Closer Examination

There is a protective clear film on top of the polished cast acrylic top to help prevent any scratches or smudges and removing it gives us a good look at the Phanteks Glacier G4090AS block for the ASUS RTX 4090 Strix/TUF GPUs. The white color version is not the easiest to photograph against a white background but I am sure the chrome finish would have been worse given how reflective it would no doubt be. Then there's the black which is more understated. All three share a lot in common and really only differ in the way the two aluminium covers on the top are finished, with the black getting an anodized treatment, the white being powdercoated, and the chrome being electroplated on the base aluminium metal. The acrylic top otherwise extends a lot longer than usual and this block is actually the longest of any in this round-up by a fair margin.

Phanteks instead saves on the height (width) of the block compared to most others by integrating the I/O ports on the side of the top rather than separately above with a terminal as usual. There's multiple Phanteks logos on the block as well as a G40 etched on the top left corner—this block was previously called G40. There are four BSP G1/4" threaded ports on the right, of which you use two to include the block in your loop and then plug the other two with the provided stop plugs—no preferred inlet and outlet markings here although logically I would choose one of the inner-most set of ports for the inlet to feed the coolant directly over the GPU core. At the bottom is a small recess in the aluminium cover through which we see a tiny connector coming off the RGB LED PCB.


Turning the block around, we see the larger aluminium cover also helps keep the coolant in place with O-rings around the flow pathway. In fact there is a thicker acrylic section here which is effectively the equivalent to the I/O terminal, although closer in functionality to a distro plate given it is more a manifold here. The cold plate itself is much smaller in size compared to the top but still makes contact with all the relevant sections of the PCB to where this is a full-length, full-cover block and then some. There is no copper-only version, as with the rest of the blocks in this round-up, and we end up with nickel-plated copper given a nice polish. The various raised sections are what make contact with the GPU core/VRMs/VRAM modules on the PCB and where the thermal pads and paste will help with heat transfer. Given the clear top and the relative urgency to finish testing for this round-up, I figured there wasn't much to gain from disassembly to see the cooling engine better since we already get a good look at it.

The coolant flows from the right to the left, as seen from the front, and then enters diagonally over the GPU core where it splits into two via a jet plate. I recall seeing a few older versions of this design where the coolant split diagonally too, which didn't make much sense to me since you would have uneven heat transfer and some sections of the microfins would not have as much contact with the coolant as others. Here at least this is purely for aesthetics before the coolant indeed splits down the middle, goes over the VRAM/VRMs, combines into a single stream again on the right before exiting out the block. Interestingly, Phanteks appeared to have tried to make the most of the previous flow pathway by extending the fins at the top and bottom along the diagonal entry path which still remains in place here. There are far fewer fins (27) and channels here and they appear quick thick too, so this block does not have the most complex cooling engine with the highest active surface area for heat transfer either. At the same time, the coolant flow pathway introduces several bends and twists, including right over the fins, so I am certainly curious how these factors would affect thermal performance and coolant flow restriction.