I use a Xylem D5 pump with a standalone reservoir, with the pump being powered through a direct SATA connection from a PSU used only for watercooling components and not part of the test system. The pump is controlled by an Aquacomputer Aquaero 6 XT in PWM mode. There is a calibrated in-line flow meter and Dwyer 490 Series 1 wet-wet manometer to measure the pressure drop of the component being tested. Every component is connected to the manometer by the way of soft tubing, compression fittings, and two T-fittings that have been accounted for when it comes to the liquid flow restriction in the loop.
There are a total of six blocks in this roundup, including this universal GPU block as opposed to the other five full-cover blocks. The iCUE Link XG3 clearly has more microfins over a larger area compared to the Hydro XG7 from the same brand so it wasn't surprising to see this being slightly more restrictive. At the same time, the cooling engine is relatively simpler than the other full-cover blocks since the coolant only really goes over the fins and not the VRAM/VRM sections, so it still manages to be in the middle of the block and not the most restrictive unit. Overall I'd say this is still less restrictive than a typical CPU block and should be fine with a decent pump.
Fan RPM
This is also the only block in this roundup with a fan integrated and thus I decided to chart the RPM response curve, or rather straight line, as a function of PWM duty cycle of 20 to 100%—you can't go below 20% in iCUE. The iCUE Link connectivity allows for software control over the fan speed as seen on the previous page. The default "silent" preset keeps the fan hovering around 1500 RPM or slightly higher depending on your loop, but you can of course set a custom fan curve if you want the VRMs to be even cooler at the expense of fan noise.