I attached a pic of the screws I think I used in the aio with super thick server fans.
Also unrelated, Wendell from L1 has some trick with painters tape he uses to make mounting these a snap but im working rn and cant find the video for you.
Water cooling of 0.5-1 kW processors
- Thread starter DanRR
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Also unrelated, Wendell from L1 has some trick with painters tape he uses to make mounting these a snap but im working rn and cant find the video for you.
Holly #$@, I took 25-26.2W for Arctic 4K from the headlines of Amazon not looking at actual data sheet of manufacturer. And last one shows strange numbers with 39W at peak. That is what you have meant by 25-36W, i.e. its actual power is much larger than I mentioned ?
Also can anybody comment how we have to understand this confusing at first glance plot with the curves going opposite to common sense. They show that at zero airflow 8K needs more power 39W (which is a peak) than at maximum airflow 220cfm. And the more fan speed the smaller pressure and power consumption. Definitely they measured and plotted something different to what I mean
Those plots are running the fan at 100% PWM speed control, varying the static pressure, to show the maximum possible performance across the pressure/flow conditions, and yes when you block the fan entirely it uses more power.
That graph appears unhelpfully backwards with hard to follow multi-axis information. If you drop an axis and mirror it, the relationships seem more readable.
As the airflow becomes more restrictive, static pressure rises and CFM drops which requires higher power consumption? The logic seems to make sense to me. Running a fan with no obstruction at high RPM maximizes CFM through, but when you restrict airflow (such as with a radiator) CFM drops from the peak value.
OK, nexox and TrashMaster, now it's clear. Thanks. But what's follows from this is that the 26.2W Amazon mentioning is just the absolute minimum, ideal, free air, never in practice realized number, not average, typical or whatever anyone may imagine
Last edited:
Amazon listings aren't generally very reliable (I see a value of 3.96 watts on some Arctic 4k 120mm listing, obviously wrong,) I always look for manufacturer specs separately when it's important.
Gentlemen. I present to you, my magnum opus:
The Silverstone stock 120mm fan tilt test: 3W -> 4W
The Noctua 120mm NF-a12x25 fan tilt test: a 1.7W puff of air
The Arctic 140mm P14 PRO tilt test: 4W -> 6W
His highness, the Arctic 120mm 4k tilt test: 5W -> 8W
And finally... (I literally had to move to a bigger power supply)
The uncontested king of suck. The one, The only...
Arctic 120mm 8k tilt test: A WHOLE LOT ----> A WHOLE LOT MOAR
You've Gotta Be Kidding Me!(tm)
The Silverstone stock 120mm fan tilt test: 3W -> 4W
The Noctua 120mm NF-a12x25 fan tilt test: a 1.7W puff of air
The Arctic 140mm P14 PRO tilt test: 4W -> 6W
His highness, the Arctic 120mm 4k tilt test: 5W -> 8W
And finally... (I literally had to move to a bigger power supply)
The uncontested king of suck. The one, The only...
Arctic 120mm 8k tilt test: A WHOLE LOT ----> A WHOLE LOT MOAR
You've Gotta Be Kidding Me!(tm)
Attachments
Last edited:
what about not to close the output ?
or just close Ait out with a plate but not tilt ?
1000 kudos to you! This test will shake to the grount the whole sucking industry.
Noctua surprised. How come no effect at all? This probably was regular not their high power sample. Or they have something like internal drain power stabilizer?
Arctic 8k contradicts the plot we discussed above yesterday where power was 39W at peak. The whopping 50 Watts at peak will overheat all cheap fan controllers money can buy. I was hoping to lower voltage on it with the individual controllers to reduce the noise. I got also 4K, not installed yet due to specific screws not supplied with it and missing in the entire city hardware stores, while Amazon is slow on mondays, but its 4,000 RPM could not be enough, and the optimum cooling will be in 5,000 or 6,000 RPM range. Also at summer time RPM needs to be higher and adjusting to the weather would be a huge plus.
Can you try to undervolt the 8K? Or it will not like it?
Last edited:
I didn't follow/read all the replies in the thread; noctua has a variety of fans for different applications (read different voltages like 5v or 24v). Going from "normal" brown fans to indusdrialppc fans will very likel improve cooling/total cooling capacity
try a custom loop with 80mm thick radiators and arctic 8k fans, that should improve cooling. or if noise is important, get a few mora iv (https://geizhals.eu/watercool-mo-ra-iv-600-schwarz-26600-a3236632.html), you can mount 9x200mm fans on these
They are discussing 4k rpm and 8k rpm 120mm fans that are 38mm thick - noctua has no comparable products
Why should any sane person undervolt a PWM Fan? just lower the pwm frequency
The profile of the blades on these high RPM fans don't actually work all that well at lower speeds (regardless of how you get them there), say less than 30% which is what you would need to try and hit a sane wattage on these. You are far better off buying a lower power fan like the 4k and running it as intended. I can hook it up to a dimmer switch after work if anybody really wants to see what it does.
How? That's what processor regulates, isnt' it? I need that both external user and the processor still controlling its speed
Last edited:
For six 8K fans THREE such power supplies are needed or one with current rating 4.2*6 > 25 A and better 50A to make variable RPM unless nilfisk_urd will show how to do that differently. And Molex will not survive 150W power for 3 fans. We need lower its voltage and crazy wattage..
Last edited:
you sound like you don't have any clue how that works. just define a custom fan curve in whatever you use to control the fan speed. for example in the bmc fan control utility ... there you should be able to limit the maximum pwm percentage.
well, then you need a better connector - crimp yourself a custom 8-PIN EPS or 8-PIN PCIe adapter, these should be able to handle that.
Or if you have experience with electric installations and know what you are doing, you could just cut the 12V and GND lines from the pwm connector, put ferrules on the strands and connect them to a proper 12V PSU (eg MeanWell LRS series) - not that noname "cheap" stuff you posted
if everything fails
get a waterchiller for your waterloop, and peltier plate preferably one you can control with your core temps.
*note if you want to really cool it down, make sure your pipes, chiller are rated for acetone, ethanol - but distilled water as long as you keep it moving shouldn't freeze either even when its below freezing point - depending on temps you need it to move faster which may require more fancy chiller that can handle sub-zero/freezing temps, pumps are typically a problem if you are doing subzero - so you might need to install heating (wire-heaters) on some parts that can't handle those temps - the pump was main failure point years ago when i played with dice/winter waterloop cooling. You will also need to cover everything in vaseline (mobo, components etc), and insulation for cables/pipes.
get a waterchiller for your waterloop, and peltier plate preferably one you can control with your core temps.
*note if you want to really cool it down, make sure your pipes, chiller are rated for acetone, ethanol - but distilled water as long as you keep it moving shouldn't freeze either even when its below freezing point - depending on temps you need it to move faster which may require more fancy chiller that can handle sub-zero/freezing temps, pumps are typically a problem if you are doing subzero - so you might need to install heating (wire-heaters) on some parts that can't handle those temps - the pump was main failure point years ago when i played with dice/winter waterloop cooling. You will also need to cover everything in vaseline (mobo, components etc), and insulation for cables/pipes.
Last edited: