(*testing components prior to installation)
intro:
this post condenses weeks of intense learning and pain into a digestible story.
i really appreciate the efforts and support from the NFC discord for helping to make this a reality.
please enjoy!
build parts:
Case: Skyreach 4 TINY
Mobo: ROG Strix X570-I Gaming
RAM: G.SKILL Trident Z RGB Series DDR4-3200 2x8GB
Storage: Samsung 970 EVO M.2
CPU: Ryzen 5700g
CPU Cooler: Noctua NH-L9i chromax.black
GPU: Gigabyte 4060 Low Profile
PSU: HDPlex 250W GaN
(+Optional: J-HACK ONE2 PLUGin 200X)
parts layout:
i was using this case with 5700g dedicated graphics for years as a media PC, but then realized super small GPUs may fit. pulled the trigger on the 4060lp after very little research. after playing around for hours, i ended up getting it to fit with the above configuration, thinking a power supply might fit underneath. while moving things around, i kept nicking the top lip of the frame above the GPU and realized no matter what i did, that part would need to be cut, so i cut it off. after sharing these happenings in the NFC discord, Josh actually put the parts together into CAD and called it "the tightest possible fit". i have to again, give a lot of credit to users for egging me on, and helping me realize that a 250W GaN would actually fit undernearth (played around with the idea and a 200W Meanwell, + DC-DC solution, but was eventually talked down off this cliff as 200W is truly not enough for a 65W CPU+115W GPU when marketed TDP values is a lie. the 5700g wattage target is actually 88w, im sure the 4060 has a higher target aswell)
directly above the GPU PCB, you can see the 'lip' of the frame that was cut. seeing this image made me realize, of course that this was crazy, as the CAD measurement was giving a value of -0.05mm clearance for everything, even with the 1.6mm savings from removing the GaN cover and GPU fans touching the GaN on top. so this was not going to work as is. i understood that more space would need to created. i thought about it for the night, slept, woke up, and realized i could re-use the same trick i implimented before -- cutting the frame. instead of cutting the lip space above the GPU as i had done prior, i knew i could cut space underneath the GaN, on the bottom of the frame, and allow it to slip right through (planning to mount the 250W GaN to the outer cover of the S4T). this would give me the additional 2mm i needed to make this all work. so, with the help of a mechanic friend, got to cutting.
(*looking back on this, i think removing the upper lip may be enough, raising the GPU that 2+ mm and keep the cover on the GaN. in that case, you would need perfectly 0mm cables coming from the side of the GaN)
cutting the frame:
leftmost picture is the cut lip above the GPU. second picture is a pencil outline of the area needing to be cut for the GaN to fit through the bottom of the frame. picture three is the results from cutting, with a grinder and dremel, and in the final picture on the far right you can see the difference between the cut frame vs. a stock S4T frame. these cuts gave me an additional 4mm in total! which was absolutely needed for my plan of running power. there are other ways to save space, which involve deshrouding the GPU and using different fans -- i wanted to aviod that as the stock fans are the most quiet, from the testing that has been done so far by Josh.
possible riser cable runs and GPU orientations:
i came up with a bending method for the riser that would allow for a very short riser cable on this build. it involves bending each of the flat cables 45 degrees (which turns the cables 90 degrees upon bending) and routing them all underneath one another. subsequently after creating this bending method, i learned about M.2 to PCIE risers, and i would have absolutely used one of those instead if i knew about them. however, here we are, i found this above bending method and it works, and is what i ended up going with for this build.
i think it's important to address this early for people thinking "hey, your GPU airflow will suck". well, it's possible to have the GPU fans face the case rather than the GaN for better airflow. this is an example of a riser run that would work, with the riser going underneath the MOBO. however you would need to create individual low profile cables for the MOBO power. there may be different MOBOs with the 24-pin power and RAM DIMM slots in another location that would make this easier. Here, a left angled riser would give best clearance and im only using the straight riser coming out of the GPU for demonstration. PCIE 4.0 cable would need to be 425mm+ to reach underneath the MOBO. you could also use a PCIE 3.0 to M.2 riser, with absolutely no performance loss. i choose NOT to do this, as i had a lot of time into the build already and thermals are "ok enough" for what i use the build for. don't feel like creating 24 more individual low profile cables for MOBO power. i did want to address this concern early as i'm sure a lot experienced builders are thinking that this will run hot.
TL;DR: if you want to have the GPU fans facing the case -- its possible with a left angled pcie riser (think a 90degree riser to m.2 would be easiest).
custom low profile molex cables (+DisplayPort cable runs)
above shows the first picture with placement of the GaN. in this picture, firstly, is a standard 8pin motherboard cable simply used for testing. that will go away soon. all of our cables will need to be low profile in order for the GPU to have space. in this above picture, custom individual molex power wires/cables have been ran from 12v and ground pins from the GaN to the J-Hack 200W unit. The J-Hack converts that 12v and ground into all the power that the motherboard 24-pin connector needs. without a J-Hack, or similar DC-DC unit for the motherboard, you would need to run 24 individual low profile cables/wires from the GaN to the 24-pin motherboard header. i learned about this bending method for crimping and bending molex pins from the NFC discord, huge shoutout to identify theif for walking me through this and user we for the suggestion on molex crimper. super happy with how these turned out. also mandatory to mention that the Power Button needs to be plugged in at this point with this exact particular hardware -- just due to the placement of the header pins and how the J-Hack lines up.
here, the individual low profile cables have been ran to the 8-pin on the motherboard. also worth pointing out, low profile connections were also required on the 8pin motherboard header, as my displayport cables from the GPU will run out of the case above this 8-pin on the motherboard, and using a standard connector here takes away the space needed.
and finally, here is the low profile cables plugged into the HDPlex for the GPU power. also worth noting is the very messed up looking chinese low profile DisplayPort cables. you can see the rough placement here for where they will eventually run out of the GPU and out of the MOBO backplate/cover. a hole is needed to be cut out of the MOBO cover for these DisplayPort passthrough cables.
it's definitely worth noting that i ended up putting kaptop tape over these individual wires on the GaN. there's one piece of wide kapton tape covering this whole area. this serves two purposes, keeping the wires as low as possible, while making sure nothing shorts or gets caught in the GPU fans. there is also a piece of foam crammed into the corner of the case -- this keeps the GPU upright and away from these low profile wires. kapton tape + foam = win.
plugging in the GPU, system posting
plugging in the GPU, putting it into it's place, and running a temporary c14 power connector for testing. wow, it boots! easier to see the DisplayPort run from here, snaking around the CPU cooler, and eventually out of the case. of course, eventually it's time to put the cover on and try again. ok cool, it works! of course, there is kapton tape on the bottom of the cover to make sure not to short out the GPU and riser cable area. actually, looking at the picture on the right that long white strand is from the roll of kapton tape. dang i'm messy. also, it's at this time i was pretty stressed as i had never soldered anything in my life before. but now was forced to solder the gx12 connector to make further progress with the build.
i didnt snap a photo while building so this is the best picture i can provide atm demonstrating the kapton tape and foam mentioned. in the bottom corner, between the GaN and GPU, you can see the kapton tape coming over the side of the GaN. and the foam piece in the corner holding the GPU up and keeping the fans spinning. this picture also helps illustrate that there is NO room for a power cable, or power button on this side of the case. unless it was some super low profile custom thing -- i think i'll pass on even thinking about that. long story short, both the AC power and power button have to be on the back side of the case with the rest of the IO.
gx12 pin for AC power and DisplayPort cable to IO shield
as there was clearly no space on the front of the case, i knew i would have to utilize one of the two back holes on the stock S4T for incoming power. since i had already purchased a 16mm power button from NFC (god, they are so beautiful), my only option was to use the 12mm hole for power (well ok, thinking back i could have drilled it larger, or perhaps cut a square for a g14. honestly if i had to do this again i would probably go the 16mm hole route). as far as i can tell, it's impossible to purchase a pre-soldered 12mm power cable anywhere. after asking around for a 12mm connector, these gx12 connecters were recommended by users on the NFC discord. i've never soldered anything before -- and to my amazment after recovering my SFF forum account to make this post -- my only 1 post from 2018 is where i mention never soldering before. and this was my first time doing so. it was honestly nervewracking. i used a voltmeter to make sure everything was safe throughout the processes and was able to finish in an hour or so. if i had read my post from 2018 i would have probably considered cutting. jeez, anyways.
in retrospect -- i would use a 12mm power button and a 16mm gx16 for incoming power (they are much larger and easier to solder, i imagine). would also consider purchasing custom gx12 or gx16 from NFC- i now realize this is an option from their store.. and custom requests are also avalible.
this is probably my favorite picture from the whole thing, i think it's an art piece demonstrating creation, really. anyways, plugging the power connector in and the DisplayPort through the new hole made in the back of the motherboard IO shield. and she powers on and posts! i ended up de-pinning the HDPlex 3-pin power connector, and crimping on my own Male Molex pins on the other ends of the gx12 recepticable wires connected to the inside of the case. the male molex pins slip right into the stock 3-pin HDPlex connector, and i was able to plug it right into the HDPlex 3-pin AC input.
put more heatshrink over the gx12 housing. when i ordered the custom DisplayPort cables, i asked the chinese supplier to put heatshrink where they could. i didn't want to bother with it and wanted to avoid shorts. anyways, somehow this translated to not sending any mounting screw holes as shown in the pictures. so, i was left to my own devices finding a way to mount these. i had definitely intended to use screws as shown in the pictures when you buy these off amazon/ebay or whatever. anyways, i ended up cutting a hole for both DisplayPorts to fit through, put 3M sticky tape and foam between them to they give each other strength, 3M tape on the MOBO IO, as well as zip ties on the outside of the case to hold them on the out of the case when you apply pressure. honestly turned out looking OK. of course, the screw holds would work a lot better but i wanted to finish! and again just plain impatience. but honestly i think it turned out ok.
final pictures and thoughts
i ended up throwing on a front bezel from my old S4T expansion kit to give the GPU some fresh air. eventually, i may look to order a bezel with skyslots for a more finished look, there are a ton of custom materials and designs that could work here. overall i'm aboslutely in love!
i have a lot of testing to do as far as thermals and potential undervolting potential. GPU sits around 60 degrees with two 4k 60hz monitors during normal work loads. in general i've ran GPU up to sustained 50% load and 60 degrees solid. running furmark+prime95 simultaneously will cause throttling after 10-15 minutes. i think that flipping GPU fans towards the case would likely help with this -- however, that's not something on my radar probably ever or at least for a very long time. have a few other builds i'm trying to work through at the moment.
i spent way longer than i anticipated putting this thread together, so i really don't have my final thoughts all together. i really appreicate this community, and NFC Josh in particular for being a torch in showing us all to never compromise.
