[SOLVED] Ryzen 9 Wraith Prism Fan Stuttering with new GPU

May 22, 2022
3
0
10
I have recently installed a new GPU into my computer and the CPU fan has started turning off for a moment after about 2 seconds, and then turning on again for another 2 (and repeat).

Background

About a year ago I built my first computer, after using 2nd-hand ex-college computers for a few years. I don't game, but do CPU-intensive programming, so I mostly focused on specs that would be helpful, and transferred an old GT 630 OEM out of the previous computer (whose PSU had failed) as a "Good enough for now", with the intention of upgrading it later. After a while the Linux drivers for it started crashing often enough (I know, crashing at all is bad, but GPU prices were sky high and I had put up with worse) that I've just bought a new Radeon RX 6600 (a bit much probably).

Problem

Right after I installed the new GPU, the CPU fan has started to stutter. It will run fine (but a little fast) during start-up, but by the time it reaches the login screen, it will start running for about 2 seconds, and cut out and slow down, and then start up again. (I can hear the fan cut out and the RGB colours (not my idea, but the CPU fan came with them so plugged them in) flicker off and restart their colour cycle).

My first response was to boot to BIOS, and MSI provides decent fan monitoring here. However, the first time I booted through BIOS, it stopped happening, so I exited and continued through normal boot, and the problem was gone (yay!) but the next morning (today), the problems returned, and now even happen during BIOS. This does mean I can monitor them, though, and sure enough the fan speed shown starts to drop and returns to normal repeatedly.

I tried looking at their connections to the motherboard, they seem pretty secure, and the fact they still work mostly, and their powering off is so regular, doesn't lead me to believe there is a problem with the connections.

Interestingly, if I force them on [FULL SPEED] in BIOS they no longer cut out, so I tried upping their normal-temperature voltage. The default is 2.4V, but upping them to 3 or 4V helps for a while and then the problem returns. 5V however, seems to be sufficient to avoid this problem, so for now, I have the fans set to a profile where their minimum is 5V and the cutting out behaviour has thankfully stopped for now.

However, I am quite worried about this, and have no idea what's going on:

  • Why might the fan be cutting out at all?
  • Why was 2.4V okay with the old GPU, and now not okay? (Is it because the power draw from the PSU is higher and so it's otherwise not getting the right current, so needs a higher voltage to get sufficient current)
  • Is running it consistently at 5V okay for the fan? (It's manageably noisy, not painfully so)
  • Is it possible I damaged something during the installation of the GPU? (I got a little forceful with it when the GPU wasn't working until I realised it needed its own direct line from the PSU)
Note, the GPU has its own 8-pin power cable from the PSU which it draws power from. radeontop shows reasonable GPU behaviour, but sensors never shows the GPU's fan on in normal usage (probably expected?).

Thank you so much for taking the time to read this. Sorry if this is a little long, I don't know what is relevant :-( Thanks in advance for any help!

Hardware specs


Motherboard: MSI B450 -A PRO MAX
CPU: Ryzen 9 3900X
Fan: As provided, Wraith PRISM
GPU: MSI Mech 2x RADEON RX 6600
Power Supply: Corsair RM650
Case: Fractal Meshify C
Fans: 2 pre-installed + 1 added
OS (if it would help if I ran any tools): Linux/Fedora 36
 
Last edited:
Solution
You are lucky things were working with no problem before this, so you did not realize what you had set up was unusual. A Wraith Prism standard CPU cooler has a fan that uses a 12 VDC power supply. That is, it turns at full speed given a 12 VDC supply, and slower if the voltage is reduced. I expect that the timing of this - right after a GPU change - is entirely coincidental.


You have not told us what mobo you have. This impacts HOW the speed of that fan is altered. For the older 3-pin fan design, the only method was to reduce the VOLTAGE supplied to the fan (via Pin #2) to slow it down. For the newer 4-pin PWM style fan (I believe this applies to the Wraith RGB fan) that voltage is always 12 VDC, but the header supplies a new control...

Paperdoc

Polypheme
Ambassador
You are lucky things were working with no problem before this, so you did not realize what you had set up was unusual. A Wraith Prism standard CPU cooler has a fan that uses a 12 VDC power supply. That is, it turns at full speed given a 12 VDC supply, and slower if the voltage is reduced. I expect that the timing of this - right after a GPU change - is entirely coincidental.


You have not told us what mobo you have. This impacts HOW the speed of that fan is altered. For the older 3-pin fan design, the only method was to reduce the VOLTAGE supplied to the fan (via Pin #2) to slow it down. For the newer 4-pin PWM style fan (I believe this applies to the Wraith RGB fan) that voltage is always 12 VDC, but the header supplies a new control signal (PWM) on Pin #4. The fan has a chip that uses that signal to modify the flow of current from that supply line though the motor windings to achieve speed control. On modern mobos virtually all fan headers have 4 pins, BUT in BIOS Setup for each header you can choose whether it operates in Voltage Control Mode (aka DC Mode), or in the new PWM Mode. This allows a user to adjust for which type of fan is installed. Another factor: the design of the new PWM fan type includes a backwards compatibility feature so that if it is connected to a header using the older Voltage Control Mode, it receives NO PWM signal and cannot use that to modify the power source, BUT that power IS a varying Voltage so the speed of this fan IS controlled by that signal system, even though it is not technically optimal. So, what mobo you have, whether or not it has these two MODE options on the CPU_FAN header, and how you have those options set is pertinent.


What you have told us is that you are attempting to control this cooler's fan speed by altering a Voltage setting in BIOS Setup. So to start, I'll talk about how Voltage is used for control of am older 3-pin fan type. It turns full speed at 12 VDC, and slower at lower Voltage. The "rule of thumb" (such rules never are universal)is that a supply of less than 5 VDC may lead to the fan stalling. When that happens it cannot re-start itself but still keeps on drawing a small current. The only way to get it to start is to increase the Voltage significantly (often to more than 7 VDC) to start it. In almost all computer fan headers the unit monitors the speed signal being returned to it from the fan for failure (no speed signal). If it detects no speed, it attempts a re-start by boosting the Voltage up to the max 12 VDC. If that fails to generate speed signal you get an on-screen warning that the fan has failed. If it succeeds, however, the header then returns the fan's Voltage to whatever it was set to previously. If that actually was the cause of the stalling, it happens again - cycle repeats continuously. THAT is what you seem to describe. The only difference here us that your failure happens around 2 VDC, not close to 5 VDC.

IF your fan actually is of the new 4-pin PWM style,this still can happen in a couple of ways. IF your header is using the older Voltage Control Mode, it is exactly the same because this fan type can act like a 3-pin fan. If the header is using the new PWM Mode, it still can happen if the PWM control signal calls for a fan speed too slow. In the BIOS Setup screen where you configure the header, SOME screens use words like "Voltage" instead of "% On" to describe the real parameter, which is a % of max power to the fan. So your might "see" the fan setting in terms of "Voltage" even in PWM Mode.

It is not clear in your post how you are using the header controls. The common default choice is often a "Standard" or "Normal" Profile in which the header automatically alters the CPU cooler's fan speed according to the temperature measured by a sensor built into the CPU chip. The common other options, though, include a Turbo (fixed max speed), Quiet (fixed or user-settable slow speed for low noise and limited cooling), Manual (fixed manually-set speed), or Custom (automatic control but using a user-specified "fan curve" of fan speed versus measured temperature). Many mobos have a parameter for you to set a minimum speed (in terms of RPM or Voltage) for the lowest measured temperatures. Others do not, but if you use the Custom option to set your own "fan curve" you get to make that specification.

So, in some manner you have chosen to fix the slowest speed that fan can run. As it happens you have made a setting much slower than most would advise, but it worked reliably anyway. Until now, that is - VERY likely just because the fan is older and slightly worn (perfectly normal), and hence has slightly higher friction than when new. So its minimum power setting is now higher that it was, and it CAN stall at settings that used to be OK.

Solution: simply set the minimum speed (or Voltage) higher until the stalling stops. You already have found that 5 VDC does this, and at some lower Voltage settings it can still stall. Consider also (IF you have not done this already) using the automatic speed control functions rather than a fixed speed. This will run at the lowest speed required for the temperature measured inside your CPU chip, and increase speed (and cooling rate) when the workload (and heat generation) increases. Also check: IF you have a FOUR-pin (PWM type) fan and the option for a how the CPU_FAN header MODE is set, it should be PWM Mode for a 4-pin fan.
 
Solution
May 22, 2022
3
0
10
You are lucky things were working with no problem before this, so you did not realize what you had set up was unusual. A Wraith Prism standard CPU cooler has a fan that uses a 12 VDC power supply. That is, it turns at full speed given a 12 VDC supply, and slower if the voltage is reduced. I expect that the timing of this - right after a GPU change - is entirely coincidental.


You have not told us what mobo you have. This impacts HOW the speed of that fan is altered. For the older 3-pin fan design, the only method was to reduce the VOLTAGE supplied to the fan (via Pin #2) to slow it down. For the newer 4-pin PWM style fan (I believe this applies to the Wraith RGB fan) that voltage is always 12 VDC, but the header supplies a new control signal (PWM) on Pin #4. The fan has a chip that uses that signal to modify the flow of current from that supply line though the motor windings to achieve speed control. On modern mobos virtually all fan headers have 4 pins, BUT in BIOS Setup for each header you can choose whether it operates in Voltage Control Mode (aka DC Mode), or in the new PWM Mode. This allows a user to adjust for which type of fan is installed. Another factor: the design of the new PWM fan type includes a backwards compatibility feature so that if it is connected to a header using the older Voltage Control Mode, it receives NO PWM signal and cannot use that to modify the power source, BUT that power IS a varying Voltage so the speed of this fan IS controlled by that signal system, even though it is not technically optimal. So, what mobo you have, whether or not it has these two MODE options on the CPU_FAN header, and how you have those options set is pertinent.


What you have told us is that you are attempting to control this cooler's fan speed by altering a Voltage setting in BIOS Setup. So to start, I'll talk about how Voltage is used for control of am older 3-pin fan type. It turns full speed at 12 VDC, and slower at lower Voltage. The "rule of thumb" (such rules never are universal)is that a supply of less than 5 VDC may lead to the fan stalling. When that happens it cannot re-start itself but still keeps on drawing a small current. The only way to get it to start is to increase the Voltage significantly (often to more than 7 VDC) to start it. In almost all computer fan headers the unit monitors the speed signal being returned to it from the fan for failure (no speed signal). If it detects no speed, it attempts a re-start by boosting the Voltage up to the max 12 VDC. If that fails to generate speed signal you get an on-screen warning that the fan has failed. If it succeeds, however, the header then returns the fan's Voltage to whatever it was set to previously. If that actually was the cause of the stalling, it happens again - cycle repeats continuously. THAT is what you seem to describe. The only difference here us that your failure happens around 2 VDC, not close to 5 VDC.

IF your fan actually is of the new 4-pin PWM style,this still can happen in a couple of ways. IF your header is using the older Voltage Control Mode, it is exactly the same because this fan type can act like a 3-pin fan. If the header is using the new PWM Mode, it still can happen if the PWM control signal calls for a fan speed too slow. In the BIOS Setup screen where you configure the header, SOME screens use words like "Voltage" instead of "% On" to describe the real parameter, which is a % of max power to the fan. So your might "see" the fan setting in terms of "Voltage" even in PWM Mode.

It is not clear in your post how you are using the header controls. The common default choice is often a "Standard" or "Normal" Profile in which the header automatically alters the CPU cooler's fan speed according to the temperature measured by a sensor built into the CPU chip. The common other options, though, include a Turbo (fixed max speed), Quiet (fixed or user-settable slow speed for low noise and limited cooling), Manual (fixed manually-set speed), or Custom (automatic control but using a user-specified "fan curve" of fan speed versus measured temperature). Many mobos have a parameter for you to set a minimum speed (in terms of RPM or Voltage) for the lowest measured temperatures. Others do not, but if you use the Custom option to set your own "fan curve" you get to make that specification.

So, in some manner you have chosen to fix the slowest speed that fan can run. As it happens you have made a setting much slower than most would advise, but it worked reliably anyway. Until now, that is - VERY likely just because the fan is older and slightly worn (perfectly normal), and hence has slightly higher friction than when new. So its minimum power setting is now higher that it was, and it CAN stall at settings that used to be OK.

Solution: simply set the minimum speed (or Voltage) higher until the stalling stops. You already have found that 5 VDC does this, and at some lower Voltage settings it can still stall. Consider also (IF you have not done this already) using the automatic speed control functions rather than a fixed speed. This will run at the lowest speed required for the temperature measured inside your CPU chip, and increase speed (and cooling rate) when the workload (and heat generation) increases. Also check: IF you have a FOUR-pin (PWM type) fan and the option for a how the CPU_FAN header MODE is set, it should be PWM Mode for a 4-pin fan.

Thank you so much! (Honestly this has been a lifesaver) This helped me diagnose the problem, but it was not quite as is stated in the answer. If anyone else has a similar problem (especially with this motherboard) here's what I was able to deduce. I also mentioned that it was a MSI B450 -A PRO MAX at the bottom under hardware specs, so I'm a little confused what more information you were looking for.

  • It was set to automatic settings (the motherboard has options for DC, PWM, and Autoselect, alongside speed profile settings, which were also automatic; I never changed this until diagnosing), and if I recall it had been using PWM under auto, but it didn't occur to me that PWM wouldn't display voltages (as I just knew that was the newer option, not what it really meant) when I ran into this problem.
  • It seems the auto-settings switched to DC after the installation, potentially the motherboard was affected by the installation of the new hardware? I don't know if that suggestion is ridiculous.
 
Last edited:

Paperdoc

Polypheme
Ambassador
Sorry, in writing my response earlier, I completely missed the proper mobo reference. Thanks for that point, now I can see what it says. There is no reason why the new installation should have made the change to DC Mode for you. But weird things do happen, and now that you've caught that you can set it to PWM Mode - maybe have done that already. Personally I don't trust Auto setting Mode.

If you want to be sure you are using the manufacturer's default "fan curve" and automatic controls, I suggest this IN CASE there are other settings in place. Go to the fan header configuration screen - manual p. 59 and select the CPU_FAN header. At bottom click on the All Set Default button. Then ensure at upper left the PWM Mode is selected. Use Esc to return to Main Menu, then F10 to get to Exit menu (p. 52) and choose Save Changes and Reboot.

Good luck, and thanks for Best Solution!
 
May 22, 2022
3
0
10
Sorry, in writing my response earlier, I completely missed the proper mobo reference. Thanks for that point, now I can see what it says. There is no reason why the new installation should have made the change to DC Mode for you. But weird things do happen, and now that you've caught that you can set it to PWM Mode - maybe have done that already. Personally I don't trust Auto setting Mode.

If you want to be sure you are using the manufacturer's default "fan curve" and automatic controls, I suggest this IN CASE there are other settings in place. Go to the fan header configuration screen - manual p. 59 and select the CPU_FAN header. At bottom click on the All Set Default button. Then ensure at upper left the PWM Mode is selected. Use Esc to return to Main Menu, then F10 to get to Exit menu (p. 52) and choose Save Changes and Reboot.

Good luck, and thanks for Best Solution!

Sorry, forgot to say, I set it to force to PWM, and now that I didn't need to have it bottom out at 5V, I reset the speed/temperature profile to default. All is well! Thanks again so much, you were very quick (and extremely helpful) to identify the exact physical problem!