So I’m planning to build a new workstation in the new year as my (belated) college graduation gift to myself, and for use in my master’s program. I’m sick to death of waiting for stuff to compile on my mid-2014 MBP; my undergrad thesis ended up being so slow on it that the workflow for a week was to make a single change and then let it run overnight. I’m also sick to death of hearing my computer.
It seems like a custom loop is the way to go for silence. I dislike fan noise, but it’s definitely way preferable to coil whine/HDD noise. I’m thinking about the 5900x + 6800xt (for some gaming and some GPU compute), so I’m roughly planning for a 450-500W heat dissipatoin budget. I’m also really liking the look of the Define 7, which also seems to have good sound dampening and build quality.
I did notice, though, that there’s going to be AIO cooling options for the 6800xt, so I could switch to the open top on the define 7 and use dual AIOs. The downside to opening the top is probably more sound leakage of e.g. pump noise or potential coil whine on the GPU. Would a custom loop with a 280mm and possibly an external 3x140 or 2x200 rad be quieter than the dual AIO option? Has anyone experimented with passive external rads? I wonder if you could have a thin 400mm mounted horizontally somewhere with lots of air clearance to get extra completely silent cooling capability.
Custom loops can be quiet, primarily because if you have enough radiator surface area, you can lower the fan speed significantly. Unfortunately, even with variable speed water pumps, in my experience, the pumps still tend to hum while running.
A well thought out air cooled machine, using quality fans has the potential of being much quieter, especially when paired with a “quiet case.”
My experience in a nutshell: water cooling can be quieter when the computer is working really hard, but an air cooled machine can be quieter at all other times.
My recommendation: water cool because it’s a new challenge, because it is fun, because it looks great, do it for bragging rights. But don’t do it because it’s quiet during those times that you are compiling code, because most of the time you won’t be compiling code. Most of the time, you’ll be surfing the web, or performing other mundane tasks and during those times, the hum of the pump will get on your nerves.
You don’t think the “quiet” case would help as much with the pump? If I went the water route, I was going to set up an arduino to ramp up/down the pump and fans based on the water temps. It’s true that I’m not compiling code all the time, but there are definitely times where I’m doing multi-hour runs of simulations and I live in a shared space, so getting a noticeable drop in noise output during those times might be worth it. I’m also worried that a well-thought out air cooled case means as few obstructions for fans as possible, which also leaves more room for noise from coil whine to escape (which I despise)
The most common reason to get a variable speed pump is so you can set it at a static speed of 50%, or 70%, or whatever you need to set it at in order to keep it quiet. I have mine running at full speed, but it’s sitting on a sorbothane pad which stops it from rattling against the case, and that alone seems to be enough.
In any case I’d be more worried about fan noise than pump noise. You’re also better off using the PWM controls in the BIOS, and ramping up the fan speed based on CPU or some other system sensor. It takes a while for overall water temperature to change where CPU temp can change more quickly.
but it’s sitting on a sorbothane pad which stops it from rattling against the case, and that alone seems to be enough.
Well that’s good to know! I assumed most of the noise was from the internal workings of the pump, but if it’s vibrations on the mounting hardware that’s good news
It takes a while for overall water temperature to change where CPU temp can change more quickly.
True, but it shouldn’t affect cooling, right? The cooling effectiveness on the CPU should be a function of the delta T between the water and the heatspreader, so basing fan RPM based on CPU temperature seems like it would needlessly ramp fans up and down for burstier workloads without actually improving CPU temperatures, and setting it based on water temperature would still allow sufficient cooling for sustained loads. Using the PWM controls in the BIOS is almost certainly a win in the reliability department, though
In any case I’d be more worried about fan noise than pump noise.
Yeah, that’s definitely a concern. Was hoping somebody had tried out passively using radiators, because then I could just put one on a stand outside the tower.
That’s my experience with the D5, at least. Other pumps may behave differently.
Let me rephrase. You might find it difficult to get much granularity in fan speeds by measuring against the water temp, but yeah, so long as it’s stable it’s fine. Most CPUs these days will just downclock if they’re not happy with thermals.
There’s a few radiators made for PC watercooling which can take like 9 120mm fans, but if you’re aiming for 500ish watts I’m not sure even that would cut it. I wouldn’t expect to find a completely passive setup for that which also isn’t completely ridiculous(perhaps in a good way ).
Again, with enough radiator surface area and quality fans (Noctua, Noise Blocker, etc.) you’ll never hear the fans. Just use your UEFI PWM controls.
Radiator thickness plays a role, but it’s not as impactful on cooling ability, as the amount of area available for fan mounting.
I use a pair of 140x280 rads in a single loop and five noise blocker fans to cool one 140W CPU and one GPU. The coolant stays below 50c and I never hear the fans, even during taxing work loads. The fifth fan draws air across the VRM and exhausts air out the back of the case.
I use a variable speed D5. I only use the faster pump speeds during the process of bleeding the system, while trying to dislodge air pockets. Once the air has been purged, I throttle the pump down to a constant ~40-50%. It doesn’t make a lot of noise, but the rest of the system is so quiet …
I’m successfully dealing with 300-400W with passive radiators in an external loop, and I’m aiming for 600W in my final configuration.
The easy and cheap way to do 200-300W or so passively is to just get a MO-RA3 with stand. Relatively affordable and well-tested. Easily does twice that wattage if you add 9 slow-spinning fans.
You want massive amounts of radiator area to do anything passively, and thickness also matters. Don’t bother with anything thinner than 60mm. The MO-RA3 can fit 9x140 fans.
My setup is a bit pricier, two 840mm Aquacomputer AMS radiators in full copper. 840, or room for 6x140 fans each. I want to double up and get two more…
As said, pump speed is more about finding comfortable rpms for noise than anything else. (Aim for 80-100l/h flow, and adjust slightly upwards or downwards to find subjectively less bad noise.)
For control, what you probably want is to tie the fans to the difference between water temperature and ambient temperature. It changes slowly, and you can reliably do a semipassive setup that keeps the fans on under low-medium loads and on under heavy load, where they don’t jump on and off.
My old setup was a 480mm rad operating semipassively (tucked away in another room.) I had it set so that the fans would kick in at 10C difference between water and air temperature, and turn off at 0.5C difference. At low-medium loads it’d stabilize at 6-8C above ambient. See graph alone for how it looks in practice. I used an Aquacomputer aquaero for control. really neat little device.
).