You may have (possibly!) noticed I have been rather quiet the last few days.
That is simple enough to explain – as I mentioned in a post last week, this last weekend was the time in which I would finally bring my current Ryzen 9 3900x CPU and the purchased and in-use RTX 3080 I had for my new build into my actual new build, plumbed up with watercooling and assembled on a new desk and monitor array.
The experience was…interesting, in a few ways. It is the most I’ve ever preplanned a PC build or decorative endeavor, kitted out with 3D models approximating the space, reading weight limits on desk frames and monitor arms, figuring out the right ways in which to build out my space to be more fun and also more usable.
When I set out after my Thursday raid night, my goal was to split the work into logical segments. I also had a lot of hesitation and doubt about building my first watercooled PC, but having already purchased the hardware, I ultimately had to work up the courage to pull the two needed parts from my existing rig and get to work, and I have.
Like with any such project, there were some challenges and complications, but I am sitting here today writing this post at my new desk on my new(-ish) PC, so for whatever challenges I did encounter, I was able to overcome them!
Let’s break it down day-by-day for fun and to share some of my learnings!
Friday was a simple day – I was planning to leak-test the watercooling hardware with the tubes I had cut myself to ensure no leakage or hardware malfunctions. Easy enough, but the doubt struck early.
To save space initially, I had unboxed the watercooling hardware and preinstalled most of it into the new system’s case. I tested my run lengths, made sure nothing was awry, and cut tubing to ensure smooth flow and no kinking in the tube runs. I got to a good spot and left it sitting, half-assembled, in this state for the last two months.
When Friday came, I had initially thought to leak-test in the case – just prop up the GPU waterblock without the graphics card attached and run some distilled water through the whole thing. Easy, right? Well…sure, but it worried me a lot. So…I ended up yanking the whole setup out of the case (I was going to have to anyways for some RGB installations and cable management, but hey) and setting it up on the stove in our kitchen, with shop towels layered down to reveal leaks easily.
And…it worked! No leaks, 10/10, hooray!
Until I had to drain it.
Theoretically, the drain valve hanging off the side of the reservoir should enable easy flow out of the system. Just open a top port to introduce air to squeeze out the water, open the ball valve and pull the stopper plug, and bam – gushing watercooling loop. In practice, this also relies on the gravity of the case, with water sources elevated over the pump allowing the air introduced to push the water down through the pump and out the valve. Because the components did not have the advantage of their proper mounting heights and positions…it quickly became an irritating and despair-inducing hour of watching the drain valve spew little bits of water as I moved component to component, disconnecting fittings with shop towels in hand to attempt to contain the gushing of small amounts of water. I learned a lot about how water flows and backflows, though!
So it was an eventful evening, after which I wrapped everything in shop towels to dry out, shook out the radiators the best I could, and left our stove out of commission, covered in a graveyard of blue disposable absorbers with components swaddled inside of them like newborn children.
My nerves were quite a bit more shot after this!
Saturday was to be the most physically intense day of the 3-4 day span I had planned to use for this project. I would have to strip down my old desk, stash away all the other tertiary pieces of furniture in use, and then remove my PC from the desk and take the desk out of the house. It took longer than I would have liked, and a large part of that was a sort of bad feeling from the drainage incident the previous night.
I ended up watching the Shadowlands Season 1 Mythic Dungeon International as I worked, taking a lot of breaks and trying to pace myself to work through in bursts (I will likely talk more about this in a separate post later this week!). It ended up taking around 9 hours in total to strip everything down and have the empty space I needed to begin the last task I had for the day – building the new sit/stand desk.
We started on the desk at 9 PM, yikes. It was both easier than I expected (very few actual steps of hand-tightening hex screws) and harder (the folding mechanism to save space in shipping for the desk frame snapped back a lot until screwed down and I almost broke my finger getting it pinched in the frame as a result), but overall, it wasn’t that bad. The side brackets to hold the tabletop use a press-fit mechanism that requires very tight tolerances, so getting them to slot in for bolting down was a giant pain in the ass, but after that, it got exponentially easier. We flipped the tabletop upside down and the frame with it to bolt it down, unlike on my wife’s desk where I crawled around on my back with a drill eating sawdust (a very stupid idea in retrospect) so drilling the pilot holes and securing the tabletop was a lot easier. The hardest part after all of that was simply moving the assembled desk – it is, as one can imagine, a heavy hunk of desk, with a solid laminate tabletop made up of dense wood and the heavy steel of the frame.
No PC to use, I watched wrestling until I was ready to go to bed.
Sunday was all about prepping the desk for the new system – setting up all of the monitor arms, UPSes, cable management, clamping accessory arms, and getting a feel for the proper heights for sit and stand adjustment. It took a few hours, but after the previous day’s slog, this felt so much easier. This is also where my planning actually paid off – the monitors fit precisely as I expected, with only a slight bit of width my plan could easily adjust for, and the only other major pitfall was that one of the clamping shelves, which I intended to use for my PC tower, would make it such that the tower would meet a monitor in the worst way, so I removed it. Lucky for me, my wife wanted one for her desk after seeing them on mine, and now she has it!
We were able to mount the monitors onto their individual arms with relative ease, and after finding the needed tightness adjustments to ensure correct positioning, everything was looking pretty great. I setup my old PC on the desk, took a few photos, and then played for a little bit – my last day using the system.
The day my nerves were really up for – time to transplant all the watercooling hardware and other components into the new system, run a proper loop bleed and final leak test with only the pump powered, and then to install Windows and make sure everything functioned. I spent most of the day with my old system running on my desk, with YouTube videos to keep me entertained and motivated to finish the work. I first had to very carefully cable manage the system the best I could, as the only real disadvantage to my Cooler Master TD500 Mesh case is that the back tray has minimal room for cables, which makes pressing the side panel back on a real challenge if not properly managed. Once I had done all the work I could with the old PC on, it was time to uninstall the old CPU and RTX 3080 and prepare them for transplant.
This phase was fraught with some peril. The last time I had removed a Ryzen CPU was the old one my wife had once used, using it to flash the BIOS on a friend’s build to allow for Zen 2 CPUs, I bent a ton of pins on that CPU and it sits on my desk as a reminder. That CPU, at least, was a Ryzen 3 2200g and was cheap and not going to be used again anyways, so it getting obliterated by my caveman heatsink removal was fine. If I fucked up my 3900x, it meant being without a PC for some indeterminate amount of time until I could either get a 5950x as planned, or if I relented and bought something different to use.
The pressure was high, and…I got it out fine. Sort of. A small patch of pins at the top of the CPU got some thermal paste on them, but it was Arctic MX4 and that is non-conductive, so I didn’t worry too much. That would cost me later…but I’m getting ahead of myself.
The GPU was much easier to remove, but also the scarier possibility. If I shredded my 3900x CPU, I could buy a 3950x 16-core today, and while it would be missing the Zen 3 improvements, it would still be a better CPU and I could just dust my hands and call it a day. If, on the other hand, I was lacking in delicacy with my RTX 3080…holy shit, my system would be done, basically, as supply shortages are forecast to last through until the second half of the year!
It was quite scary to remove the stock heatsink, as MSI bolted the backplate down both from the front and back of the card, and there is a point in every GPU disassembly where you have to pry against the sticky thermal pads to get everything separated, and it never gets easy to do with your own hardware that cost almost $800. I triple-checked for every screw, and then pried away, and finally…a naked RTX 3080. NSFW! (Kidding, of course)
So it was time to mount the waterblock to it. The process was mostly simple, with one major caveat.
I’m an experienced system builder who has been tinkering with hardware for over 20 years at this point, I am conditioned, as a result, to note that when a thermal pad has a shiny film atop it, it must be removed. So I set out to remove the film from atop the preapplied thermal pads on the Corsair XG7 waterblock. However, a moment of RTFM – the manual does not specify to do this. It also doesn’t say not to do this, however. So I did this for all of the RAM thermal pads, before finally searching to see that in fact, it is intended that you do not remove them. However, as they made good contact and are still thermally conductive all the way through, it wasn’t bad and I was able to proceed, but yes – it was a massive pain in the ass to remove the film, only to discover that the reason it is that way is because you don’t have to touch it. Newbie watercooling indeed! So, oops – spent 20 minutes pawing at those damn things only to find out I didn’t have to. I applied Thermal Grizzly Minus Pad 8 to the backplate on the hot spots, including for the VRMs, GPU core, and around the power input section of the card, so I’m leveraging the backplate even more than Corsair intended, and the card makes an appealing sandwich in the waterblock and backplate!
With the graphics card entombed in copper and plastic for liquid cooling, it was finally time to install all of the watercooling hardware. I mounted the CPU and put the waterblock on it, mounted the GPU and locked it in with screws, had prepared all 3 radiators with fittings and set aside the lengths of tubing I had cut with the compression collars still on there to simplify things, and set about slowly moving through the case to secure everything together, with shop towels down in case the small amounts of fluid that I could still hear in the radiators (particularly the top-mounted 240mm one!) decided to make a grand escape.
Then I had to pull a chunk of it out because I forgot to plug in the adressable RGB controller that hides behind the pump to manage all of the LEDs not attached to the motherboard. Oops! Got that all wired up and remounted the front radiator, pump/res, and fans.
Everything came together and I triple-checked, as I explained the loop order and flow pattern to my wife as an extra precaution, forcing me to examine my narration to ensure stop plugs, fittings, and tubing were all securely mounted and tightened.
The time for talk was over. I grabbed my bottle of Corsair purple coolant, my squeezy fill bottle, and a funnel, and filled the squeeze bottle with 200ml of coolant, squirting it into the reservoir with an untold amount of caution (and shop towels tightly wound throughout the case). A flick of the power supply switch, and the sound of suction as the pump pulled the grape Kool-aid-looking fluid through, with an almost entrancing pattern mixing with air bubbles. Repeat about 5 times, and the loop seemed pretty full, so I left it running to bleed and leak test once more as we ate dinner and watched Cutthroat Kitchen.
Dinner ended, and as we were eating, I realized something critical – in my nervous state, I had forgotten to apply thermal paste to the CPU when mounting it, so it was sitting dry behind the waterblock with inefficient heat transfer. So I had to pull the waterblock, now full and with full tubing, off the CPU and get enough space for my sausage fingers to push in a tube of thermal paste and apply a dollop of the stuff. I did it, and reseated the waterblock. Then, I pulled the power supply jumper off the 24-pin cable – it was time to properly power on the new PC and get things running.
So I fumbled with it, and then finally hit power on.
A POST error.
The manual to the ASRock B550 Taichi included no POST code error table for me to read, so I had to Google what the C5 code meant (memory error, it turned out). The memory was brand new and I hadn’t even been into BIOS to set an XMP profile yet, so it either meant that the memory was bad (oh fuck please not that!) or that the CPU memory controller wasn’t working. I stressed out, hard, before I realized that perhaps, just maybe, the thermal paste that had gunked up the pins responsible for memory controller interfacing. They were, by position, the ones closest to the DIMM slots, so it made logical sense.
So I had to pull a full waterblock and tubing set away again, this time enough for me to pull up the retaining arm, grab out the CPU I had just caked up with thermal paste, and then clean the pins. I got it fairly easily, and plopped it right onto a shop towel soaked in isopropyl alcohol. Not quite good enough, so I poured some of the alcohol right onto the bottom of the CPU and let it sit there for a minute, then flipped it back onto the towels, gave it a gentle massage, and then grabbed a SIM ejector tool to scrape out any larger bits of thermal paste I could get at. I was losing my mind the whole time, because then, of course, I had to maneuver my large-ass hands into the small opening I had between tubes to socket the CPU correctly without full visibility, lock in the retaining arm, and then screw down the waterblock one more time, without bending any pins, failing the socketing, or putting excess pressure on the tubing and watching as my new system died in a flood of purple drink.
After a lot of frustration trying to find the screwholes for the waterblock in the dark (phrasing), I finally got everything locked in, restored power to the system, hit the button, and….it worked!
I had smaller problems identifying the correct 2 TB NVME drive to install Windows to (curse me for getting both a PCIE Gen 3 and Gen 4 drive to use the motherboard’s NVME capacity to the fullest!), but finally, late last night, it was up and running.
It also crashed because of XMP not working on the 3900x with 64 GB of RAM, but I was able to fix that and get it running at the proper 3200 MHz today by loosening up timings slightly.
And so, while it still has some work awaiting in syncing up the RGB illumination and getting my liquid temperature display fished out of a corner of the case and taped up in a visible spot, I now have Shadowbringer up and running!