. lifetime on a coin cell.
Compute & headroom:
nRF54L15 runs a 128 MHz M33 (+ RISC-V coprocessor) with up to 1.5 MB NVM / 256 KB RAM. MG24 tops out at ~78 MHz M33 with up to 1.5 MB / 256 KB. If you need more CPU cycles per joule, 54L15 has the architectural advantage.
MG24 has SKUs with +20 dBm—fantastic for range, but you’ll pay for it (~156 mA at max power). If you don’t need that PA, the comparison should be apples-to-apples at 0 dBm.
So is MG24 “more favourable”?
If your priority is quickest path to Zigbee/Thread/Matter and Python-based prototyping:* sure, MG24 is easier out of the gate for now. 
If your priority is battery life for BLE-centric sensors/trackers, plus more CPU per µA: the nRF54L15 is hard to beat right now. 
I can say though at 4-5 weeks the info is avalanching with updates coming in a timely manner. Giving it time to marinate and propagate you can’t beat the support they offer over the other companies.
GL PJ
more to come, forsure from the NORDs…
1 Like
Thanks for the correction.
I have a few weeks “spare” before my customer’s prototype boards arrive so hopefully “the Nords, Vikings etc” have a suitable solution to these “teething” problems?
1 Like
