RivaTuner v2.07

I keep reading postings like "I've just changed thermal table and it still doesn't work, where is the promised fix?" from AMD 38x0 owners in different forums. So the post #1 has been appended with special notes (marked with bold font) for 38x0 owners.

 

Some thoughts on simplifying transfer function explained in the post #4:

The equation

Duty_cycle = Duty_cycle_min + 6.25 * (T - T_min) * T_slope / Duty_cycle_100%

can be replaced with

Duty_cycle = Duty_cycle_min + (T - T_min) * (100 * T_slope / Duty_cycle_100%) / 16

And the equation marked with bold physically represents desired increase of duty cycle (in percents) per each 1/16C.
E.g. in our examples in post 4 we use Duty_cycle_100% = 135 and T_slope = 22, so our marked equation transfroms into 100 * 22 / 135 = 16% per each 1/16C. That gives us 1% per 1C duty cycle adjustment. Anyone find such representation easier for understanding?

 

Mike89

You've to recreate the profiles. From the release notes:

- Now all low-level profiles are linked with target display device by physical display adapter location identifier instead of logical display adapter name. Due to this change all previously created low-level profiles will be invisible to new version and you will need to recreate them.

 

I know that it sounds insane, but it seems like there are way more bugs in fan controller calibration in some AMD VGA BIOS’es than I could imagine. I’ve just found that besides screwed automatic fan speed calibration parameters, which can be now edited and fixed by 2.07, AMD BIOS guys also managed to screw fan controller’s acoustic management system calibration in certain VGA BIOS images. Acoustic management system is dedicated to get rid of unwanted fan speed fluctuations caused by temperature fluctuations. Also it serves for making fan speed change soft and less annoying for hearing (rapid stepwise fan speed change usually annoys hearing much more than soft one). There are two acoustic management related settings involved:

PWM hysteresis – defines hysteresis for resulting duty cycle generated by the fan controller. Hysteresis effect is used to get rid of unwanted duty cycle fluctuations and all duty cycle changes within [Duty_cycle-PWM_hysteresis; Duty_cycle+PWM_hysteresis] range are treated by controller as unwanted noise and ignored. In other words, if PWM hysteresis is equal to 10% and duty cycle is currently set to 50%, fan speed won’t be able to change anything within [50-10=40%; 50+10=60%].
PWM ramp – defines ramp for duty cycle changes. When ramp is defined and duty cycle needs to be changed (either by temperature lookup table or by transfer function), the controller increases current duty cycle by ramp value instead of directly setting new value. In other words, if duty cycle is currently set to 50%, ramp is enabled and set to 2% and duty cycle have to be changed to 60% (e.g. according to transfer function) then the controller is starting to increase duty cycle by ramp value (i.e. set 52%, then 54% etc until desired duty cycle is reached).

Now we’ve came to new problems in AMD BIOS:

1) Due to some odd reason AMD set PWM hysteresis to insanely high values (up to 25%) in some BIOS images. It means that the transfer function must change duty cycle by 25% or more to get new duty cycle applied. Furthermore, smart AMD controller developer decided to apply PWM hysteresis at hardware level to fixed (!) fan speed mode too. It means that those who have such high PWM hysteresis in BIOS won’t be able to adjust fan speed in steps less than 25% even in fixed mode. “Bravo” AMD.
2) Furthermore, AMD combined abnormally high PWM hysteresis value with abnormally low PWM ramp value (e.g. hysteresis = 25%, ramp = 1%). That is just plain stupid. It means that even if duty cycle is allowed to change (i.e. when delta between current and desired duty cycles is more than 25%), the controller is allowed to change duty cycle just by 1%.

Both of these parameters will be also added to the list of editable fan controller parameters RT in v2.08. I’ll also forcibly reset PWM hysteresis in fixed fan speed adjustment mode to provide soft fan speed adjustment possibility to those who is affected by incorrectly programmed PWM hysteresis.
In meanwhile, if you're affected by such BIOS, you can use RivaTuner's command line interface to access PWM controller directly and reset incorrectly programmed PWM hysteresis:

http://forums.guru3d.com/showpost.php?p=2625760&postcount=8

 

Forget about it. You're doing ambigious settings, it is up to you to avoid programming the techniques which are supposed to cause interference for sure.

 

Pausing scheduler sounds like a good idea and something to be planned for future versions.

 

No, 8800 series use different fan controller:

http://www.analog.com/UploadedFiles/Data_Sheets/206619465ADT7473_0.pdf