I was playing around with a 60mm Delta, which was strong enough to make into a mini hovercraft (made out of a sheet of cardboard, and a plastic bag as the skirt), but I accidentally got my finger too close and it chopped down to the bone :( Also broke the fan blade :(
@@phuzz00 Ouch, I've had a couple of deep cuts from 92mms where the grille leaves a "wide enough to be dangerous" gap around the edge but nothing that bad. I think so at least, I immediately superglued the wounds shut so I'm not sure how far it actually went!
11:34 with all that airflow, you could have reversed the flow of air through the radiator and had it blowing across your fan controller board, and probably the tubing and motherboard for even more cooling XD
@@der8auer-en isn't it possible to just stick some rads on top of the chips? I mean it's still quite common to see rads on transistors, controllers or SoCs to improve dissipation. Here it seems the controller had some over temp protection but a custom PCB with high current rated TO-220 mosfets screwed on a rad can easily handle switching a 7-10A load.
Kinda crazy how good noise suppression has gotten passively for everything. That thing must have been deafening, like hearing protection levels but on the mic speaking comes through clearly. 15 years ago nothing would make it through.
@@Adamgreen735closed loops are designed to handle higher water temps - 35c is more than fine. But 60c water in customer loop is asking for trouble, it can damage the tubes and kill the pump
There are certain kinds of hard-line that would melt with that water temp. Its also not good for any of the seals or o-rings. The general rule of thumb with water cooling is 100W input per 120mm fanspace, which will achieve a 10C coolant temp over ambient. If you run the loop hotter, you can reject more heat per unit of radspace at the cost of increased noise and stress on the loop components. And then there's my oddball X299 setup where I'm cooling 88W on a 92mm radiator.
@@George_McGregor no, it's only applicable for the metal housing variant "Max. system temperature -10 to + 95°C for pumps with brass housing (non-freezing) +/- 0 to + 60°C for pumps with plastic housing (non-freezing)" it's in the official PDF that details preassure and flow curves, exact dimmension, rotor speed ...
The fans will work fine, hard wired. There is no need for a PWN controller. Also, you were running those fans unrestricted through a free-flowing radiator. There is a lot of air moving, but what end up happeneing is that air at the surfaces of the fins become sort of a "Hot Coating" that lubricate the "Good Cold Air" directly out. Those highspeed fans need to be presnted with a restriction system on the back side to increase the prissure in the radiator, causing the air to compres more and force the hot layer to interact with the cool layer. This is a nice experiment, just not on an expensive GPU like that. Dust off a heat-hog and give it another go. With back pressure baffles. This is the reason rack servers have so many baffles and relatively small openings for cooling. It's more about pressure management than CFM. I wish RUclips allowed the poasting of diagrams and pics as a response. So you can see the phenomenon I am talking about. Those fans create a high static pressure in the enclosure, causing lots of mixing, then smaller vent slits let "some" of the air out.
I have a 3000 RPM exhaust fan in my system. There's gaps around it, so I tried covering them with electrical tape, to reduce recirculation. After a few hours the fan had blown the tape off the back, even though it wasn't in the path of the fan. The tape didn't overlap the fan at all. That just shows the huge pressure differentials these fans can create.
You can create a diagram and upload it to imgur then post a direct imgur link in the youtube comment to link to your diagram. I've done that before and it's allowed.
I would like to learn more about this. Do you have any video sources discussing it further? I've worked with servers randomly for many years, and was always surprised how restrictive the air flow is.
Hello Roman, maybe what happened is when you tried to add two more fans, those additionnal fans were producing current (turning the fans in a wind turbine) and that current going back to the controller may have fried one of the main component of your fan controller. Stil happy to see those experiments. Cheers.
Next time time for fun, use fan cone or cylinder. I had my share of cooling projectors and lasers, there is a huge dead zone in the middle, basically you left your center radiator fins air starved, it is proven until you hit equilibrium.
I assume that these 4 fans plugged probably tripped OCP protection on PSU, not sure what their starting amps are for fans, I tripped few times OCP protection on Aquaero and Quadro Fan controllers with few fans That's what we like from you Roman, crazy things who nobody does, awesome video mate
Yea! I have done that on server boards with those Blowiemetron delta fans too! They blow a lot, but it doesnt really gain much in terms of thermals wrt the cost or the difficulty of installation. Edit -> I ended up making a separate fan control board with a separate SMPS using the RP2040 to control the ramps from the mobo commands. Waste of money, it was.
Startup amps for those brushless fans is little if any higher than at their full speed. You might be thinking of brushed motors, especially with more load on them than just pushing air. They're probably no more than about 1.5A peak rated and in use, probably 1A or less which is pretty trivial compared to powering a computer and gaming video card, but yes that could overwhelm a fan controller not designed for that much current. It wouldn't trip OCP on the PSU.
These are server radiators, and Alphacool also offers water-cooled server chassis that you can install these in. In my homelab, I have two 2U chassis, which significantly help in reducing noise compared to air-cooled servers.
Nice and quiet, that's all you need to know 😂 From the time I've spent in data centers you have to raise your voice when you are around those little fans at full tilt, they are so loud!
I feel like when Roman talks about something, that means it’s worth considering in terms of technology. Not like he’s a puff piece sort of guy, right? Silly weird artifacts nonwithstanding, he’s a real source of good information
Have to agree... While I deeply enjoy those deep technical reviews where Der Bauer shows his dedication to getting all the science right.... NEVER stop with these kind of fun things... Just to prove people wrong ;) It's almost like those radial fan blowers we used to have on GPU's.... were not the most stupid idea ever!
@@FreakyAngelus that's what I kinda meant, yeah, those technical videos are awesome, but kinda everybody is doing them, on the other hand, those "experiments" are way more entertaining, you can't find many channels doing it on youtube, old LTT was perfect place for it but they stopped them, everybody is like "I'm not cOmFoRtAbLe WiTh 4090 RuNnInG oN 360 rad" meanwhile derbauer is like "hello, in this video we will try to cool 4090 with 40mmx2 radiator" xD
0:20 There really is no actual answer to that. It depends on how cool you want the GPU to be. It could range anywhere from 'just cool enough to work without dying' to 'The coldest the GPU could be allowing for the laws of thermodynamics on planet Earth'. There is a LOT of room there.
At some point, the radiator would melt, so there's your actual maximum. I would imagine the liquid inside would boil and blow the radiator up before that. Now i want to see a radiator pushed to literal failure. Thanks.
Delta fans always bring a smile on my face with how delightfully absurd they are. I _swear_ Extreme-G used tiny fans like that for their bike engine sound, with only minimal editing required.
There is some weird greek letter based math that can calculate a decent prediction of inlet/outlet temp based on the thermal conductivity and specific heat of the aluminum, coolant, area of the radiator fins, and the air flow over them. IIRC, there is a counter intuitive relationship with air dwell time in contact with the heatsink, air can move too fast past a heat sink to pick up as much energy as it could. That being said....when you're cramming cubic meters of air across 80x40mm of zigzag metal, that relationship falls to the good ole American proverb "effectiveness by volume". Usually its used by artillery, bit the sentiment is the same 😆 Sidenote: I was just looking at these kinds of fans for a side project unrelated to PC hardware and cooling. So big thumbs up for incidentally covering something I wanted to see a live demo of! 😎👍
A trick I use when filling the loop is to use boiling hot distilled water, heated in the microwave in a pyrex beaker to avoid contamination. It gets the air out quickly as the air volume shrinks massively when cooled, plus the water is now sterile so as long as the loop remains sealed it shouldn't need a biocide additive! No guarantees but it works amazingly well for me! I always do a double flush of a new loop as well, to be sure there is no crap in it.
Proper way to attribute heat dissipation properties is so-called "thermal resistance". It can make an abstraction of elements in the cooling loop, starting with the heat source and all the way to the fans.
Impeller and pressure is very important. A good impeller design will prevent cavitation at higher flow speeds. If that checks out, you just need a good fan setup to get rid of that heat quickly.
Going push-pull with counter-rotating fans on a 12 FPI radiator is so hilariously overkill, you’ve got enough pressure to lift a car. Need a ridiculously dense radiator.
Actually, there is a smaller double 40mm radiator with barbs permanently attached that I bought almost 15 years ago. It is only 25mms thick. I still have it and never used it, for obvious reasons :D.
@fermitupoupon1754 Yes, let's put a full loop together to just cool an SSD, or better yet, use it in an already built loop and cause massive flow restrictions :D.
You can usually guess if a push/pull will make a large change. If it's a thinner rad or has wide fin spacing, there's not much resistance at all and adding extra fans won't produce a considerable change, if any at all. If it's a beefier rad with a thick core or tight fin spacing, push/pull can have an impact.
As a Delta Electronics fangirl(and fan nerd), I loved this hahaha Now you only need to put the fan controller into the airflow and it won't overheat :D
I've had a Mobo that did do 3A each on 2 fan connectors, I ran 2.7A fans on it. The magic of the pwm on fans is that you can just pipe a trace directly from the 12V connector, the pwm itself is done on the fan, so the controller shouldn't get warm unless it's actually doing dc speed control.
These 40mm rads are very interesting units. I first came across them when an ex-customer involved in High Frequency Trading and crypto in the early days requested for my company to procure the servers. I believe it was an ICC Vega "server" where they had a Core i9 overclocked to 5GHz (you want high single-threaded performance for HFT to validate and push trades faster rather than high core counts) and used a triple 40mm rad to handle the heat. The IT requirements in that part of the finance industry is rather interesting. They'll ditch any form of cybersecurity necessary just to get lower latencies - no firewalls or managed switches (for VLAN segmentation) because they add latencies. There are even specialized unmanaged switches for this industry that are designed to purely focus on ultra low latency switching
That was a really interesting experiment! One probably wouldn't expect a radiator that small to work with that much wattage. It is far too loud outside of a server room, but it does work.
I had a similar discovery with a cheap fan controller. I bought one that has 4 rheostats and support 8 5v fans. I bought a 4U 24 bay case and the air flow for the drives was lacking, so I put better (and quieter) fans in there, but cranked them all up to 100%. While I was messing around with other cables it moved once and hit the side of the case. Everything shut off. To my horror, it turns out that some of the contacts/pins from the factory weren't grounded or insulated in any way and it shorted out when it touched the case! So I coated anything metal in superglue and that fixed that issue. Then I noticed that it was putting off A LOT of heat, I touched one of the 4 ICs and burned myself like you did and was like "holy shit that's hot!". Luckily I had a bunch of small heatsinks that I purchased for a Pi 3 or 4 long ago and they fit perfectly! 1 heatsink on each IC definitely helped drop the temps down to reasonable, but still warm/hot to the touch, the moderate airflow in the case also helped cool them down. We always forget how hot tiny things get when we force a lot of power through them hahaha
This is awesome, impractical, and exactly why I love this channel. 👍👍👍 Considering the hi-perf fans that rad will use, I am absolutely shocked at the sparse tube/fin density. I would expect with such powerful fans that you could barely see light through the stack.
When Roman first powered those fans on I was reminded of the time my buddy had his switch go out at his house and as a joke I brought over an old Cisco 3750X. When I powered it on and it started screaming during the fan(s) power cycle test he just about lost it. I said "what, its fine, it works great." People have no idea how loud a datacenter can be.
Some of the HPE servers at my workplace have fans that pull over 40w. IIRC the doublestack fans pull 60w... And theres 6 of them in a server. The noise and air movement is something to behold
12:05 I mean you could use a splitter cable, you don't need to power the fans with the fancontroller you could just get the PWM signal (if they are even PWM fans) and power the fans themselfs straight from the psu.
I love these tiny radiators. Mine are a bit larger, designed for 80mm fans in 2U chassis instead of these 1U rads, but I have a grand total of 2x80mm cooling a 14900K and 7900XTX. Noise isn't a concern, but size was for a custom case.
The reason the clearance doesn't fit a socket head bolt, is because it's intended that you use hex head bolts or install the nut on that side. The fan body is then able to act as one side of a hex socket when tightening it down. In extremely cramped situations you can simply push the hex head against the fan with your thumb, if you can't get a spanner into that spot.
This reminds me of the time I used a Vantec Tornado on a Swiftech MCX4000 cooler without a fan controller. Thanks Roman for the flashback with that fan noise
Please use hearing protection around those fans. The sound they make is more than loud. It is not bandlimited. It contains energy at extreme frequencies. The energy in a wave is equal to its amplitude times the square of its frequency. Loud sounds will make your ears ring, high sounds will make your ears ring, loud high sounds will destroy your hearing!
Good luck on getting your part. I ordered my parts from watercool a couple weeks ago last year and my parts we to delivered on the first of the month and they haven’t even shipped yet.
just at the beginning of the video i know there is limitation of how much a powerful fan can extract as heat from a radiator, the limit is the conductivity between the water inside and the copper of the radiator. it takes time to move the heat and spread through the surface of that radiator, no matter how fast the fans are. you can increase that speed by just increasing the surface. the same happens with the am5 IHS small and thick, creating a "jam" in the heat transfer.
Having some stators on the back of the front fans might help shove more air through the heatsink by redirecting the swirling coming out of the fan into the fins and eliminating the relatively dead spot directly behind the fan hub.
I love experiments like this! They're wacky, and maybe aren't directly applicable, but hey, now I know that it's possible to cool 300w with minimal space in a server chassis if I ever needed to.
2:18 finally somebody says it the standard curves in an AIO are set sooo agressive often that it isn't really much more quite, but if you allow the water temp to get hotter you can mentain said tempreture easier because the same radiator will dissipate much more energy. Ofc it will not cool the cpu to the lowest temp anymore but that is fine. Turned my rig from being pretty loud whenever I fired up a game thats semi hardware intensive to staying at a comfortable noiselevel just by raising the watertemperture (about 20 degrees I think funnily enough, the stock settings of the corsair radiator are really quite agressive, I made sure to stay in the specs as to not risk damage to the tubing and pump tho)
A fan usually has a 4pin connector, +12V, gnd, pwm and a feedback of the rpm You could cut the 12v and gnd wires and hook them up directly to the psu and plug the 4pin connector directly into the motherboard With this you can control any pwm fans without frying your mb
Quality content and precise tests are really cool and all, but sometimes people just need some funny video for goofy questions answered and im glad you're doing just that 😂
Is it just me or those rads could have a denser fin array when matched with those fans? 2 of those probably good enough for a 4090 and 4 for a whole system. That would be a good, why not video xD
@@WayStedYou I don't believe so. Such 40mm form factor are used in servers only. And even thinner ones are very good at static pressure so the fins could have easily been twice the density without an issue.
Using plenums between the fans and the radiator would have given a much more uniform flow through the radiator matrix, significantly increasing the efficiency.
Cool experiment. Weird radiator. Fin density and surface area is extremely low. Would have been interesting to see the performance with double or triple the fin density, but no point in wasting time and money on it. No matter how perfectly the fans gets matched to the airflow restriction, the setup would still be way to loud.
Yeah, those delta fans are really optimized for static pressure over just pure airflow. They would've been perfectly happy with much denser fins, or a radiator that was 4* as long. Normally they are used in 1u rack servers to blow air throughout the entire 3 foot long server case and through all sorts of different heatsinks.
What's funny is these little thickbois can handle just as much of not more than a generic thinner radiator 2-3x their size. I got an EK240 AIO a couple years ago and I was sorely disappointed by the radiator and promptly replaced it with a thick 240 from my old loop and the difference was significant. It's the first and the last time I'll ever buy an AIO.
This feels like it's sized to enable an internal console liquid cooling setup, while Lian Li back in the day made an external case for the xbox 360 to be able to transfer the guts of such a console into a pc sized case to enable a larger amount of airflow to keep things in control. Granted, that was probably made before knowledge of the target temp behavior the console was doing.
Clearly static pressure isn't an issue, you should stack a few of the radiators in series to get closer to a counter flow heat exchanger. See if gets any closer to being practical
I had built one of those that was double size (4x fans). Ripped out 8 Dell Server blade fans like what you have, just without the double stack option. Im glad we're all a bit mad
Not to mention, running hot water through your loop's going to lead to faster evaporation. My GPU loop's lost over an inch of water more than my CPU loop after a year of hard use.
Where does it evaropate to though? Anyways, would it not be rather question bigger airgaps in rad? Try running 100% pump to see bubbles out. If you have opening to air in your loop, disregard my comment.
Always love this kind of videos showing these weird cool things about computer hardware. Also question: Is the fan controller controling the voltage to fans or only sand out PWM signal but fans get power from PSU?
it seems that controller board provides the whole power with a buck regulator. Otherwise, there is no reason a simple PWM signal circuit would overheat as such.
Now THIS is content. I have not played with counter-rotating fans, but the ones we did were powerful on their own. San Ace baby! If you can try their 120mm that runs at 24 volts (9HV1248P1G?) You should be using PWM control with the fans, not the usual CPU 2-pin or 3-pin (2 pin with RPM) voltage control. That way you would not burn up fan controllers, the fans themselves have control. Or if you did it this way...fan power should be straight +12 and negative, no PWM. OR - is it a Delta fan thing, my Delta blower has weird PWM control and didnt really work with my homemade controller, unlike the San Ace style (PWM wire: ground for minimum speed, open for maximum speed, PWM to vary the speed) which worked fine. Hope you can test it!
I basically did this when i tried to make the thickest 120mm radiator sandwich ever. push pull configuration with 2x38mm thick 120mm server fans, The sound levels were tremendous! Terminated both fans directly to molex since each one drew ~45W
As you probably already figured out, there is a limit to how much radiators can radiate. Even with excessive air throughput. Its just a limitation of the heat passing to the surface area from the water to the fins. IE the lack of surface area is the bottleneck.
If you plugged the fan while it was rotating from the airflow, it probably was generating negative voltage, that got dumped into the fan controller killing it
Now to watch you cool that card with the MoRa IV 600 while it sits outside of your window dangling over the street below! Subambient cooling like we used to do in the early 2000s ;)
I would love to see a system that incorporates all of the liquid and air cooling tech in one package. Even a NOS system that drag cars use would be cool. Maybe some kind of water to air intercooled computer, like wild cars use to cool inlet air temps. I know that sounds very tong in cheek, but Im being serious. Could be interesting to brainstorm! : ]
brave brave man i appreciate you so much for doing such experiments. i guess you maxed out your psu + fan controller thermal and wattage limits with those tiny fans.
I think you have to leave some gap between the fans and the radiator, otherwise the area behind the fan hub has no airflow. You were effectively only using like 50% of your radiator if you place the fans on the radiator, maybe add a 3d printed duct in between. This should help tremendously with push/pull too, fans don't suck in air properly when it is right up against a rad.
I know this was a fun project but a slightly more practical test would be to use a single 120mm radiator. I ask this because sometimes a mini-ITX case like the Fractal Ridge or the Terra can fit only a single 120mm radiator and it would be fun to see if it can cool both the CPU or the GPU with that.
You can use an electric domestic under sink water heater maintaining a target temperature to measure the radiators heat dissipation rate by measuring the heaters power consumption over time and the difference in temperature between the water and the room's air. I'm currently doing this. For a 350mm x 120mm radiator with three cheap fans I'm getting 22 W/K max, and 960 W max. That's at a dT of 40°C - 45°C. Meaning that if the room is 20°C then the water would be 60°C to 65°C.
You can do this the characterize the performance of any radiator and fan combination. You could use this to objectively compare PC radiator performance. W/k/CFM is possibly a good metric? The higher the better.
This metric would also enable you to know before you build a PC what the water temperature would be for the water loop given the system component wattage. Also given the fans cfm vs db(A) you could estimate how loud it would need to be to maintain your target operating temperature, OR what temp the system would be at a given loudness.
Nince idea Roman. I would have used the Aquacomputer fan controllers, like the aquaero with the water cooling backplate for 30W per channel from the built in fan controllers( I have the air cooled backplate). It will also expand with their other single 30W adaptors. I use one channel for my D5 pump and the other 3 as fan controllers. Six fans on my MORA 360 and the other six are run off the pump's fan controller. I've got temp sensors on every connection in my loop. Aquasuite is a good app but a memory hog, that still requires HWINFO64 for full functionality. My 1st one from 2017 is still on the job. I expect long lived devices. It's not cheap but it's programmable with a mono lcd screen that has past the test of time. It is still very readable across the entire screen with no burn in. Although from day one I've only used the lowest brightness settings for everyday use. Only when I access the controller directly via remote or the touch buttons does it light to full brightness, no burnt out pixels but I do have a little pixel dimming on low brightness, not noticeable unless you stare at the screen for the full cycle of displays. I've got a second one that I was going to use on the MORA 360 as a stand alone but I've got extensions from the one in my pc. So it sits as a "Hot Spare". Do you really need a color lcd screen that burns out in a year or two or a oled that may suffer burn in? I can see why after 7 years of constant use (it displays time and date when the pc is off) they made their choice of display tech to use.
Maybe, 2 fans in a "pull" config would have worked the best. So that the extra few watts of heat from the fans themselves are not being dumped into the radiator.
It makes me wonder if the heat sinking performance of the tiny rad under burst loads could be improved if the plumbing inside were zig-zagged, increasing the surface area by lengthening so it is an endless chicane from start-to-finish.
But they are server motherboards that has a lot sturdier circuitry for the fan connectors. Server fans can be pretty brutal and the motherboards are certainly not designed the same as normal motherboards.
Maybe not in stable operation, but those sorts of fans are used in 1U servers all the time, and if that server was cranking dual 300W CPUs and/or a power hungry GPU to 100%, you would hear them spin up. Just very rarely would they hit 100% speed because there’s probably a bank of 6 to 8 of them at least (plus maybe one more in each PSU).
now you need to watercool your fan controller too.
Could have at least aircooled the fan controller.
There are water blocks for Aquacomputer Aquaero 5 and 6 and its fan relays. Heck, at some there were waterblocks for DDC pump electronics....
Solid state cooling soon
Feedback loop hehe
Never going to happen@@arcadealchemist
Quick, someone confiscate Roman's Dremel before he really does go too far!
😅😂😅😂
No worries he has a hammer and tape to fix anything.
You cant; he's German so precision engineering tools are in his blood!
😅
I was hoping he would upgrade to a sawzall
I know those fans. I used to babysit half a rack of 1U servers, where each of them had 6 of these at 100%. It was a safety hazard.
I own 32.500 RPM as well as 38K RPM fans for my 3D printers.
I was playing around with a 60mm Delta, which was strong enough to make into a mini hovercraft (made out of a sheet of cardboard, and a plastic bag as the skirt), but I accidentally got my finger too close and it chopped down to the bone :(
Also broke the fan blade :(
No joke, the tip speed of the blades on the 21000RPM side are travelling at 300kph and can do some serious damage to fingertips.
@@phuzz00 Ouch, I've had a couple of deep cuts from 92mms where the grille leaves a "wide enough to be dangerous" gap around the edge but nothing that bad. I think so at least, I immediately superglued the wounds shut so I'm not sure how far it actually went!
@@phuzz00 was it the model with aluminum fan blades? Those look brutal if you ever stuffed a finger in there by accident.
11:34
with all that airflow, you could have reversed the flow of air through the radiator and had it blowing across your fan controller board, and probably the tubing and motherboard for even more cooling XD
Yes, fan controller and power source should be in air path
yea next step would be to have a custom block for the fan controller xD
😂
@@der8auer-enI was thinking that while you were talking about how hot it was.
@@der8auer-en isn't it possible to just stick some rads on top of the chips? I mean it's still quite common to see rads on transistors, controllers or SoCs to improve dissipation.
Here it seems the controller had some over temp protection but a custom PCB with high current rated TO-220 mosfets screwed on a rad can easily handle switching a 7-10A load.
Kinda crazy how good noise suppression has gotten passively for everything.
That thing must have been deafening, like hearing protection levels but on the mic speaking comes through clearly.
15 years ago nothing would make it through.
When you have a watercooled 4090 but also absolutely need the airplane sounds for immersion in MS FlightSimulator :)
now time to link fan speed to throttle in the game. "I cannot stop or my card will nuke itself I must fly"
12:38 the 60°C water is also not viable for long period of time because this is the max operating temperature of D5 and DDC pumps
Yeaah i always thought my ICUE AiO was hot at 35c and was always turning the fans up to cool the water more.... Now i know it will be okay at 35c.../
@@Adamgreen735closed loops are designed to handle higher water temps - 35c is more than fine. But 60c water in customer loop is asking for trouble, it can damage the tubes and kill the pump
There are certain kinds of hard-line that would melt with that water temp. Its also not good for any of the seals or o-rings. The general rule of thumb with water cooling is 100W input per 120mm fanspace, which will achieve a 10C coolant temp over ambient. If you run the loop hotter, you can reject more heat per unit of radspace at the cost of increased noise and stress on the loop components. And then there's my oddball X299 setup where I'm cooling 88W on a 92mm radiator.
Original D5 (Laing Xylem Lowara) Standart Work TF95C = 95°C
@@George_McGregor no, it's only applicable for the metal housing variant
"Max. system temperature
-10 to + 95°C for pumps with brass housing (non-freezing)
+/- 0 to + 60°C for pumps with plastic housing (non-freezing)"
it's in the official PDF that details preassure and flow curves, exact dimmension, rotor speed ...
RTX 4090: *93°C memory*
My laptop's memory is laughing in the lava: 103°C in a stress test, I had to make custom cooling.
u forgot about your (and mine) 13(4)980hx at 95 all the time.
My razer blade 16 with only a 4070 reaches 110 and that is 100% normal which is fucking mad
@@butters2160 tbh it starts to throttle down at 95, at max fan I can keep it below in benchmarks or at 95:)
I get maximum 80 C on memory with my Asus TUF 4090 . I don`t understand how is he getting 93 with water cooling.
@@HDRGamingHub I thought we are talking about 4090 mobile lol.
9:32 5090 😳
freudian slip
The fans will work fine, hard wired. There is no need for a PWN controller. Also, you were running those fans unrestricted through a free-flowing radiator. There is a lot of air moving, but what end up happeneing is that air at the surfaces of the fins become sort of a "Hot Coating" that lubricate the "Good Cold Air" directly out. Those highspeed fans need to be presnted with a restriction system on the back side to increase the prissure in the radiator, causing the air to compres more and force the hot layer to interact with the cool layer. This is a nice experiment, just not on an expensive GPU like that. Dust off a heat-hog and give it another go. With back pressure baffles.
This is the reason rack servers have so many baffles and relatively small openings for cooling. It's more about pressure management than CFM. I wish RUclips allowed the poasting of diagrams and pics as a response. So you can see the phenomenon I am talking about. Those fans create a high static pressure in the enclosure, causing lots of mixing, then smaller vent slits let "some" of the air out.
you are a genius
Yeah, it'll work, but you heard how loud those suckers were. The inside fans run at 70,000 rpms and will run at that speed constantly.
I have a 3000 RPM exhaust fan in my system. There's gaps around it, so I tried covering them with electrical tape, to reduce recirculation. After a few hours the fan had blown the tape off the back, even though it wasn't in the path of the fan. The tape didn't overlap the fan at all. That just shows the huge pressure differentials these fans can create.
You can create a diagram and upload it to imgur then post a direct imgur link in the youtube comment to link to your diagram. I've done that before and it's allowed.
I would like to learn more about this. Do you have any video sources discussing it further? I've worked with servers randomly for many years, and was always surprised how restrictive the air flow is.
Hello Roman, maybe what happened is when you tried to add two more fans, those additionnal fans were producing current (turning the fans in a wind turbine) and that current going back to the controller may have fried one of the main component of your fan controller. Stil happy to see those experiments. Cheers.
Great observation!
Next time time for fun, use fan cone or cylinder. I had my share of cooling projectors and lasers, there is a huge dead zone in the middle, basically you left your center radiator fins air starved, it is proven until you hit equilibrium.
I assume that these 4 fans plugged probably tripped OCP protection on PSU, not sure what their starting amps are for fans, I tripped few times OCP protection on Aquaero and Quadro Fan controllers with few fans
That's what we like from you Roman, crazy things who nobody does, awesome video mate
Yea! I have done that on server boards with those Blowiemetron delta fans too! They blow a lot, but it doesnt really gain much in terms of thermals wrt the cost or the difficulty of installation.
Edit -> I ended up making a separate fan control board with a separate SMPS using the RP2040 to control the ramps from the mobo commands. Waste of money, it was.
Startup amps for those brushless fans is little if any higher than at their full speed. You might be thinking of brushed motors, especially with more load on them than just pushing air. They're probably no more than about 1.5A peak rated and in use, probably 1A or less which is pretty trivial compared to powering a computer and gaming video card, but yes that could overwhelm a fan controller not designed for that much current. It wouldn't trip OCP on the PSU.
*no air planes were harmed in the making of this video* ✈️
Experiments like this are stupid, impractical and a waste of everyones time.
I love it, more of this sort of thing please.
These are server radiators, and Alphacool also offers water-cooled server chassis that you can install these in. In my homelab, I have two 2U chassis, which significantly help in reducing noise compared to air-cooled servers.
A dB measurement would have been fascinating. Still stunning to see such a tiny solution is workable for those niche situations.
Nice and quiet, that's all you need to know 😂
From the time I've spent in data centers you have to raise your voice when you are around those little fans at full tilt, they are so loud!
bro, ALL WE WANT TO SEE is weird stupid things like that, we don't care about some motherboard reviews xD
hahaha
I feel like when Roman talks about something, that means it’s worth considering in terms of technology. Not like he’s a puff piece sort of guy, right? Silly weird artifacts nonwithstanding, he’s a real source of good information
Have to agree... While I deeply enjoy those deep technical reviews where Der Bauer shows his dedication to getting all the science right....
NEVER stop with these kind of fun things... Just to prove people wrong ;)
It's almost like those radial fan blowers we used to have on GPU's.... were not the most stupid idea ever!
@@FreakyAngelus that's what I kinda meant, yeah, those technical videos are awesome, but kinda everybody is doing them, on the other hand, those "experiments" are way more entertaining, you can't find many channels doing it on youtube, old LTT was perfect place for it but they stopped them, everybody is like "I'm not cOmFoRtAbLe WiTh 4090 RuNnInG oN 360 rad" meanwhile derbauer is like "hello, in this video we will try to cool 4090 with 40mmx2 radiator" xD
💀
0:20 There really is no actual answer to that. It depends on how cool you want the GPU to be. It could range anywhere from 'just cool enough to work without dying' to 'The coldest the GPU could be allowing for the laws of thermodynamics on planet Earth'. There is a LOT of room there.
At some point, the radiator would melt, so there's your actual maximum. I would imagine the liquid inside would boil and blow the radiator up before that. Now i want to see a radiator pushed to literal failure. Thanks.
I love these kind of "mad scientist" videos. They are very entertaining! Thanks Roman!
Delta fans always bring a smile on my face with how delightfully absurd they are. I _swear_ Extreme-G used tiny fans like that for their bike engine sound, with only minimal editing required.
9:34 HAHAH! the fans are so damn loud and deafening, he calls the 4090 a 5090. (or maybe he's just excited)
he's probably already testing the 5090, so hence the confusion
@@simocity99 idk about that, the cards have not even been announced. Once they are announced thez will be sent out to reviewers. Edit: “they”
There's so much news about the TDP and TGP of the 5090 it's understandable if it's a Freudian slip.
Those fans are, but I'm getting used to them slowly, they just call "servers" so much!
There is some weird greek letter based math that can calculate a decent prediction of inlet/outlet temp based on the thermal conductivity and specific heat of the aluminum, coolant, area of the radiator fins, and the air flow over them. IIRC, there is a counter intuitive relationship with air dwell time in contact with the heatsink, air can move too fast past a heat sink to pick up as much energy as it could. That being said....when you're cramming cubic meters of air across 80x40mm of zigzag metal, that relationship falls to the good ole American proverb "effectiveness by volume". Usually its used by artillery, bit the sentiment is the same 😆
Sidenote: I was just looking at these kinds of fans for a side project unrelated to PC hardware and cooling. So big thumbs up for incidentally covering something I wanted to see a live demo of! 😎👍
Delta-T?
ϵ or σ? Emissivity and Stefan-Boltzmann constant.
Or ΔT
Lambda
A trick I use when filling the loop is to use boiling hot distilled water, heated in the microwave in a pyrex beaker to avoid contamination. It gets the air out quickly as the air volume shrinks massively when cooled, plus the water is now sterile so as long as the loop remains sealed it shouldn't need a biocide additive! No guarantees but it works amazingly well for me! I always do a double flush of a new loop as well, to be sure there is no crap in it.
The PWM signal should be used to directly control the fan instead of adjusting the voltage and current from the fan's power input.
Proper way to attribute heat dissipation properties is so-called "thermal resistance". It can make an abstraction of elements in the cooling loop, starting with the heat source and all the way to the fans.
That's a linear approximation for non-linear heat transfer, and for a radiator it also depends on water flow (and air flow, ofcourse).
@@-szega your point? I only pointed towards Navier-Stokes equation elements where all those resistances can be presented as a sum
Impeller and pressure is very important. A good impeller design will prevent cavitation at higher flow speeds. If that checks out, you just need a good fan setup to get rid of that heat quickly.
weird to see a fan controller overheat. i think this fan controller did not use PWM signal but an adjustable buck converter?
Has to be
Going push-pull with counter-rotating fans on a 12 FPI radiator is so hilariously overkill, you’ve got enough pressure to lift a car. Need a ridiculously dense radiator.
I wonder how many of those radiators you could stack together before those fans even noticed. We must construct the cooling plank.
Actually, there is a smaller double 40mm radiator with barbs permanently attached that I bought almost 15 years ago. It is only 25mms thick. I still have it and never used it, for obvious reasons :D.
it'd be perfect to cool a single SSD.
@fermitupoupon1754 Yes, let's put a full loop together to just cool an SSD, or better yet, use it in an already built loop and cause massive flow restrictions :D.
@@Hostilenemy you're right, better put 2 DDCs in series to cool that SSD. That should be sufficient to overcome the restrictions in the loop.
You can usually guess if a push/pull will make a large change. If it's a thinner rad or has wide fin spacing, there's not much resistance at all and adding extra fans won't produce a considerable change, if any at all. If it's a beefier rad with a thick core or tight fin spacing, push/pull can have an impact.
Roman is ready to send that radiator to supersonic speeds with those fans! Very interesting tests!
6:02 looks/sounds like RC F-15C with twin 90mm EDFs at full crank...
As a Delta Electronics fangirl(and fan nerd), I loved this hahaha
Now you only need to put the fan controller into the airflow and it won't overheat :D
I've had a Mobo that did do 3A each on 2 fan connectors, I ran 2.7A fans on it.
The magic of the pwm on fans is that you can just pipe a trace directly from the 12V connector, the pwm itself is done on the fan, so the controller shouldn't get warm unless it's actually doing dc speed control.
This is so ridiculous. Love it!
These 40mm rads are very interesting units. I first came across them when an ex-customer involved in High Frequency Trading and crypto in the early days requested for my company to procure the servers. I believe it was an ICC Vega "server" where they had a Core i9 overclocked to 5GHz (you want high single-threaded performance for HFT to validate and push trades faster rather than high core counts) and used a triple 40mm rad to handle the heat.
The IT requirements in that part of the finance industry is rather interesting. They'll ditch any form of cybersecurity necessary just to get lower latencies - no firewalls or managed switches (for VLAN segmentation) because they add latencies. There are even specialized unmanaged switches for this industry that are designed to purely focus on ultra low latency switching
This was an awesome video, I love taking things to the extreme just to find out where physics limit us. Thanks Roman!
That was a really interesting experiment! One probably wouldn't expect a radiator that small to work with that much wattage.
It is far too loud outside of a server room, but it does work.
I had a similar discovery with a cheap fan controller. I bought one that has 4 rheostats and support 8 5v fans. I bought a 4U 24 bay case and the air flow for the drives was lacking, so I put better (and quieter) fans in there, but cranked them all up to 100%. While I was messing around with other cables it moved once and hit the side of the case. Everything shut off. To my horror, it turns out that some of the contacts/pins from the factory weren't grounded or insulated in any way and it shorted out when it touched the case! So I coated anything metal in superglue and that fixed that issue. Then I noticed that it was putting off A LOT of heat, I touched one of the 4 ICs and burned myself like you did and was like "holy shit that's hot!". Luckily I had a bunch of small heatsinks that I purchased for a Pi 3 or 4 long ago and they fit perfectly! 1 heatsink on each IC definitely helped drop the temps down to reasonable, but still warm/hot to the touch, the moderate airflow in the case also helped cool them down. We always forget how hot tiny things get when we force a lot of power through them hahaha
This is awesome, impractical, and exactly why I love this channel. 👍👍👍
Considering the hi-perf fans that rad will use, I am absolutely shocked at the sparse tube/fin density. I would expect with such powerful fans that you could barely see light through the stack.
When Roman first powered those fans on I was reminded of the time my buddy had his switch go out at his house and as a joke I brought over an old Cisco 3750X. When I powered it on and it started screaming during the fan(s) power cycle test he just about lost it. I said "what, its fine, it works great." People have no idea how loud a datacenter can be.
Some of the HPE servers at my workplace have fans that pull over 40w. IIRC the doublestack fans pull 60w... And theres 6 of them in a server. The noise and air movement is something to behold
12:05 I mean you could use a splitter cable, you don't need to power the fans with the fancontroller you could just get the PWM signal (if they are even PWM fans) and power the fans themselfs straight from the psu.
I love these tiny radiators. Mine are a bit larger, designed for 80mm fans in 2U chassis instead of these 1U rads, but I have a grand total of 2x80mm cooling a 14900K and 7900XTX. Noise isn't a concern, but size was for a custom case.
Those 40mm fans are absolute beasts, not the least bit surprising some cheap un-heatsinked controller wants to cry when trying to power them.
The reason the clearance doesn't fit a socket head bolt, is because it's intended that you use hex head bolts or install the nut on that side. The fan body is then able to act as one side of a hex socket when tightening it down. In extremely cramped situations you can simply push the hex head against the fan with your thumb, if you can't get a spanner into that spot.
This reminds me of the time I used a Vantec Tornado on a Swiftech MCX4000 cooler without a fan controller. Thanks Roman for the flashback with that fan noise
Please use hearing protection around those fans. The sound they make is more than loud. It is not bandlimited. It contains energy at extreme frequencies. The energy in a wave is equal to its amplitude times the square of its frequency. Loud sounds will make your ears ring, high sounds will make your ears ring, loud high sounds will destroy your hearing!
Good luck on getting your part. I ordered my parts from watercool a couple weeks ago last year and my parts we to delivered on the first of the month and they haven’t even shipped yet.
just at the beginning of the video i know there is limitation of how much a powerful fan can extract as heat from a radiator, the limit is the conductivity between the water inside and the copper of the radiator. it takes time to move the heat and spread through the surface of that radiator, no matter how fast the fans are. you can increase that speed by just increasing the surface.
the same happens with the am5 IHS small and thick, creating a "jam" in the heat transfer.
As soon as I heard delta I knew where this was going. I was not disappointed.
Keep up the antics.
Having some stators on the back of the front fans might help shove more air through the heatsink by redirecting the swirling coming out of the fan into the fins and eliminating the relatively dead spot directly behind the fan hub.
I love experiments like this! They're wacky, and maybe aren't directly applicable, but hey, now I know that it's possible to cool 300w with minimal space in a server chassis if I ever needed to.
2:18 finally somebody says it
the standard curves in an AIO are set sooo agressive often that it isn't really much more quite, but if you allow the water temp to get hotter you can mentain said tempreture easier because the same radiator will dissipate much more energy.
Ofc it will not cool the cpu to the lowest temp anymore but that is fine.
Turned my rig from being pretty loud whenever I fired up a game thats semi hardware intensive to staying at a comfortable noiselevel just by raising the watertemperture (about 20 degrees I think funnily enough, the stock settings of the corsair radiator are really quite agressive, I made sure to stay in the specs as to not risk damage to the tubing and pump tho)
A fan usually has a 4pin connector, +12V, gnd, pwm and a feedback of the rpm
You could cut the 12v and gnd wires and hook them up directly to the psu and plug the 4pin connector directly into the motherboard
With this you can control any pwm fans without frying your mb
11:48 Put the fan controller in the fans exhaust stream to cool it.
Quality content and precise tests are really cool and all, but sometimes people just need some funny video for goofy questions answered and im glad you're doing just that 😂
Is it just me or those rads could have a denser fin array when matched with those fans?
2 of those probably good enough for a 4090 and 4 for a whole system. That would be a good, why not video xD
Yeah but i guess they assume most people arent mounting 21k rpm fans to it
@@WayStedYou I don't believe so. Such 40mm form factor are used in servers only. And even thinner ones are very good at static pressure so the fins could have easily been twice the density without an issue.
@@WayStedYou radiator that size could be end up in server, so i dont think alphacool didnt think user not slap a mega rpm fans
Will you be making those water blocks available for the 50 series?
Using plenums between the fans and the radiator would have given a much more uniform flow through the radiator matrix, significantly increasing the efficiency.
Cool experiment.
Weird radiator. Fin density and surface area is extremely low. Would have been interesting to see the performance with double or triple the fin density, but no point in wasting time and money on it. No matter how perfectly the fans gets matched to the airflow restriction, the setup would still be way to loud.
Yeah, those delta fans are really optimized for static pressure over just pure airflow. They would've been perfectly happy with much denser fins, or a radiator that was 4* as long. Normally they are used in 1u rack servers to blow air throughout the entire 3 foot long server case and through all sorts of different heatsinks.
If Roman smiling that smile and holding up a Dremel isn't a meme, it ought to be. 😂🤣
This is what RUclips was made for!
At least what I always wanted to be....
I just love such wired ideas that are professionally executed.
I wonder how well noise cancelling headphones would handle the fan noise.
What's funny is these little thickbois can handle just as much of not more than a generic thinner radiator 2-3x their size. I got an EK240 AIO a couple years ago and I was sorely disappointed by the radiator and promptly replaced it with a thick 240 from my old loop and the difference was significant. It's the first and the last time I'll ever buy an AIO.
Good videos as always, happy new year x
This feels like it's sized to enable an internal console liquid cooling setup, while Lian Li back in the day made an external case for the xbox 360 to be able to transfer the guts of such a console into a pc sized case to enable a larger amount of airflow to keep things in control. Granted, that was probably made before knowledge of the target temp behavior the console was doing.
Clearly static pressure isn't an issue, you should stack a few of the radiators in series to get closer to a counter flow heat exchanger. See if gets any closer to being practical
I love the sound of the server fans. Maybe I am used to it working around them
I love when you do these impractical and silly experiments.
I had built one of those that was double size (4x fans). Ripped out 8 Dell Server blade fans like what you have, just without the double stack option. Im glad we're all a bit mad
Not to mention, running hot water through your loop's going to lead to faster evaporation. My GPU loop's lost over an inch of water more than my CPU loop after a year of hard use.
Where does it evaropate to though?
Anyways, would it not be rather question bigger airgaps in rad? Try running 100% pump to see bubbles out.
If you have opening to air in your loop, disregard my comment.
Always love this kind of videos showing these weird cool things about computer hardware. Also question: Is the fan controller controling the voltage to fans or only sand out PWM signal but fans get power from PSU?
it seems that controller board provides the whole power with a buck regulator. Otherwise, there is no reason a simple PWM signal circuit would overheat as such.
Thanks for the chuckle. this was great content.😁
13:13 No noise-canceling headphones?
This is a great PSA as to why you always want to get the biggest fans and radiators you can fit.
Blow the exhaust air flow over the controller board (in a way the exhaust air flow mixes with ambient air, to get a cooler flow)
“We’ll say goodbye to my ears and I’m sorry you have to suff(turns fans way up) OH MY GOD!” 😂
Now THIS is content. I have not played with counter-rotating fans, but the ones we did were powerful on their own. San Ace baby! If you can try their 120mm that runs at 24 volts (9HV1248P1G?)
You should be using PWM control with the fans, not the usual CPU 2-pin or 3-pin (2 pin with RPM) voltage control. That way you would not burn up fan controllers, the fans themselves have control. Or if you did it this way...fan power should be straight +12 and negative, no PWM.
OR - is it a Delta fan thing, my Delta blower has weird PWM control and didnt really work with my homemade controller, unlike the San Ace style (PWM wire: ground for minimum speed, open for maximum speed, PWM to vary the speed) which worked fine. Hope you can test it!
I basically did this when i tried to make the thickest 120mm radiator sandwich ever.
push pull configuration with 2x38mm thick 120mm server fans, The sound levels were tremendous!
Terminated both fans directly to molex since each one drew ~45W
As you probably already figured out, there is a limit to how much radiators can radiate. Even with excessive air throughput. Its just a limitation of the heat passing to the surface area from the water to the fins. IE the lack of surface area is the bottleneck.
If you plugged the fan while it was rotating from the airflow, it probably was generating negative voltage, that got dumped into the fan controller killing it
Great content! Absolutely love stuff like this!
Now to watch you cool that card with the MoRa IV 600 while it sits outside of your window dangling over the street below! Subambient cooling like we used to do in the early 2000s ;)
I would love to see a system that incorporates all of the liquid and air cooling tech in one package.
Even a NOS system that drag cars use would be cool.
Maybe some kind of water to air intercooled computer, like wild cars use to cool inlet air temps.
I know that sounds very tong in cheek, but Im being serious.
Could be interesting to brainstorm!
: ]
brave brave man i appreciate you so much for doing such experiments. i guess you maxed out your psu + fan controller thermal and wattage limits with those tiny fans.
The spec for radiators should be published the same as it is for heatsinks, that is deg C per watt at various airflows.
I think you have to leave some gap between the fans and the radiator, otherwise the area behind the fan hub has no airflow. You were effectively only using like 50% of your radiator if you place the fans on the radiator, maybe add a 3d printed duct in between. This should help tremendously with push/pull too, fans don't suck in air properly when it is right up against a rad.
I know this was a fun project but a slightly more practical test would be to use a single 120mm radiator. I ask this because sometimes a mini-ITX case like the Fractal Ridge or the Terra can fit only a single 120mm radiator and it would be fun to see if it can cool both the CPU or the GPU with that.
well, to do that you'll have to trade room temperature and use a noisy server fan. That's how datacenters manage the thermal dissipation.
You can use an electric domestic under sink water heater maintaining a target temperature to measure the radiators heat dissipation rate by measuring the heaters power consumption over time and the difference in temperature between the water and the room's air.
I'm currently doing this. For a 350mm x 120mm radiator with three cheap fans I'm getting 22 W/K max, and 960 W max. That's at a dT of 40°C - 45°C. Meaning that if the room is 20°C then the water would be 60°C to 65°C.
You can do this the characterize the performance of any radiator and fan combination. You could use this to objectively compare PC radiator performance. W/k/CFM is possibly a good metric? The higher the better.
This metric would also enable you to know before you build a PC what the water temperature would be for the water loop given the system component wattage. Also given the fans cfm vs db(A) you could estimate how loud it would need to be to maintain your target operating temperature, OR what temp the system would be at a given loudness.
Nince idea Roman. I would have used the Aquacomputer fan controllers, like the aquaero with the water cooling backplate for 30W per channel from the built in fan controllers( I have the air cooled backplate). It will also expand with their other single 30W adaptors. I use one channel for my D5 pump and the other 3 as fan controllers. Six fans on my MORA 360 and the other six are run off the pump's fan controller. I've got temp sensors on every connection in my loop. Aquasuite is a good app but a memory hog, that still requires HWINFO64 for full functionality. My 1st one from 2017 is still on the job. I expect long lived devices. It's not cheap but it's programmable with a mono lcd screen that has past the test of time. It is still very readable across the entire screen with no burn in. Although from day one I've only used the lowest brightness settings for everyday use. Only when I access the controller directly via remote or the touch buttons does it light to full brightness, no burnt out pixels but I do have a little pixel dimming on low brightness, not noticeable unless you stare at the screen for the full cycle of displays. I've got a second one that I was going to use on the MORA 360 as a stand alone but I've got extensions from the one in my pc. So it sits as a "Hot Spare". Do you really need a color lcd screen that burns out in a year or two or a oled that may suffer burn in? I can see why after 7 years of constant use (it displays time and date when the pc is off) they made their choice of display tech to use.
I honestly like these DIY videos, waiting for the motorcycle rad waterloop edition
How is 80 degree celsius water ok? Almost all PC watercooling flexible tubing starts breaking down and leaching plastic around 40 degrees celsius.
This kind of experimental testing is exactly what we want to see.
Maybe, 2 fans in a "pull" config would have worked the best. So that the extra few watts of heat from the fans themselves are not being dumped into the radiator.
It makes me wonder if the heat sinking performance of the tiny rad under burst loads could be improved if the plumbing inside were zig-zagged, increasing the surface area by lengthening so it is an endless chicane from start-to-finish.
A lot of rackmount servers have these fans and run up to 6 of them on the motherboard directly.
But they are server motherboards that has a lot sturdier circuitry for the fan connectors. Server fans can be pretty brutal and the motherboards are certainly not designed the same as normal motherboards.
It's like having a jet engine reving up near you. Not even server coolers make that noise, Jesus
Maybe not in stable operation, but those sorts of fans are used in 1U servers all the time, and if that server was cranking dual 300W CPUs and/or a power hungry GPU to 100%, you would hear them spin up. Just very rarely would they hit 100% speed because there’s probably a bank of 6 to 8 of them at least (plus maybe one more in each PSU).
What about placing the radiator and the fan controler out of the window?