[SOLVED] How to best utilize 3 fan headers for fans and AIO pump

Feb 7, 2021
8
0
10
0
Greetings! My mITX motherboard (ASUS Z170i Pro Gaming) has three fan headers only: CPU_FAN, CHA_1 & CHA_2. I plan on using a Y-Splitter to connect both Kraken X53 radiator fans to the CPU_FAN header and utilizing PWM. For the remaining two fan headers (CHA_1 & CHA_2), I will be connecting three case fans and the AIO pump. Two of these case fans are 120mm 3-pin , and one is a 92mm PWM. The pump is a 3-pin connector.

Specifically, I am uncertain as to whether it is okay to connect the AIO pump to a Y-Splitter, as opposed to connecting it directly into the motherboard header. As the pump has a 3-pin connector, does this mean it will run at 100% speed even if connected to a PWM Y-Splitter along with a PWM fan that is using a custom speed? I understand that the pump should run at 100% constantly.

The Noctua 92mm is blowing directly onto the GPU, and therefore I would like for this fan to be PWM performing accoring to GPU temp, if possible.

Your recommendations and support will be very much appreciated.
 

Paperdoc

Glorious
Ambassador
And the last shall come first....
The problem with your plan is that there never has been any way for a GRAPHICS CARD to send to the mobo a GPU chip temperature, so that the MOBO can use that to control a fan aimed at the card. So that Noctua fan aimed that way can have its speed controlled a couple of ways.
(a) Connect it as you plan, to a CHA_FAN header, and configure that header to use PWM Mode of control and the MB (motherboard) temperature sensor. That is NOT the temperature inside the GPU chip by any means, but it does reflect the air temperature in the case.
(b) Connect the same, but set that header to use the CPU chip temperature sensor, betting the heat generation due to workload changes in the CPU chip will be VERY similar to heat generation and workload in the GPU chip.
(c) Forget control of this fan. Just set it to run full speed all the time, by powering it directly from the PSU with an adapter cable. This does NOT make any use of its PWM capability. It may over-supply air to the graphics card, but that does no harm.
(d) Connect as in (c) above, BUT look what came with the Noctua fan. SOME of their fans come with little items they call "LNA" for Low Noise Adapter. This is used when your fan is connected to a full 12 VDC power supply and has no speed control. It simply inserts into the fan power supply connector and is a resistor to reduce the fan speed (and airflow, and noise).
You have not told us details of your graphics card. Most actually have their own cooling fan built in, and most of those do their own fan speed control based on their GPU chip's internal temperature. They just cannot send those signals back out to the mobo for use there.

Your thoughts about powering the pump and fans of the Kraken X53 system are on the right path. The PUMP part is wired just like a 3-pin fan. Any 3-pin fan (and this pump) plugged into a 4-pin mobo header that is using PWM Mode for control (and your mobo's CPU_FAN header DOES do it that way) will always operate at full speed all the time. That IS what your pump is supposed to do. So I recommend that you get a simple 3-output 4-pin fan SPLITTER like this

https://www.amazon.com/Splitter-Computer-Extension-Converter-TeamProfitcom/dp/B07F8LV1BY/ref=sr_1_4?crid=1B5YJ1RA6HC0W&dchild=1&keywords=fan+splitter+4+pin&qid=1612756562&sprefix=fan+splitter,aps,183&sr=8-4

and plug that into your CPU_FAN header. Look closely at the three output arms. Only ONE of them has all four of its pins, and that is the only one that will send back to the CPU_FAN header the speed of its device. So plug into that one the 3-hole cable from your PUMP, then plug the two rad fans into the other outputs. The CPU_FAN header has an important second function: it monitors its device's speed signal for FAILURE. If that happens it will alert you immediately, and SOME mobos may take quick action to shut everything down even before the CPU internal temperature sensor shows high temps. In an AIO system, the critical component to monitor for failure is the pump. As long as it runs, even with partial rad fan failure, you get CPU cooling.

If you connect this way, the PUMP will run full speed all the time as designed and its signal will be monitored for failure. The RAD FANS both WILL be controlled for speed by the PWM Mode of the header, but that header will never "see" nor display those fans' speeds. From time to time YOU should verify they both still work.

So that leaves your two front 3-pin case fans to connect. If you are already using one CHA_FAN header for the Noctua aimed at the graphics card, connect them both to the other header using a Splitter. (Yes, you can use a 4-pin Splitter for this.) If not, you can connect each to its own separate header. In either case, since these are 3-pin fans, ensure that they are configured to use DC Mode and to use the MB temperature sensor.
 

Paperdoc

Glorious
Ambassador
And the last shall come first....
The problem with your plan is that there never has been any way for a GRAPHICS CARD to send to the mobo a GPU chip temperature, so that the MOBO can use that to control a fan aimed at the card. So that Noctua fan aimed that way can have its speed controlled a couple of ways.
(a) Connect it as you plan, to a CHA_FAN header, and configure that header to use PWM Mode of control and the MB (motherboard) temperature sensor. That is NOT the temperature inside the GPU chip by any means, but it does reflect the air temperature in the case.
(b) Connect the same, but set that header to use the CPU chip temperature sensor, betting the heat generation due to workload changes in the CPU chip will be VERY similar to heat generation and workload in the GPU chip.
(c) Forget control of this fan. Just set it to run full speed all the time, by powering it directly from the PSU with an adapter cable. This does NOT make any use of its PWM capability. It may over-supply air to the graphics card, but that does no harm.
(d) Connect as in (c) above, BUT look what came with the Noctua fan. SOME of their fans come with little items they call "LNA" for Low Noise Adapter. This is used when your fan is connected to a full 12 VDC power supply and has no speed control. It simply inserts into the fan power supply connector and is a resistor to reduce the fan speed (and airflow, and noise).
You have not told us details of your graphics card. Most actually have their own cooling fan built in, and most of those do their own fan speed control based on their GPU chip's internal temperature. They just cannot send those signals back out to the mobo for use there.

Your thoughts about powering the pump and fans of the Kraken X53 system are on the right path. The PUMP part is wired just like a 3-pin fan. Any 3-pin fan (and this pump) plugged into a 4-pin mobo header that is using PWM Mode for control (and your mobo's CPU_FAN header DOES do it that way) will always operate at full speed all the time. That IS what your pump is supposed to do. So I recommend that you get a simple 3-output 4-pin fan SPLITTER like this

https://www.amazon.com/Splitter-Computer-Extension-Converter-TeamProfitcom/dp/B07F8LV1BY/ref=sr_1_4?crid=1B5YJ1RA6HC0W&dchild=1&keywords=fan+splitter+4+pin&qid=1612756562&sprefix=fan+splitter,aps,183&sr=8-4

and plug that into your CPU_FAN header. Look closely at the three output arms. Only ONE of them has all four of its pins, and that is the only one that will send back to the CPU_FAN header the speed of its device. So plug into that one the 3-hole cable from your PUMP, then plug the two rad fans into the other outputs. The CPU_FAN header has an important second function: it monitors its device's speed signal for FAILURE. If that happens it will alert you immediately, and SOME mobos may take quick action to shut everything down even before the CPU internal temperature sensor shows high temps. In an AIO system, the critical component to monitor for failure is the pump. As long as it runs, even with partial rad fan failure, you get CPU cooling.

If you connect this way, the PUMP will run full speed all the time as designed and its signal will be monitored for failure. The RAD FANS both WILL be controlled for speed by the PWM Mode of the header, but that header will never "see" nor display those fans' speeds. From time to time YOU should verify they both still work.

So that leaves your two front 3-pin case fans to connect. If you are already using one CHA_FAN header for the Noctua aimed at the graphics card, connect them both to the other header using a Splitter. (Yes, you can use a 4-pin Splitter for this.) If not, you can connect each to its own separate header. In either case, since these are 3-pin fans, ensure that they are configured to use DC Mode and to use the MB temperature sensor.
 
Feb 7, 2021
8
0
10
0
Paperdoc; your reply is greatly appreciated.

The graphics card for this build is an R9 Nano, and chassis an NZXT H210. Due to the very close proximity of the GPU to the PSU shroud in this case, the 92mm Noctua fan is zip-tied under the PSU shroud and blowing directly onto the R9 Nano like in this video:
View: https://youtu.be/XQaAY9vHTKU?t=454
(I'm using an SFX power supply as well, of course).

An additional question that comes from this: With this 92mm fan running at 100% as you've suggested, do you foresee an issue with it blowing air directly into the GPU fan at such close proximity?

The 3-output 4-pin Splitter that you've suggested is conveniently included in-box with the Kraken X53! Is it okay to connect all three components (two 120mm PWM rad fans and the AIO pump) to this single motherboard header? The mobo manual states that the header supports a maximum of 1A fan power, and I believe the total amperage of these three components combined will be under 1A. With that said, does this not cause any strain on the mobo, especially considering that two of the fans are PWM?

Following your recommendation of using the 3-output Splitter, and if also using a Y-Splitter for the two case fans, this frees up the third header for the 92mm Noctua, thus allowing for PWM, if desired. You mentioned the LNA (Low Noice Adapter), which I hadn't heard of and is a great thing to know. Would you recommend running at 100% with the LNA, or using a custom fan speed with PWM?
 

Paperdoc

Glorious
Ambassador
Handy that you have the Splitters you need already. The specs for the Kraken X53 system show a max current load of pump plus both rad fans as 0.94 A, so that is OK on one header.

If you connect the Noctua to the second CHA_FAN header, do NOT use the LNA thing IF you even have it. Just use the header's controls on PWM Mode. You can set it to a fixed speed in BIOS Setup (and change that easily if you decide to), or set it to control speed automatically, perhaps using the CPU temp sensor on the assumption that workloads on both CPU and GPU are correlated. Given the layout of your parts, the temperature of the air inside the case (as measured by the mobo temp sensor) is pretty much irrelevant to graphics card cooling, since its air supply will be almost all from outside air drawn in by that Noctua fan.
 
Feb 7, 2021
8
0
10
0
A quick update—I received the following advice from NZXT and curious to know others' thoughts:

It's generally not recommended to connect both the pump and the fans to the same header. We would recommend having the fans connected to one of the CHA_FAN headers and then run the pump itself off the CPU_FAN. This makes sure that the pump is only supplying TACH information to the motherboard while the fans are safely handled from the CHA_FAN header. Adding all of these to the same header is not a great idea, as the fans will not be able to be controlled properly if chained with the pump sense cable.

Per the advice above, my understanding is that component connected to the 4-pin connector of the 3-way splitter will supply the TACH information. Is it true that the fans will not be able to be controlled properly if chained with the pump to the same header?
 

Karadjgne

Titan
Ambassador
The problem with your plan is that there never has been any way for a GRAPHICS CARD to send to the mobo a GPU chip temperature, so that the MOBO can use that to control a fan aimed at the card.
SpeedFan can do that. Simple as changing the addressing as to which sensor controls which header. Can set all fans to respond to gpu temps if you wish. Can even change a sys_fan header to respond to cpu temp, or the cpu_fan header to respond to motherboard temps.

There's 2 ways to install a basic aio. Nzxt/corsair and pretty much everybody instructs that the pump go to cpu_fan and fans go to sys_fan. That's to cover their posterior since it's assumed that the pump will fail before the fans do. The cpu has a built in safety check, if cpu_fan is not populated or falls below a set threshold, the cpu shuts down/refuses to boot. So if the pump fails, pc is protected. That works great, except the sys_fan sensor is not the cpu temp, but the motherboard temp somewhere else, and rarely ever sees over 40-50°C. So your fans are not controlled by cpu temps. Sucks for gaming.

The other method is what most ppl do, against recommended instructions, and install the aio fans to cpu_fan header and the pump to a sys_fan. Cam controls the pump by the reported tach, Cam controls the fans according to cpu temp. Everything works as it should. When you game, cpu temps climb, so does the fan rpm.

Take your pick, neither way is wrong, one way just works better, the other has a better sense of security.

Cam controls pump speeds by the usb, the tach wire is just the reporting tool. The pump will climb as high as 4000rpm in some series, fans do not. Pwm works on a % scale, not a temp scale, so when a sensor reports a specific temp, Cam fan curves say to move the header pwm pulse to x% , but that may not be what Cam wants the pump to change to. Having 1 tach means only 1 % change, not 2.

There's also other factors, like startup amperage is not running amperage, fans can and do spike (listen to an air conditioning compressor, same thing at start) amperages. So that 0.9A total, isn't, it can be considerably higher, which can burn out a header. Also not all headers are the same. It's generally accepted that a header is 1A at 12v. In reality there's motherboards (cheap ones) that may only max out at 0.9A. So those 2 fans added with the pump is really pushing your luck when the cpu is hitting high loads.
 
Last edited:

Paperdoc

Glorious
Ambassador
I do disagree with what NZXT said. A fan header does NOT need the fan speed signal to be able to control that speed. In fact, it never uses that info for that function. It uses a speed signal for only two things: to display it for you for info and interest (not critical); and to monitor it for fan FAILURE. Furthermore, when you use a Splitter, only ONE fan's speed can be reported to the header, so that failure monitoring can only check ONE fan. YOU still have to do your own checks of the other fan(s). Hence, having NO rad fan speeds fed to the mobo header means only that YOU have to check BOTH of them for failure, not just one. NO difference there! Plus, you can't "see" those speeds if you are curious.

Secondly, the details of how failure monitoring is done differ from one mobo to another, and they never tell you those details in the manuals! But in general, the CPU_FAN header in particular (and likely any AIO_PUMP header if you have one) does much more that just sending you a screen notice when failure is detected (by lack of a speed signal). On some mobos after a relatively short wait, the system may actually shut down your system to prevent CPU overheating without waiting for the CPU's internal temp sensor to report excess temperatures. It also MAY refuse to allow the system to boot if it gets no speed signal immediately on starting. Now, in an AIO system, the most important item to monitor for failure is the PUMP. Without that, no fluid flows to remove heat, so CPU overheating can happen rapidly. On the other hand, if ONE of the fans on the rad fans fails but the pump keeps going, there is SOME heat removal and the rise in internal CPU temperature is slow, so that the normal CPU temperature limits, including both CPU speed throttling and possible shut-down, can act quickly enough to deal with that. So, if you have to choose which component of an AIO system should be monitored very closely with rapid response available in the event of failure, between the PUMP and the RAD FANS, I say the PUMP is the one that should be monitored by the CPU_FAN header. So that is where you should connect the PUMP. IF you are using a Splitter for that and including other items (like rad fans) on that header, then you must ensure that the PUMP is the item whose speed is sent to that header by using the only Splitter output arm with ALL FOUR pins.

No matter which way one chooses that, it is IMPORTANT in the Kraken system (and others that use the mobo's built-in fan control systems) that the RAD FANS be controlled by a header that CAN use the CPU internal temperature sensor for guidance. Your mobo DOES allow you to make that choice in configuring any of its CHA_FAN headers, so for you it IS possible to connect the PUMP to the CPU_FAN header and the RAD FANS to a CHA_FAN header properly configured. Not all mobos have that option.

Karadjne raises a valid point on header load. Most headers can supply up to 1.0 A max current to the connected load. (I have seen a few that are less, and a few that are more.) Several threads earlier have commented that these header's hardware components CAN withstand a load higher than that for a short time, and that is sufficient to allow for start-up surge current with no damage to the header. However, that assurance is NOT guaranteed for any and all mobos, and you will never find any mention of that in a manual. So, for your mobo, OP, I said the max load is spec'd at 0.94 A, and that is acceptable. But maybe not - we cannot really know. If you are concerned about that, then your best option is to put only the PUMP on the CPU_FAN header, and use a Splitter to put the TWO rad fans on a CHA_FAN header configured to use the CPU temp sensor. That then means that your two front case fans will also need a Splitter to connect to the other available CHA_FAN header.
 

Karadjgne

Titan
Ambassador
then your best option is to put only the PUMP on the CPU_FAN header, and use a Splitter to put the TWO rad fans on a CHA_FAN header configured to use the CPU temp sensor. That then means that your two front case fans will also need a Splitter to connect to the other available CHA_FAN header.
That's the best way to do it regardless.
MITX is a tiny box, even the larger optioned cases. CFM is cubic feet per minute and a single 120mm averages about 50cfm. A cubic foot is 1' x 1' x 1' and an mITX isn't close to those dimensions, considerably smaller overall. You'd be looking at a single intake and single exhaust fan effectively replacing the entire volume of air inside the case every 1 second, ±.

That's a ton of air movement. So setting equal fans at same rotation and cfm in pairs maximizes potential. Keep intakes as a pair, exhausts as a pair and pump on the last header.
 
Feb 7, 2021
8
0
10
0
Paperdoc, as always, and Karadjgne, thank you for the informative responses. You guys are beasts.

At this point then, I'll probably opt for the consensus method which is to utilize the CPU_FAN header for the pump only, and use Y-Splitters to connect the remaining four fans (two and two) to CHA_FAN headers. That leaves the additional fan (92mm zip-tied under the PSU shroud) without a header. I'm thinking a good option, per Paperdoc's advice above, is to connect this guy directly to the power supply and use a Low Noise Adapter to preserve its lifespan as much as possible.

However, would it be viable to use a Y-splitter to connect both the pump and this 92mm fan to the CPU_FAN header (ensuring that the pump is occupying the 4-pin connector, of course)? This way, the pump provides TACH info to the mobo for fail-safe, and the 92mm fan can be controlled with PWM, as opposed to just running 100% if connected to the power supply.
 

Paperdoc

Glorious
Ambassador
I suggest a different arrangement. At the core of this is that the rear fan should really be controlled according to the temperature sensor on the mobo, NOT by the CPU chip's internal sensor.

The way you have things allocated right now is:
CPU_FAN header us solely for the PUMP of thre AIO system. It is set to PWM Mode and always uses the CPU temp sensor - that header has no choice on this latter item.

CHA_FAN 1(?) both rad fans via a 2-output Splitter, set for PWM Mode and using the CPU internal temp sensor for guidance, since it is for CPU cooling.

CHA_FAN2 (?) for two 3-pin case ventilation fans via a Splitter, set for Voltage Control Mode (aka DC Mode) and using the mobo temp sensor for guidance.

FIRST: check the ratings of those two 120mm 3-pin case vent fans. I expect their max current rating is less than 0.25 A each, but verify. Then check similarly the rating for that 92 mm rear fan; I expect also it is below 0.25 A.

SO, what I suggest (assuming the total load is under 1.0 A) is that you use a three-output Splitter to connect both of the two 3-pin 120 mm case fans AND the rear 92 mm fan to that second CHA_FAN header. This would put that rear fan under automatic control based on the mobo temp sensor. Now, that is a PWM style fan and Voltage Control Mode is not quite ideal for its control. However, it WILL work that way - that is part of the backwards compatibility features of this design. Among those three fans, it does not matter which of them is plugged into the only Splitter output that has all four pins.
 
Reactions: Chrisaka
Feb 7, 2021
8
0
10
0
Thank you for these points and suggestions.

The 92mm fan, however, is mounted under the PSU shroud (not the rear) bringing air up from the bottom vent of the case. In addition to circulating more air in general, I did this particularly to help cool the GPU which is otherwise mildly choked by its close proximity to the PSU shroud (which is perforated, of course, thank god).

Per my research, the max input current of the fans is a follows:
  • 120mm NZXT fan (AER F case model): 0.18A
  • 92mm Noctua NF-A9 PWM chromax.black: 0.1A
Based on this, it seems minimal risk to burning out the header, even considering start-up spike and fluctuations noted above.

Is there any major drawback to connecting the pump to the CPU_FAN header via Y-Splitter along with the 92mm Noctua (per Karadjgne's advice, ensuring the pump is connected to the TACH wire, of course)?
 

Karadjgne

Titan
Ambassador
Electrically you are fine either way, we'll under amp ratings for any of the headers. The difference will be in response. Assuming decent fan curve settings, with that 92mm on a 3way splitter to case fans, it'll respond the same as the other case fans, spinning up/down with motherboard temps/case temps. If it's split from the pump, it'll use the pumps settings as per Cam. Almost universally, the pump (mostly nzxt uses variable speed pumps, many don't) will be activated by one of 2 settings (which you choose in Cam). It'll respond to coolant temp or cpu temp. Your choice. I prefer coolant temp as coolant takes far longer to saturate and change vs an erratic cpu under loads. Eliminates fans ramping up/down at idle or low loads with windows services startups.

So as long as you are electrically sound, you can do either way, or try both and see which makes better sense as thermals are a variable depending on case choice. See which has a better impact on gpu temps when gaming longer than 20-30 minutes, does that affect noise outputs, etc.

There's very few times ppl can claim absolutely that any setup is wrong, if things work as intended. My case has almost zero defined airflow, I've seen ppl use the rear and top as intakes (BTX especially), my old case I didn't use a rear exhaust at all since the 280mm worth of fans on top created a chimney affect with an aio. The nzxt H500-H510 series benefits considerably by not using intakes at all. The Silverstone Ravens don't have any exhaust fans yet are highest ranked airflow cases. So there's no perfect answer, no absolute yes or no you did it wrong, unless you do something completely insane like have all intakes or all exhausts or mount the tower backwards to case airflow.

You'll figure out what works best for you as to which placement and connection, as long as you stay well within electrical parameters for the header.
 
Reactions: Chrisaka

ASK THE COMMUNITY