Serious talk: Building an AM4 system - Hear my thoughts and tell me your recommendations
In the next month or so, i will be finally be building a new system based on an AMD 3950X.
I was seriously thinking of air cooling for various reasons, but turns out there will be some limitations; mostly clearance/size related and also noise related as you are forced on having good airflow THROUGHOUT the case, and are limited by the surface area of the specific aircooler heatsink you plan on using.
With watercooling you can add as many radiators as you can fit so you can reduce the airflow needed and therefore the noise (not technically true in reality, just up to a point). You just need an appropriate air intake (assuming you don't mount them externally), and have the radiators exhaust only; thus you don't really care about the airflow inside the case (assuming supporting components won't need active cooling)
I've been following the PC space for years, but haven't build a personal system since ages ago, so i am not fully informed in specific existing watercooling products. However, i'm not just a random guy, i've been modding since the Pentium 3 days (what's RGB leds when you can have fluorescent IDE cables with Neon tubes lol), and i'm actually a mechanical engineer now.
The main problem i see with the existing watercooling products are weakly designed coldplates. From disassembly photos i've seen, the fin stack mostly don't even cover the full area of the IHS and sometimes even die(s) which is detrimental to performance. Also the fin height is pretty short in all of them (however fin height reduces in effectiveness due to the linear hear transfer from the plate). I know this increases manufacturing time and cost since most coldplates are slotted with a disc, but with skiving machines entering this market density, area and fin height should be no problem and would expect about 10mm of fin height. Especially now with multi-die CPU's becoming the normal (even Intel will be using special glue lol) this is more important, since you can't expect the IHS to transfer heat effectively, meaning you need to close up the distance to the dies by having the fin stack on top of the dies.
Does anyone make such a waterblock? Because from what i've seen pretty much all are close in effectiveness and design of the coldplate; i am looking for the best* one. *For me best is: it does the job it's supposed to, effectively for it's cost. There is no bad product, just bad price.
I need one that will not only have the capacity (assuming overclocking, about 180-200W) and be effective at it (so that temps stay lower to allow for better overclocking in the first place). Many are really overpriced in my opinion for their performance. Much like other PC products in the last 5 years but that's another story.
In terms of the rest of the build, what i am planning is to have is 2 or more radiators (like 2x 360 or bigger/similar total area) connected in series. The outlet goes to a reservoir (or 2 that are interconnected), and then 2 pumps (which could be integrated to the reservoir/s) feed separately the CPU and GPU, which output to the inlet of the radiator(s). This way i don't have one radiator not contributing when only one of the CPU/GPU is at load, and i can control the flow to each one separately. Generally, the bigger, the better.
As for the case i still haven't got around to it, but ideally it needs to fit an eATX board (even though i am getting an ATX board for now), extended length GPU (or 2 ideally), 3 HDDs min (4 ideally), 2 SSDs min, dimensions: 74cm height limit (should be more than enough), 54cm or less deep (not hard limit, but ideally), and no limit in width but lower is better. Glass window panel and air filters are a plus.
I don't care about RGB or info oled displays, i was doing these when i was a 15year old, these are just optional things for me now.
In all seriousness, if any of the products available for me to achieve decent performance are overpriced i will go with the custom route. It's not the first time and actually i am almost finishing now my functional mechanical keyboard prototype (pcb, case, etc all custom). It's not the first time i design and manufacture a custom solution when there is no effective solution to my problem. Although i do mostly automotive stuff it's still the same process. You might say it will cost you more, well yeah in some cases yes, but generally in such cases it will be better performing, i get to gain more experience and if i make a few and sell them i get back my development cost.
I was going to talk about the elephant in the room but decided not to, because it is potential for patenting. So let's talk about 2 different elephants:
- Why aren't you using automotive coolant? It's been here for ages; it's reliable, does mixed metal, doesn't grow bacteria or corrode stuff, and it's cheap.
- Why nobody makes a waterpump that has a mechanical coupling from the motor to the shaft, instead of the "magnetic coupling" used by D5/DDC? These pumps have *****ty head pressure. Ever used a magnetic stirring bar? It is weak as heck. Plus they are inefficient (in terms of energy transfer) and from what i read they are unreliable as well. The only thing they are good about is possibly noise and no motor-to-pump shaft seal.
- I would start to complain about fittings and tubing also but let's leave it for another time.