Yesterday I’ve put Noctua NF-A12 PWM on my 2080 Super Hybrid FTW3 and today I have noticed PX1 seems to have a bug that comes to surface when trying to use custom fan curve with it. To be specific:
 
If I take fans out of auto and I set value for fan 2 manually to 40% everything is working fine, fan shows to be at 800 RPM (this fan has 2000 RPM at 100%).
 
If fans are on ‘auto’ but no custom fan profiles are used everything is still working fine, if PX1 says fan 2 is supposed to be at 40% it still shows 800 RPM.
 
However, moment I enable ‘fan curve control’ RPMs of fan 2 (Noctua) start oscillating all over the place, from whatever they are supposed to be based on % for that temperature to much, much lower (as low as 0, 30, 60 …) and GPU-Z seems to confirm that.
 
Of course, moment I close PX1 behavior of fan 2 goes back to normal as confirmed by GPU-Z.
 
This is exactly same issue another owner experienced here https://forums.evga.com/RTX-2080-Hybrid-inconsistent-fan-speeds-with-Noctua-NF12-fan-m3048579.aspx
 
Considering fan is working just fine when card itself is controlling it and issue starts only when PX1 is in charge of fan that to me indicates bug in PX1.
 
Can EVGA please look into this?

DeadlyMercury
This is known bug: every time px1 updates fan settings - it actually stops and runs fan. But default update period is 1 sec, so it stops your fan every second.

Call me stupid but why would change in voltage (%) require restart of fan? How are others doing it? Nvidia obviously doesn't have problem doing it right (when it is card itself controlling fan).

DeadlyMercury
Well, first of all we are talking about pwm signal, not voltage, …
 

 
I stand corrected :)
 

DeadlyMercury
And there is no such bug with vrm fan or with aircooler.

 
Great point. IMHO, just the fact that there are no problems with Noctua, neither at auto nor manual set speed, when it is the card controlling the fan is enough to point in the direction of solely PX1. Otherwise all of them would be having a problem, none of them would be able to do it right.

DeadlyMercury
Anyway because pump is loud too - most users connects both pump and fans to motherboard to have ability controlling pump speed via voltage, so nobody actually cares about that bug.

 
I -DO- care about that bug and I feel others should too.
 
I bought Hybrid as pre-assembled card by EVGA for the sake of convenience and simplicity. If I wanted to spend time and money on building my own solution I wouldn’t have bought something pre-made to start with, somebody else would’ve gotten my business.
 
And I -DO- need custom fan curves.
 
Custom fan curve is only way one can avoid that annoying grating / clicking sound from card fan (not rad fan) every time GPU temperature change makes it kick in from zero RPM.
 
Also, if I disconnect rad fan from card I will be able to have custom fan curve for card fan but I will lose ability to have custom fan curve for rad fan based on GPU temperature. Are you aware of any aftermarket solution that lets you control fan based on temperature of -GPU-?
 

DeadlyMercury
And you can also check, that stock fan has same behavior: no problem on auto or fixed speed - turning on and off with custom curve.

 
I haven’t seen such behavior with stock fan, PX1 reports back its rpms are what they should be.
 

DeadlyMercury
Well, answer for the first part - "then dont use custom fan curve" :) But you already trying to fix another "bug"  or kind of build "own sollution" :)

 
Not using custom fan curve is what I will have to settle for if EVGA doesn’t solve this. However, I hope it won’t come to that as that intermittent grating/clicking sound from fan 1 is really really REALLY annoying. I would prefer 1,000x more to have it never fall below certain RPM level than have that sound.
 

DeadlyMercury
Yeah, vrm fan clicks when starts and also it generate much noise on rpm settings lower than 10% … but evga still didn't make any. … I have no idea why. It is either not possible for some reasons … or they don't have many tickets about that bug.

 
I wouldn’t call it a bug, I would call it an aftermath of bad design choice. Noise seems to be coming from fan’s motor. Zero RPM mode might be in Nvidia’s BIOS, not EVGA’s choice nor something they could change. This might be result of combining the two. Which makes it even more important to have correctly working custom fan curve.
 

DeadlyMercury
I used thermal sensor for that, my motherboard have one thermal sensor header, so I used that and placed sensor under backplate (pushed through hole on it).

 
If that is what I would have to do then things would become more complicated than I am in the mood to deal with.

I’m sorry I couldn’t post my reply sooner due to issues EVGA forums were having.
 

DeadlyMercury
Then 2nd bios is a solution. But I am talking about 2080ti and don't know does 2080S have it.
Try to search for bios switch on your gpu, it could be located near external power. If you card has one, then switch it to OC bios and that will prevent fan from stopping.

 
Thank you! Thank you! Thank you!!! I had my card in front of me like 1,000 times and never noticed this switch! I guess small gray font on blackish background makes it invisible to my eyesight. Switching to OC BIOS addressed the issue of grating and gave me bit better thermals :)
 

DeadlyMercury

ZoranC
I haven’t seen such behavior with stock fan, PX1 reports back its rpms are what they should be.

Hm, did you tried custom fan curve with stock fan at it was working normally?
Mine wasn't.

 
I stand corrected again :) The other day for the first time ever I saw same issue with stock fan on standard BIOS, just in much much lesser frequency than with Noctua which is what was making me think I don’t have it with stock fan. But once I watched long enough … yeah, it was there, 60 RPMs is definitely not 40% of stock fan 2.
 
So to recap: When one tries to use PX1 custom fan curve with either stock or Noctua fans reported fan 2 speeds end up oscillating between 0 and whatever they are supposed to be. I didn’t observe it happening when it is card controlling fans, neither with standard nor with OC BIOS.
 
Which to me strongly points in direction of PX1. It is either interfering with fan speeds or it is interfering with readout of fan speed. I can’t say which one of those two it is as I don’t have a way, nor knowledge, to find out what fan was actually instructed to do nor what are its actual RPMs.
 
-HOWEVER- that does NOT mean everything is working fine when card itself is driving fan!
 
When I rely on card to drive fans rad one intermittently stays “stuck” at 800-ish RPMs. In turn temperature hits, for example, up to mid 70s C when running TimeSpy Extreme. Otherwise fan reaches as much as 1400+ RPMs and temperature stays in mid 50s C. I’ve observed that happen with both stock and Noctua fan, just more frequently with Noctua.
 
So in the end I gave up fighting with EVGA (for now) and did exactly what you suggested I should do: I am using OC BIOS for driving of the fan 1 so I don’t have to ever hear that grating sound and rad fan I’ve connected to SYS2_FAN on mb at 80%. Even at 80% Noctua is still very quiet, quiet enough that at the moment I can’t notice it / it doesn’t bother me, yet TimeSpy Extreme temperature doesn’t exceed low 50s C.
 
Thank you again!

DeadlyMercury

ZoranC
Which to me strongly points in direction of PX1. It is either interfering with fan speeds or it is interfering with readout of fan speed. I can’t say which one of those two it is as I don’t have a way, nor knowledge, to find out what fan was actually instructed to do nor what are its actual RPMs.

Well, it actually stop fans, not just read fan speed wrong. And with default update interval (1 second) it means that fan is barely spinning and that leads to overheating.

 
That is what I too am suspecting but don't have a way of proving because even if rad fan is not stopped every second but never moves beyond that magical number of 800 rpm (when PX1 is not running and it is card controlling it) I still get similar temperature rise.
 

DeadlyMercury

ZoranC
So in the end I gave up fighting with EVGA (for now) and did exactly what you suggested I should do: I am using OC BIOS for driving of the fan 1 so I don’t have to ever hear that grating sound and rad fan I’ve connected to SYS2_FAN on mb at 80%. Even at 80% Noctua is still very quiet, quiet enough that at the moment I can’t notice it / it doesn’t bother me, yet TimeSpy Extreme temperature doesn’t exceed low 50s C.

So if you gave up that way - you actually can use msi afterburner to control vrm fan (custom fan curve works fine there, but msi afterburner cant control rad fan) and even switch back to normal bios to get that zero rpm at idle.

 
Thank you again for the advice :) As of now I don't see need for Afterburner, VRM fan seems to be behaving fine when controlled by OC BIOS keeping temperatures at the good level, but it is good to know what I might be able to do if something changes.