EKWB De-Lids an AMD Processor for Better Cooling
EKWB de-lids the AMD APUs.
We've long heard of folks de-lidding Intel Ivy Bridge and Haswell processors, but we have not heard of de-lidding AMD processors; that is, up until now.
EKWB claims that it is the first to support cooling of de-lidded AMD A-series CPUs. The de-lidding is only supported with the Llano, Richland, Trinity, and Kaveri APUs. Any socket AM3+ CPUs do not support de-lidding, plainly because the CPU dies are soldered to the IHS.
The product from EKWB that makes this possible is the EK-Supremacy PreciseMount Add-on Naken APU. This is a rather long name for what is essentially a different screw set that allows EKWB's water blocks to be mounted lower, actually reaching the CPU die while providing sufficient pressure.
Note that if you de-lid your processor, your warranty is void; if you fail in the process, only you can be held accountable for any damage done.
EKWB has listed the screw set on its webshop with a price tag of $5.48.

I don't think anyone would go to the expense to custom liquid cool an apu.
But i'm sure someone has done it so feel free to prove me wrong.
In my experience 100w TDP and lower are also not soldered and can be delidded, but I do believe that several 95w TDP Phenoms and FX CPUs are soldered.
And lastly, AMD has used the exact same mounting method since AM2, so anything that supports de-lidded FM2 will work on delidded AM2/AM3 as well.
For reference, I use a Cooler Master X6 Elite on a delidded G2 Brisbane 4800+.
Some de-lidding techniques seem awfully simple and relatively fool-proof if they really work as well as shown. Not sure how helpful it is on AMD's deliddables but on Intel chips, the main benefit is eliminating the 100+ microns thick thermal paste gap between the CPU and IHS caused by the IHS glue acting as a shim, preventing the IHS from making physical contact with the die. Removing the glue and putting the IHS back on reduces Intel chips' core temperatures by 8-15C. Putting paper shims to replicate the stock gap and using "better" thermal paste between the IHS and CPU produces worse results, indicating that Intel's "crap paste" actually outperforms the vast majority of "high-end" aftermarket pastes.
The uncovered chips received broken corners quite often because of heatsinks that were installed unevenly. The heatspreader was more a way of protection for the silicon of the chip rather than more evenly spreading the heat. A decent heatsink could spread the heat as well if not better.
Many people were accidentally crushing their CPU dies and another problem is that without a minimum amount of thermal mass attached to the CPU die, the CPU could overheat and destroy itself faster than thermal management could shut down clocks.
The IHS eliminates or significantly reduces two of the most common failure modes in retail CPUs.
kind of if you are on a budget, the gpus on board are better for decent gaming compared to intels offerings if you dont have a dedicated gpu.
"Look we replicated the gap for lulz, and this somehow proves Compound X isn't as good as the factory stuff even though we're not using it properly. Teh Lulzzzzzzz"
This might be useful for 750K/760K chips for a real budget overclock or for the experience. Other than that, I don't see a lot of use cases for delidding FM chips.
If you could apply Intel's IHS paste without the gap, it would likely perform even better than "aftermarket pastes applied properly" except applying Intel's paste properly probably requires a hydraulic press to pack the near-solid paste properly on the IHS before slapping the IHS on the CPU.
How do you ensure that vapor will uniformly hit every hot spot across the die area with sufficient flow to prevent local overheats? You cannot. So you would still need to attach something to spread heat around to prevent local hot spots and increase the heat transfer area from the die to vapor.