VROOM! VROOOM! ASUS HD7850 DCU fluctuating fan

[TRWOV]

A video about an issue with my ASUS HD7850 DirectCU: when 2d clocks are active the fan control works as intended but as soon as the card gets a 3d load the fan speed begins to fluctuate a lot, making this annoying vroom! vrooom! sound.

Setting a fixed fan speed on Afterburner or Overdrive doesn't work... well, it actually does, since the speed reported by GPU-Z, Overdrive and Afterburner is the speed you set the fan to but the RPM readout and the sound clearly show that the fan revs up and down based on some kind of thermal management algorithm since once the temps go over 50C the fluctuation decreases a lot and by 60C it practically flat-lines. Not even setting a 100% fan speed gets rid of the issue since the fan sometimes goes to 9000RPM which is 2x of the 100% setting. As you can see on 2:15, as soon as I close the render test the 100% setting works as intended.

I have a Coolviva Z1 cooler that I can use on this card so this issue isn't a dealbreaker but I'm curious as to what causes this. Doing a search for "ASUS fluctuating fan" gives out several links and for some reason it seems that DCU versions are the most (only?) affected by this.

BIOS: Asus.HD7850.1024.130221 (latest on TPU GPU database)
Tested drivers: Cat 13.4, 14.4, 14.9
Tested OSes: W10 Preview, W8.1 Pro

 

[Blue-Knight]

Amazing, it sounds like a F1 car.

 

[Aquinus]

That sounds like a bearing. Don't be surprise if the fan stop spinning soon.

 

[TRWOV]

That sounds like a bearing. Don't be surprise if the fan stop spinning soon.

I don't think so, I've had fans fail before and I know how they sound. Besides that doesn't explain why manual control works fine in 2d mode: a bad bearing would grind always, not just with 3d clocks. On 2:15 you can see that I set a 100% speed while in 3d mode and as soon as I close the render test the fan spins at 100% normally.

 

[Aquinus]

I don't think so, I've had fans fail before and I know how they sound. Besides that doesn't explain why manual control works fine in 2d mode: a bad bearing would grind always, not just with 3d clocks. On 2:15 you can see that I set a 100% speed while in 3d mode and as soon as I close the render test the fan spins at 100% normally.

I get that, I'm diagnosing the pattern of sounds it makes, how it slows down and speeds up at a weird rate. If the bearing is going, it will have to feed it more juice to get it moving. So the fan starts spinning like a bat out of hell because static friction was overcome by the motor, it spins way faster than its supposed to, so it cranks it back, but since the bearing is theoretically failing, the fan will spin slower at a faster rate than it would otherwise, resulting in the fan controller jacking up power on the fan to overcome it again.

I seriously think it's the bearing track. If the bearing itself failed, it would have seized up already. You might be able to reproduce it at lower speeds by changing the orientation of the fan on the video card. For example: Testing it where gravity is pull up on the fan (away from the GPU) and where gravity is pulling down on the fan (towards the GPU). It sounds like to me, when the fan speeds up, it's lifting and hitting a bad part of the bearing track where at idle there isn't enough lift on the fan for that to occur.

This is all guessing though. 9000 RPMs at full speed seems a bit high though, regardless of what's wrong if we forget the strange behavior for a moment.

 

[TRWOV]

I just installed the Coolviva Z1 on the 7850 and noticed that the ASUS cooler matches the 3850's mounting holes. I mounted it on my 3850 AGP and the fan works normally.

I'll connect the 3850's fan to the 7850's header and see what happens. If it does the same thing then we can rule out the fan.

 

[Aquinus]

I just installed the Coolviva Z1 on the 7850 and noticed that the ASUS cooler matches the 3850's mounting holes. I mounted it on my 3850 AGP and the fan works normally.

I'll connect the 3850's fan to the 7850's header and see what happens. If it does the same thing then we can rule out the fan.

I would be seriously surprised if the fan controller is going nuts. I don't know how a fan can spins faster than normal without more voltage than normal. Do you have a volt meter? Can you measure the output on the +12v? If by chance you have a oscilloscope, I would also check the PWM signal.

 

[TRWOV]

So I connected the HIS 3850 fan to the 7850 and the same happens:

Already installed the Coolviva on the 7850 so I won't be hearing that anymore but I'm still curious. Not even a bios flash got rid of it so I guess it's some form of (retarded) thermal management by ASUS.

 

[Schmuckley]

You know,everything you need to fix that is right here on this site..
hmm...BIOS flash did not help..weird..
I would've made it flatline @ an acceptable speed.

 

[TRWOV]

Yes, I flashed the latest bios ( http://www.techpowerup.com/vgabios/141769/asus-hd7850-1024-130221.html ) on the card. Mine originally had the 120523 bios. In fact I tested several things before posting as I've had the card for a week now (changed drivers, OSes, etc).

I can't make it flat line since even after setting a fixed clock (50% and 100% in the video) as soon as 3d clocks become active the RPMs fluctuate. In 2d mode it doesn't do it: I made another test by disconnecting the Coolviva's fan and even after getting to 55C the RPMs didn't fluctuate on the HIS fan.

As stated, this only happens when 3d clocks are active. Now it's not a problem since I installed the Coolviva; I've had that cooler for 5 years now so I don't really mind since it was just collecting dust. But still I'm curious as to what causes this. It's not an isolated happenstance as I've read some posts on the matter and for some reason DCU models were the ones discussed. I even found one about a guy that got three cards with the issue (1 new, 2 RMAs). Maybe it's an issue that only presents itself in certain hardware configurations or something?

I've posted the videos on the ASUS forums. I'll report back with whatever they answer.

I would be seriously surprised if the fan controller is going nuts. I don't know how a fan can spins faster than normal without more voltage than normal. Do you have a volt meter? Can you measure the output on the +12v? If by chance you have a oscilloscope, I would also check the PWM signal.

Could it be that the actual maximum speed of the fan is 9000RPM and ASUS decided to set the 100% setting at 4500RPM or something? The 20% setting is 1200RPM so the RPM aren't in a linear correlation (1200*5=6000).

 

[andrewsmc]

You know,everything you need to fix that is right here on this site..
hmm...BIOS flash did not help..weird..
I would've made it flatline @ an acceptable speed.

I think that is why he is here..... to get help to fix it?

 

[PHaS3]

I had an MSI 6970 that did that. It was a weird "revision 2" received after an RMA. Also tried flashing the BIOS to get rid of it, to no avail. Nothing I could do fixed it, so i sold it. I had it as a spare card anyway.

What I assumed was that the fan controller itself had failed somehow.

 

[Schmuckley]

pull fan,use waterblock..
or pull fan from heatsink..use regular fan

 

[TRWOV]

pull fan,use waterblock..
or pull fan from heatsink..use regular fan

Already did that as stated on the very first post. I just wanted to see if there was some fix for this, you know, just in case someone else gets the same problem.

 

[TRWOV]

I tried one last thing: a non-DCU bios. It went as well as you'd expect but I was prepared and flashed the card with the original bios again. So I guess this is thread is finished unless someone comes across with a fix.