radeon software, custom (inverted) fan curve for modded vega64.

[Pesky Ngon]

im trying to mod a reference vega64 with a much more powerful fan. replaced blower with new fan. wired new fans PWM wire to GPU. the fan speed changed with Radeon software fan curve BUT it was inverted. eg: as the GPU gets hotter, the fan spins slower. Radeon software wont let you go from fast to slow, so I cant just invert the fan curve to make it work. is there anyway around this like an alternative software or a registry edit or something? I would be extremely grateful for any advice or suggestions on this because ive been working on this project for months and am so close to making it work.

[AbydosOne]

@Pesky Ngon I'm an EE, so like to think in hardware solutions...

 

My first guess is that the PWM signal from the card is inverted (0 for 1, 1 for 0) from what the fan is expecting for some reason. If you knew the voltage of the signal, it would be pretty trivial to hack together an inverting logic circuit that would sort it out. There is also the possibility that the signal is being pulled high by the old fan and now that the pull up is missing the signal is getting inverted weirdly.

 

I don't suppose you have access to an oscilloscope to snoop at the PWM signal? What is model fan you're replacing the stock one with?

[Pesky Ngon]

thanks! the inverting logic circuit is a great suggestion. I don't fully understand what you mean by "pulled high" but I feel like I should give you some more info in case its relevant:
the new fan is a DELTA that draws something like 55w. waaay to powerful for normal fan headers. so I have the positive and negative hooked up to a molex 12v and ground. then the tachometer is detached and the signal wire is hooked up to the GPU for temp control. the only way I was able to get the GPU header to accept the new fan was by leaving the old fan attached and having both fans share the PWM signal (its not a problem if I have to leave the old blower fan attached, you can barley hear it compared to the replacement that makes 70db). ive tried a lot of combinations and this was the only successful one.

im afraid im on a very tight budget, so an oscilloscope is way more than I can afford.
the exact fan model: PFR0912XHE|

edit: also need to mention the old fan still works like normal when both are attached, ie as temp goes up, fan gets faster.

Edited April 8, 2020 by Pesky Ngon

[AbydosOne]

17 minutes ago, Pesky Ngon said:

PFR0912XHE

I think I found the issue:

 

 

From what I understand, normal PC fans are the opposite (so that with no signal (i.e. 0% duty cycle) on the PWM line, they will run at DC voltage (i.e. 100%) speed). You will either have to invert the PWM signal, or see if there's a different variant of the fan with the opposite profile.

 

EDIT: Link to datasheet

[Pesky Ngon]

3 minutes ago, AbydosOne said:

From what I understand, normal PC fans are the opposite

ahhh ok, I didn't know that. that is super helpful, thank you so much. its really nice to know what the issue is for once, most of this project has been trial and error. so in theory, all I need to do is build an inverting logic circuit that matches the signal voltage? would a multimeter work for reading the signal voltage? if it doesn't I imagine ill be able to find it somewhere like a spec sheet.

[AbydosOne]

1 minute ago, Pesky Ngon said:

would a multimeter work for reading the signal voltage?

If you can set the fan to 0%, that should get you a pretty close reading for 100% duty cycle (i.e. always on PWM signal).

 

My guess is that it will be either 5V or 12V. The Delta fan (per the linked spec sheet) isn't picky about the PWM voltage, as long as it's above 0.8V. Pulling either of those voltages from the PSU shouldn't be hard.

[Pesky Ngon]

1 minute ago, AbydosOne said:

If you can set the fan to 0%, that should get you a pretty close reading for 100% duty cycle

that's perfect, I cant quite get 0% because the minimum is 10% speed, but I imagine that will be close enough that I can work out if its 12v or 5v. 

and I can get 5v and 12v from molex. ill do some research into how to make the inverted logic thing. thank you so much for the information dude. you've probably just saved me several days of work.

[Pesky Ngon]

On 4/8/2020 at 4:52 PM, AbydosOne said:

inverting logic circuit

I now understand that I need an NPN bipolar transistor (I think) and I get how they work for the most part. ive been having a bit of trouble finding information as most search results are talking about giant AC/DC converters or on the correct topic, but beyond my understanding. ive seen a lot of people demonstrating inverters in different circuits, but I wasn't able to simulate the circuits I was seeing, I had issues like "max current exceeded" and I don't know enough to solve the issues. 

ive attached a pic of the circuit im planning to make. it doesn't work in the sim, but would you expect it to work in a real circuit?
you mentioned before that I would need to know what the GPUs PWM voltage was, im guessing id have to make the source above the pull up resistor match that voltage?
a lot of people talked about general rules, but I don't know which exact components I should be using, i.e. a 1k resistor vs a 100k or which of the hundreds on NPN bipolar transistors to use. it was mentioned that for the resistor you don't want one with too little resistance because it wont have any effect, but if you make it too strong it wont work either. im afraid this is quite far out of my depth and ive been struggling to work things out. sorry for all the questions.