PWM Slope Setting

gpw928

gpw928

There's any number of documents on the Internet that tell me about the PWM settings in my AMI BIOS. They are:
  • Fan off temperature limit: FAN will stop work if temperature is lower than the Fan off temperature limit value (default 30 C).
  • Fan start temperature limit: If the temperature is higher than fan off temperature limit, FAN will start work (default 32 C).
  • Fan Full Speed Temp limit: If the temperature is higher than the FAN Full Speed temp limit value, the FAN will work at full speed (default 56 C).
  • Fan start PWM: If the temperature is higher than the FAN start value, the FAN will start work at this PWM (default 63 PWM).
  • PWM slope setting (default 8 PWM). Options are:
    • 0.125 PWM
    • 0.25 PWM
    • 0.5 PWM
    • 1 PWM
    • 2 PWM
    • 4 PWM
    • 8 PWM
    • 15.875 PWM
    • 4.5 PWM
I get that "PWM slope setting" describes a linear relationship between temperature and fan voltage, and voltage translates to speed.

I believe that PWM values in the AMI BIOS run from 0 to 127, and that 127 PWM means 100% fan speed.

So what exactly does an 8 PWM slope setting mean in terms of temperature and fan speed?
 
gpw928 said:
There's any number of documents on the Internet that tell me about the PWM settings in my AMI BIOS. They are:
  • Fan off temperature limit: FAN will stop work if temperature is lower than the Fan off temperature limit value (default 30 C).
  • Fan start temperature limit: If the temperature is higher than fan off temperature limit, FAN will start work (default 32 C).
Click to expand...

So, 2°C hysteresis off to
slow, but 32C's very cool for a running system. Desktop?

gpw928 said:
  • Fan Full Speed Temp limit: If the temperature is higher than the FAN Full Speed temp limit value, the FAN will work at full speed (default 56 C).
Click to expand...

So, only 24C from on low till on at max speed. Hmm.

gpw928 said:
  • Fan start PWM: If the temperature is higher than the FAN start value, the FAN will start work at this PWM (default 63 PWM).
Click to expand...

Which, if 127 is max speed, means only a 64 range between start and max fan.

gpw928 said:
  • PWM slope setting (default 8 PWM). Options are:
    • 0.125 PWM
    • 0.25 PWM
    • 0.5 PWM
    • 1 PWM
    • 2 PWM
    • 4 PWM
    • 8 PWM
    • 15.875 PWM
    • 4.5 PWM
Click to expand...

At 8-sized steps, that gives 64/8 = 8 steps (8 speeds)
IFF I'm guessing right (no warranty!)

gpw928 said:
I get that "PWM slope setting" describes a linear relationship between temperature and fan voltage, and voltage translates to speed.
Click to expand...

True enough here for the average voltage seen by the fan, smoothed.

gpw928 said:
I believe that PWM values in the AMI BIOS run from 0 to 127, and that 127 PWM means 100% fan speed.
Click to expand...

I'm a bit surprised they don't use 255.

gpw928 said:
So what exactly does an 8 PWM slope setting mean in terms of temperature and fan speed?
Click to expand...

Sounds like maybe 8 steps, starting at half speed at 32C, to full speed at 56C, so 3°C per step?

"PWM" used as units here seems weird, it's a method.

Certainly strange temps for a laptop, where start nearer 40C and max nearer 64C may be suitable, but I haven't had a desktop for 20 years.

Anf then, what is BIOS doing if you select say 4 or 16 size steps - unless I'm way off. And are they using hysteresis to stop 'hunting' around those (eg) 3C steps?

Further, does this apply to non-doze OSes?

cheers
 
As SirDice pointed out, the fan speed control is by Pulse Width Modulation, so it's pulsing current, and not the voltage, that controls the speed of PWM fans. A significant point, but the original question remains.

My guess is that "PWM slope setting" defines the slope of the fan speed relative to temperature (as the independent variable), and that fan speed may not reach 100% at exactly 56 C. So there will be "knee" with two straight line lines in the graph. The first line will show fan speed rising monotonically with temperature. The second line will either be horizontal (100% speed reached before 56 C, or a vertical line (100% speed not reached before 56 C).

I have no metrics available to accurately determine fan speed. I only have the fan sound as a rough guide. This is a pity, as knowing the actual fan speed would greatly assist the investigation.

I'm still hoping that somebody familiar with hardware design can tell me exactly what an "8 PWM slope setting" means in terms of temperature and fan speed (and how that differs from the other slope setting options).
 
i was looking for a way to fix fan noise on the astaro appliance
probably i will hack something based on superio(4)
 
gpw928 said:
As SirDice pointed out, the fan speed control is by Pulse Width Modulation, so it's pulsing current, and not the voltage, that controls the speed of PWM fans. A significant point, but the original question remains.
Click to expand...

Well, current and voltage are co- dependent, given a fixed resistance. With fans, the smoothing is mostly due to inductance, but there may be some capacitance involved also. The Wikipedia article shows the waveform with and without smoothing.

At say 20kHz, 50% duty cycle (square wave) it is maximum voltage (so, current, ignoring rise/fall slew time) for 25us and zero V (so I) for 25us.

However your voltmeter shows half supply voltage, and the fan spins smoothly without measurable speedup and slowdown per 50us cycle, when one revolution is (say) 15ms @ 4000rpm.

Sorry to so labour this point, but it does matter.

gpw928 said:
My guess is that "PWM slope setting" defines the slope of the fan speed relative to temperature (as the independent variable), and that fan speed may not reach 100% at exactly 56 C. So there will be "knee" with two straight line lines in the graph. The first line will show fan speed rising monotonically with temperature.
Click to expand...

I can picture that, though it will go up in steps rather than monotonically, eg one per ~3C if 8 speeds, as I've speculated. OTOH if you choose tiny steps, like 1 for 64 speeds per (default) 24C range, or even 0.125 for ridiculously teensy steps, but at what processing penalty?

Check out 'slope' in Covacat's PDF, though it maybe something different.

gpw928 said:
The second line will either be horizontal (100% speed reached before 56 C, or a vertical line (100% speed not reached before 56 C.
Click to expand...

Vertical I can't imagine, nor that it'll be anything like that precise. I've written fan control scripts for thinkpads and have a fair grip on temp vs fanspeed parameters, though BIOS is barer metal.

gpw928 said:
I have no metrics available to accurately determine fan speed. I only have the fan sound as a rough guide. This is a pity, as knowing the actual fan speed would greatly assist the investigation.
Click to expand...

Now this I can maybe help with. Get hold of PitchLab app for iphone or (better) android, which will map the fan sound to a music scale, quite accurately. Since the frequency emitted is directly proportional to rotational speed, a bit of maths and Bob's your second cousin twice removed :)

PM me for an URL for the best android version, officially withdrawn but freely licenced. I use it for planetary orbital music ...

gpw928 said:
I'm still hoping that somebody familiar with hardware design can tell me exactly what an "8 PWM slope setting" means in terms of temperature and fan speed (and how that differs from the other slope setting options).
Click to expand...

You can try non-default settings, no? I'd try values like 4 or 16, and experiment with different temp. ranges.

What's the machine?
 
I'm back in the office, and have had some more time to examine the cooling issues.

Yesterday I posted a summary of my tests on the Samsung SSD 970 EVO Plus SSDs with various combinations of fan and heatsink. Bottom line is that on a well ventilated passively cooled case, the 970 EVO Plus gets up to 100C under sustained load (may be more -- I chickened out at 100C). It's rated for 0 to 70C. No wonder their failure rates are making waves all over the Internet.

The PWM settings may be used to abate the fan noise a little, during moderate load, but running the fan flat out, even with a good heatsink, will not prevent chronic over-heating of the SSDs at 100% duty cycle. So I have not persevered the PWM configuration. I remain disappointed that the AMI documents are so murky.

I have, however, changed "Platform Thermal Management" in the BIOS (not PWM):
Code: 
    Advanced Thermal Configuration
    Platform Thermal Configurations
        Active Trip Point 0             71 ->  55C
        Active Trip Point 0 fan speed  100 -> 100%
        Active Trip Point 1             55 ->  47C
        Active Trip Point 1 fan speed   75 ->  75%
What I have found is that under sustained I/O load, the "Active Trip Point" temperature sensors used will always reach 47C, and will get up to 55C if there is any appreciable CPU activity (beyond that created by the I/O load itself). On these settings with a heatsink on one SSD, and the other SSD directly under a fan (3 mm clearance), the SSDs seem to top out at 85C (maybe more, but the increase was slow). With this setting the fans are not audible unless there is heavy CPU and/or I/O load. The SSDs still get way too hot when busy, but this setting is as good as I think is possible, given the spatial restrictions (heatsink clearances, fan size and position) of the case.

I have tried setting "Active Trip Point 0" to 47C, and that gets the fans running 100%, way too often, and way too noisy for a desktop. Critically, it does not reduce the temperature that the SSDs eventually reach under sustained load.

Given that I found no combinations of fans and heatsinks that prevented thermal overload on these SSDs, I have to accept that they can't ever be safely run at 100% duty cycle for more than a few minutes. So I have little incentive to fine tune much further.

I'm pretty sure that Australian consumer law would be on my side if I asked for a refund. And I'm contemplating that as my best option.
 
Hi again,
after our PM conversation I spent some time googl'n and intended sending my results but had a bout of illness so distracted myself elsewhere.

Took a while to picture where your fan lives, and had originally thought something like this 120mm external fan.

That'd have advantages over the 40mm internal fan of
. much greater airflow
. lower rpm, thus noise
. more room inside case

Also found bits of possible interest re CPU overheating due to poor thermal contact on some of these here.

Maybe more here.

Some report of cooler running vertically rather than flat - which leads me to ask, is yours the A, B or C model of Topton at AliExpress in terms of airflow? Apart from maybe the Model C it's hard to imagine much airflow happening, assuming a 40mm hole at the top, and another (?) at the bottom.

Hot air rising, I'd be going for the larger external fan as an exhaust fan, pulling heat out.

Hope it works out, it sure looks cute enough. Cheers
 
It's an "A"model, not quite like the photos which have shallow fins on the side of the case, but exactly like the architectural drawings.

I am aware of the poor thermal contact issues. Most of the reports I saw were old. There's been a lot of "evolution" and I don't believe I have that problem.

The vent holes are at the bottom of the case. The 40mmx40mmx10mm fan mounts in the exact space occupied by the optional 2.5" SATA SSD, which blocks a lot of the vent holes. So the fan and the 2.5" SSD are mutually exclusive.

I never really wanted fans at all, as it's a desktop system. So bolting on an external large fan is not an option because of the noise.

I'm reasonably confident that better behaved SSDs will work well. I have asked for a refund, as the SSDs are completely unfit for purpose.

I'm confident that Australian consumer law is on my side, but expect a fight. We'll see...
 
gpw928 said:
It's an "A"model, not quite like the photos which have shallow fins on the side of the case, but exactly like the architectural drawings.
Click to expand...

Ah, I didn't come across any drawings, hence my queries.

gpw928 said:
I am aware of the poor thermal contact issues. Most of the reports I saw were old. There's been a lot of "evolution" and I don't believe I have that problem.
Click to expand...

Ok. I meant to post this one.

gpw928 said:
The vent holes are at the bottom of the case. The 40mmx40mmx10mm fan mounts in the exact space occupied by the optional 2.5" SATA SSD, which blocks a lot of the vent holes. So the fan and the 2.5" SSD are mutually exclusive.
Click to expand...

Isn't there a vent in the top too? Hot air, um, rises ... sorry, I did a year's Mech Eng before programming.

gpw928 said:
I never really wanted fans at all, as it's a desktop system. So bolting on an external large fan is not an option because of the noise.
Click to expand...

Fair enough, but these nvme things seem to need active cooling, or bolting to a huge heatsink. Good quality 120mm fans are almost silent at moderate revs.

I guess a question remains re nvmes vs SATA6 SSDs: is the extra speed and cost worth these overheating and reported unreliability issues?

gpw928 said:
I'm reasonably confident that better behaved SSDs will work well. I have asked for a refund, as the SSDs are completely unfit for purpose.

I'm confident that Australian consumer law is on my side, but expect a fight. We'll see...
Click to expand...

Good luck, perseverance still furthers, hopefully.
 
You must log in or register to reply here.
Back
Top