Thermalright True Spirit 120 Rev. A BW 4U fan mount rebuild (#P29)
Technically the second step to the Fujitsu Esprimo board modification in #P28, but not necessary for regular ATX cases that are just big enough to fit a tower cooler, especially nowadays. Well, I always got special problems with my special builds, so I needed to address that.
With the Intel board mount, the Thermalright True Spirit 120 sits ever so slightly higher up in the case compared to the AMD one. It not only interferes with the card holder which is fixed by drilling two more holes in it, but the appended fan holder also touches the case lid. That one will not be drilled into! Luckily the fan holder is a plastic part that can be completely separated from the heat sink and the fan. It is constructed to have no fan guard/protector to lower noise, which makes it impossible to modify to my needs, it would simply fall apart. 3D printers to the rescue…
The thingiverse STL visualizer is once again extremely slow or broken, so I’ll just add the rendered image later. Basically I recreated the entire holder in three separate pieces, the fan base and two identical fingers that mount via M3 screws left and right. This design is only necessary as the original piece has isotropic material properties due to the injection moulding process, but the FDM print yields different properties along and perpendicular to the print layers. Both the fingers and the main part flex quite a bit during mounting, so there’s no optimal printing orientation for one big part. I made it into separate pieces that do have an optimal print orientation each.
Here’s the assembled unit with the stock fan:
The fingers are fixed in multiple axis (two screws, three 90° edges from the main piece) so they won’t move. The main piece is screwed to the fan after mounting the finger units, which is not part of the original design. That one has little grips on the part that is no longer present here, and springy bits on the fan mounting holes to avoid vibration noise.
The fingers are also slightly longer than the original ones as they wouldn’t grip right. Well, the original unit didn’t as well, so that’s an improvement. There is a tiny recess on the heat sink fins that would accommodate the fan, but this never worked for me. So this holder entirely clips to the heat sink by the finger grips, not using the fan alignment on the fins.
Two more comparison shots – the fingers grip tightly enough to actually reduce their number by about two, but since it worked after printing the second base and third pair of fingers, that’s what I’ll stick with.
Not sure why the original part is not symmetric, as the heat sink clearly is.
Now about mounting the entire thing: The motherboard base mount is elevated by a single M3 stainless steel washer on each post, as the stock nylon washers cannot be mounted on the Fujitsu board. It has a backplate that also holds the CPU mount in place, so it cannot be exchanged for the Thermalright parts. The washers just compensate the minor height difference so the heat sink mount sits flush. The washers can also be put on top of that mounting plate which would only require two of them, doesn’t really matter.
The heat sink design is symmetrical but the base mount isn’t, so it is oriented in such a way that there is room for the fan between the CPU socket and the PCIe slots. That way, the fan can be mounted with airflow direction towards the heat sink and towards the power supply fan, which should result in lower temperatures of the graphics card PCB.
The red fan connector barely clears the base plate, so that’s dope. Note also the USB3 front connector is in close proximity, and it also clears the fan assembly just barely. So there is a strict order of assembly/disassembly, but it works without further hardware modification.
The fingers could have been made shorter to fit the leftmost ridge in the heat sink, but this way there is less strain on the material. Also the stock holder does it the same way (well, it doesn’t really reach over so it sits more in the middle, but as I said it never had perfect fit)
The recess of the heat sink fins can be seen here. The bare fan does fit in here, but neither the original piece nor my version does use this additional alignment. Again, air flow is directed through the heat sink and the out of the case via the power supply fan instead of heating the GPU card from above.
And this is the thing completely mounted to the chassis with the card holder screwed in place, barely passing the ends of two heat pipes. Yes, the wiring looks terrible, no I won’t bother cleaning this up much further, I won’t even take out additional SATA and power cables that are not in use at the moment. This is not a modern “gaming” (show-)case with double bottom architecture to hide any wiring, this is a 4U = 18 cm server case that doesn’t have room for that kind of games. Airflow isn’t affected all that much, and I won’t be seeing this mess in the rack at all, hence the original motivation to make it fit the lid once again. This will do for me, YMMV.
Idle temperatures are down from like 60°C with the tiny stock cooler to 40°C. They won’t go over ~65°C at 100% Prime load, that’s what a large surface heat sink and a quiet 120mm fan can do compared to a beefed up laptop cooler. Benchmarks can be seen in the previous P28 post.
In terms of cost, printing this was the only expense (apart from time and a bit of electricity) that I had. The old board, CPU and the surplus OEM power supply sold for 47€ on eBay (three auctions, plus postage), subtract the greedy 10% that eBay takes, and that’s exactly the 42€ that I paid for the new board/CPU/PS and 4GB of RAM. So the swap from an FX-6300 to the more modern i5-4570 was basically free and also generated two blog posts. Excellent yield
Here’s the 3D model files on thingiverse, the RSDOC (AiO file), and the three exported files in STL format: main, side part 1, side part 2
Shoutout to WordPress that once again fucked up STL and RSDOC uploads with the v5.5 update. You guys really suck in making updates that do not brick existing installations