Need help with PWM signal for blower fan

[Hybrid]

So the PWM controller on my HD 6970 is dying and started freaking out,

So i made an adapter and connected the blower fan of the card to the motherboard, but the PWM signal from the motherboard is not good for this blower fan.

 

Now i don't really now what to do,

I thought about using something like a rasperry pi or arduino to provide a PWM signal to the fans.

But i don't know how the signal should be since i don't have a HD 6970 with good PWM signal to measure.

Is there a way to ensure that the duty cycle and RPM are synchronized?

If there is a better way to do it i would like if you let me know

 

[Leeghoofd]

Just invest 80 bucks in a 2nd hand card and get more performance... A 6970 is like 20-25 bucks here on fleabay

[Hybrid]

On 10/13/2022 at 6:56 AM, Leeghoofd said:

Just invest 80 bucks in a 2nd hand card and get more performance... A 6970 is like 20-25 bucks here on fleabay

As far as i know it's a common issue with those cards,

The ADT7473 is not just a PWM controller, it also monitors the temp sensors and reports the thermals.

So the temperature reporting is also affected (On some sensors it reports 1,000C, on the GPU core it reports crazy low values like -2,147,483,648C)

Examples:

https://www.reddit.com/r/graphicscard/comments/ocqs5i/my_diagnosis_temperature_sensors_in_gpu_are_hosed/

https://forums.tomshardware.com/threads/radeon-hd-6970-fan-is-spinning-at-very-high-rpm.3365747/

https://forums.tomshardware.com/threads/amd-hd-6970-reports-128-c-and-fan-at-100.2638437/

 

Also i wouldn't throw a card for such minor issue just because it's cheap on the 2nd hand market, It's a waste.

Edited January 16, 2023 by Hybrid

[wutske]

One of the Reddit threads suggested that re-installing the driver fixed the issue, I assume it didn't for you ?

 

If the IC is damaged in some sort, it's always a possibility to replace it, though the question will always remain why and how long it takes for the issue to return. Does it also effect the performance of the GPU ? I mean, if it think it's at 1000°c it might be throttling.

 

As for the PWM signal, there are a couple of options: high (22.5kHz) or lower frequencies (see p29 of the datasheet), it just affects the pulse widths but not the duty. Assuming it's a 4-wire fan they've probably set it to a higher frequency.

Looking at the datasheet (p26) it seems like they're really stressing out the fact that the pull-up of the PWM signal should not exceed 3.6V (because the chip can't handle more). I'm not exactly sure how this should affect the compatibility with the motherboard though.

 

Anyhow, you could try using an Arduino, though it's PWM pins are limited to 500Hz. It's probably safest to also use a diode to pull the PWM signal below 5V in case the fan's pull-up voltage is above the 5V limit of the Arduino. It's probably also best to power the fans from something else than the Arduino, not sure what current it pulls but Arduino are pretty weak.

 

And as a final option, you could use your motherboards PWM signal to drive an NPN and use that to power the fan with (fig 36 of the datasheet).

 

Hope that helps you save another card from the bin ?

Edited January 16, 2023 by wutske

[Hybrid]

5 hours ago, wutske said:

One of the Reddit threads suggested that re-installing the driver fixed the issue, I assume it didn't for you ?

Doesn't work, i even tried flashing different VBIOS images and modifying the VBIOS to no avail.

5 hours ago, wutske said:

If the IC is damaged in some sort, it's always a possibility to replace it, though the question will always remain why and how long it takes for the issue to return.

I am certain that the IC has gone bad, from what i know the issue manifests itself after 6-7 years of use.

5 hours ago, wutske said:

it also effect the performance of the GPU ? I mean, if it think it's at 1000°c it might be throttling.

Luckily it doesn't throttle.

5 hours ago, wutske said:

As for the PWM signal, there are a couple of options: high (22.5kHz) or lower frequencies (see p29 of the datasheet), it just affects the pulse widths but not the duty. Assuming it's a 4-wire fan they've probably set it to a higher frequency.

That reminds me something about the issue i have with the PWM signal of the motherboard header,

I can clearly hear each and every duty cycle which is annoying, in theory could it be solved with a frequency increase?

At 100% duty cycle the fan is too loud but at least the fan doesn't ramp up and down every duty cycle

5 hours ago, wutske said:

Anyhow, you could try using an Arduino, though it's PWM pins are limited to 500Hz.

Can that be an issue?

5 hours ago, wutske said:

It's probably safest to also use a diode to pull the PWM signal below 5V in case the fan's pull-up voltage is above the 5V limit of the Arduino.

How do i do that?, just a diode in series?, also what diode should i use?

5 hours ago, wutske said:

It's probably also best to power the fans from something else than the Arduino, not sure what current it pulls but Arduino are pretty weak.

I will power it from the motherboard.

Edited January 16, 2023 by Hybrid

[wutske]

I honestly don't see any issue why a 500Hz PWM signal would be a problem, it's just turning the fan on and off at that rate. If I hadn't killed my motherboard I could have checked with my multimeter at what frequency it was working.

The best way is to test and see at what frequency you can't hear it, I can image that a blower creates an annoying sound when PWM'ed at a wrong frequency.

 

As for the diode, easiest way is to just put a diode from the PWM signal to ground like figure 37 in the datasheet. As for what type, I'd just pick whatever you have lying around and put them in series until their foward voltage is ~5V.

[Hybrid]

13 hours ago, wutske said:

I honestly don't see any issue why a 500Hz PWM signal would be a problem, it's just turning the fan on and off at that rate. If I hadn't killed my motherboard I could have checked with my multimeter at what frequency it was working.

The best way is to test and see at what frequency you can't hear it, I can image that a blower creates an annoying sound when PWM'ed at a wrong frequency.

 

As for the diode, easiest way is to just put a diode from the PWM signal to ground like figure 37 in the datasheet. As for what type, I'd just pick whatever you have lying around and put them in series until their foward voltage is ~5V.

I will try that, Thank you for the help!