SiR4242DP works well, as PPtk suggested it and probably works well for him in his applications.
Had a bad layout before, removed old links to not confuse someone.
Was shoring source and gate. Vishay shows bottom view of the part, not top view…
For surface mounted parts, I don’t know how that is useful because it is not.
Wrong layout on the PCB, but it shows you how it is done.
Had to turn the part around.
Here are two options of connecting it, with the first one you need to connect the gate by a trace that you need to add to connect the gate pad to a signal trace.
The second one allows to put the mosfet more inside the PCB’s center and does not need any new traces. Problem is, it may short signal trace and output trace if it manages to solder on the trace cut edge. It did not for me, although I used extra new traces for source and drain because I had cut there before and wanted to make it more solidly routed. I doubt it would solder there especially when not putting any paste near it but one never knows and has to check before plugging the driver anywhere.
And the result:
Works with an XM-L U3 fine but it pushes 3A soon so a resistor was used for testing the higher currents, 0.47ohm, 5W only so I have to watch not to burn it.
The XM-L U3 was hitting 3A at 3.3V or 3.4V hence I had to switch to the resistor.
Powered from my PSU although the PSU screams a little with the load, dunno why, probably because it’s lines are unbalanced and the load is PWM.
Can’t measure what it gives as voltage under load so it’s an idle voltage, 5V is loaded and should be stable, 3.3V line shows 3.4V but I guess it drops to 3.3V once it’s loaded.
nlite firmware, 18kHz PWM, 5–33–100% modes.
Load
0.47ohm (5W ceramic under a fan, poor thing even smelled once)
Test #1
U: 3.3V
L: 0.3A
M: 2.1A
H: 5.8A
Calculated total resistance 3.3/5.8 = 0.57ohm, means 0.1ohm on all connections and the mosfet combined if resistor really has 0.47ohm.
Test #2
U: 5V
L: 0.4A
M: 2.7A
H: 7.9A
Calculated total resistance 5/7.9 = 0.63ohm, means 0.16ohm on all connections and the mosfet combined.
Soldered connections apart from the PSU connection and I hold two wires on/off as a switch.
Can measure up to 20A, indirect, ACS712, 0.1A resolution on my meter.
Cables on the driver are maybe 4”, probably 0.5mm^2.
It should be doing more, but with all the resistances that add up on the route it does the above.
PSU should give 3.3V/28A and 5V/30A, hopefully it does under load.
The driver is dead cold, mosfet especially.
Even pushing it at the 8A I did not notice the mosfet getting any warm. Only a smell and a hot resistor J)
At 5V, 6mohm, it really doesn’t give a damn about 0.384W being dissipated on it. It’s rated 3.1-4.8W depending on temperature. At lower voltages it goes to 10mohm at 3.3V and 20mohm at 3.0V (datasheet). Even if it would give 15A at 3.0V with 20mohm, it dissipates 4.5W. At 3.3V, 10mohm, 20A with 4W on the mosfet. 4.5V, 7mohm, 25A, 4.375W. It’s rated around 20A.
I think the traces or flashlight would caught on fire sooner than the mosfet would die.
IMPORTANT:
When I drive it on 5V PSU and have it in low mode, it switches after a short while automatically into medium mode. I think this is because the capacitor on the ATtiny13A does not filter the voltage enough. It’s the only mode and combination when it does that, consistently. Maybe it interferes with the PSU, dunno.
I also wasn’t able to reprogram the ATtiny when the capacitor was missing, dunno why it needs it, but without it it won’t reprogram… it acts as if it does not see the ATtiny.
With my DIY clip it can be reprogrammed even in this position when the mosfet is closer to the MCU. And should be fine with other clips too unless you have some crazy beefy clip, then just file the plastic off the clip 
I will try moon mode now and try running it from an 26650, protected, may even try the unprotected Samsung 18650 beast, just don’t want to accidentally short it.
Might add 4-5 AMC7135s to make a medium mode without PWM and avoid the crazy spikes on low mode. That’s always an option.
