You probably set the UEFI (BIOS) to regulate fan speed depending on load/clock/temperature.
Fans maximum voltage is 12V, they all have a control line.

I think 12v. Wired 55mm cooler to 5v recently and sometimes took a while before starting to move. I don't see how much larger CPU fans can use less than 12v. I'm thinking fan speeds change with resistance (you can only use 5v, 7v or 12v afaict).

A fan can spin very fast, but be very inefficient. Plus, its more in the current than it is in the voltage.
Pretty much all fans the sell for PCs are Chinese crap. High quality fans, like Sanyo, are much more expensive, but have 10x better airflow.

Stojke wrote:

You probably set the UEFI (BIOS) to regulate fan speed depending on load/clock/temperature.

I did. But what's happening is that the maximum fan speed seems to be different every so often, even though it's set to 100% in the BIOS when the CPU reaches 60c.

That said... right now I'm rendering my next ADG episode again (noticed a typo and had to correct it from an earlier render) and I'm noticing that the fan speed is still very slowly increasing, even though the CPU temperature has been 61c for quite a few minutes now, leaving me to wonder if perhaps the sensor controlling the speed is different from the sensors I have access to. *shrugs*

I've never seen the temps break 62c under full load, even with this fan weirdness, and I know the maximum safe temp for my CPU is 65c, so I'm not really too concerned, it's just odd. :P

62c is nothing. When it reaches 85c and the fan is still jumping up and down than its the time to worry.

Holering wrote:

I think 12v. Wired 55mm cooler to 5v recently and sometimes took a while before starting to move. I don't see how much larger CPU fans can use less than 12v. I'm thinking fan speeds change with resistance (you can only use 5v, 7v or 12v afaict).

Modern PWM is very granular, and older voltage controlled setups (3-wire with tach return) will usually have somewhat more granular control than 5-7-12. For example the 3-wire on my A8N has at least 5 "tiers" it can select for the CPU fan. The 4-wire setups on newer graphics cards you can pick whatever RPM % you want and it will get fairly close to the specified target. With conventional fans that just hook up via 2-wire you can still feed them off a pot which will vary voltage from 0-12, usually starting voltage for most fans (especially big ones) is relatively high (>5V) but they may be okay actually running at lower than starting voltage (e.g. I have some Antecs that won't start on lower than 7V, but if you hook them up to 3-4VDC supply and start them with your finger they'll run just fine).

Now, as far as what Gemini000 is seeing: it's likely that the sensor reading RPMs (or the fan itself reporting RPMs) is not very accurate (remember folks: the sensors in most computer parts are barely enough to keep the part from cooking to death, they aren't dead-to-rights accurate measurement instruments). It may also be that the software controlling whatever ramp you'd like to control isn't the best thing in the world as well. All of this stuff is "just good enough" - accurate enough for its needs, but not for a shot to the moon. If the CPU isn't overheating, that's all that really matters - the sensor, fan, etc is doing its job as well as it can be expected to. 😊

The box of my aftermarket Cooler Master Hyper 212 EVO says:

Rated Voltage: 12VDC
Rated Current: 0.19A
Power Cosumption: 2.28W

The fan almost surely runs on 12V.
If you look close at the fan, I bet you'll find a little blue thermistor in there. This device acts as a variable resistor on the 12v input - it's resistance changes according to temperature. It's supposed to regulate the fan speed, but it's really an unwelcome feature IMO. Due to their location, these thermistors are very slow at sensing temperature changes. They lag minutes behind the motherboard and on-chip sensors.

The result is that the motherboard/CPU detect an increasing temperature, the fan speed is set for 100%, but the thermistor hasn't warmed up yet. So for the next several minutes, the thermistor's resistance prevents the fan from hitting it's maximum speed. It takes it's sweet time slowly reacting. The board commands some percentage of maximum, but the meaning of maximum is defined by the ever-changing state of that silly thermistor.

On a modern motherboard, a thermistor in the fan is just redundant and interferes with the board's own control mechanisms. Yet it's common. AMD provided these type of fans with many (maybe most) of their K8 and K10 CPUs. I don't know if they're still doing it.
With modern multi-core, variable Vcore, variable speed power managed CPUs, a load spike can cause dramatic shifts in the heat output of the CPU. It can go from 10W to 80W in the blink of an eye. In that situation, the thermistor is a liability.

For some strange reason, it wasn't until AMD started making these types of processors, where thermistors are the most unneeded and problematic, that they started including thermistor fans. At the same time, they started using 4pin PWM connectors - so they introduced 2 conflicting ways to control the fan, operating in serial. It's very odd. If they had used thermistors in the K7 era, it might have made some sense. On the K8 and later, they're just dumb.

I'm picking on AMD here, but whether yours is an AMD or not, I think you're seeing the same problem.
If it bothers you as much as it does me, then my suggestion is to put a short across that thermistor. On the 2 fans I've done this to, I just desoldered it and bridged some solder across the connection points. On some fans this is easy to do, on others it takes some disassembly.
This makes the fan become instantly responsive, and it always has the same maximum speed available (if commanded by the board). Board says 60%, it runs 60%, period. Depending how the fan speeds are programmed on your motherboard, the accelerations might become abrupt, so that might annoy some. But it's no different than if you installed a fan that didn't have a thermistor to begin with.

When my brother bought a quad-core K10, he had overheating problems in the first few minutes of video rendering. Bypassing the thermistor fixed that.
Later, I ran into a similar problem on a K8 Opteron 280. The problem was made worse because I undervolted it, so it has an exaggerated difference between idle and load power.

I know this is old, but I'm converting game footage for my next filler video right now and have been monitoring my CPU fan speeds relative to temperatures... Once the temps stabilized at around 60c, the CPU fan speed continued to slowly increase for a few minutes, eventually stabilizing about 250 RPM higher than when the temperatures stabilized, and the 12v line has read 11.994v the entire time, so I'm fairly convinced the CPU fan speed is being controlled by a sensor I'm unable to actually read, and given that I've got the BIOS set to max out the fan speed at 60c, and that the max fan speed is almost 500 RPM higher than it presently is at, I'm going to guess that either the sensor controlling the fan speed is under-reporting the core temp, the sensors I can actually read are over-reporting, or the truth is somewhere in the middle. :P

Stock cooler?

Gemini000 wrote:

I know this is old, but I'm converting game footage for my next filler video right now and have been monitoring my CPU fan speeds relative to temperatures... Once the temps stabilized at around 60c, the CPU fan speed continued to slowly increase for a few minutes, eventually stabilizing about 250 RPM higher than when the temperatures stabilized, and the 12v line has read 11.994v the entire time, so I'm fairly convinced the CPU fan speed is being controlled by a sensor I'm unable to actually read, and given that I've got the BIOS set to max out the fan speed at 60c, and that the max fan speed is almost 500 RPM higher than it presently is at, I'm going to guess that either the sensor controlling the fan speed is under-reporting the core temp, the sensors I can actually read are over-reporting, or the truth is somewhere in the middle. 😜

Is this on an AMD FX? 😊

Yes, stock cooler, and yes, AMD FX-8350. :B

Gemini000 wrote:

Yes, stock cooler, and yes, AMD FX-8350. :B

🤣 what do you expect 😀

I got a FX-6300 in my capture computer and the stock cooler is not very good.

AMD FX do not report real-world values. They report "AMD degrees" - the sensor is designed to only be very accurate in a certain threshold to kick-in and prevent the chip from cooking, not to provide consistent measurement of the die temperature (and you read that right: sensor, not sensors - there is a single sensor for the entire chip). See here for more:
http://help.argusmonitor.com/index.html?Tempe … forAMDCPUs.html

The stock cooler, as Phil points out, is also not a paragon of quality. Especially on a chip like FX-8350. I would probably upgrade the cooler, if it were my equipment.

Interesting...

Highest temp I've seen my monitoring software report was 63c after leaving it at full load for several hours doing video conversions, yet even then it hadn't triggered the maximum fan speed of 4200 RPM and was just grazing the 4000 RPM mark, which is important to note because I've seen 4200 RPM with a reported temp of 61c, yet more recently under load I'm seeing 61c with a fan speed of 3700 RPM.

I do actually have HWMonitor installed but only for reference sake. Going back and looking at it I can confirm the temps I've been reading this whole time are ACTUAL physical temperatures taken from the motherboard sensor. The actual internal sensor for the chip gives... very odd readings at idle, though under load they seem to be several degrees less than what the physical sensor on the motherboard is reporting.

The other possibility is that some setting I've changed in the BIOS at some point in time has altered what the BIOS considers to be the "maximum" fan speed, which would be weird since I have no direct control over this, just the "percentage" of maximum, which I have set to max out at 60c.

*shakes head* Well, the thing isn't melting off the board so I guess it's not THAT big an issue. I know stock coolers aren't anything too spectacular, but the case itself is extremely well ventilated with a total of three case fans, two 200 mm, one 120 mm, so I figured it should suffice well enough.

It's just really weird that I can't figure out what's actually determining what speed to run the CPU fan at. :P

Gemini000 wrote:

I do actually have HWMonitor installed but only for reference sake. Going back and looking at it I can confirm the temps I've been reading this whole time are ACTUAL physical temperatures taken from the motherboard sensor. The actual internal sensor for the chip gives... very odd readings at idle, though under load they seem to be several degrees less than what the physical sensor on the motherboard is reporting.

MB sensor is (always) inaccurate - it's too far away, so while it's reporting actual degrees, they aren't accurate to inside the chip. Internal sensor doesn't report actual degrees, see previous link. The internal sensor is calibrated around 45* C, so it should be considered accurate (as in it should kick-in and protect the chip) when the chip is at heavy load and running hot. If you aren't getting any shut-downs that's good, but from past experience running CPU/GPUs right under their thermal limit does shorten lifespan even though it doesn't immediately kill the chip. There is an advantage to effective cooling beyond lower noise. 😀

The other possibility is that some setting I've changed in the BIOS at some point in time has altered what the BIOS considers to be the "maximum" fan speed, which would be weird since I have no direct control over this, just the "percentage" of maximum, which I have set to max out at 60c.

Also consider that's not a dead-to-rights target (we are not dealing with calibrated measurement equipment); so if it says 100% it's going to say ~4200 RPM +/- X%. And that's both the fan itself not hitting whatever value dead-to-rights (it may be somewhat higher or lower), and the thing that's reading it not being dead-to-rights (it may read somewhat off). And all of this is happening in realtime.

*shakes head* Well, the thing isn't melting off the board so I guess it's not THAT big an issue. I know stock coolers aren't anything too spectacular, but the case itself is extremely well ventilated with a total of three case fans, two 200 mm, one 120 mm, so I figured it should suffice well enough.

It's just really weird that I can't figure out what's actually determining what speed to run the CPU fan at. 😜

Very likely it's being dictated by the CPU's sensor when you aren't in the OS, in the OS it's whatever software is taking over. If you don't have some software taking over, it's still BIOS dictated by CoreTemp I would assume. For example on my A64X2 machine, it stays under BIOS control until Windows loads, and then the MB's software starts up and takes over - it has roughly the same settings/features as the BIOS, but it's more granular (the BIOS basically can turn the fan on or off, the software has 3 or 4 "tiers" of speed). The only potential issue is if it's switching to software control and then relying on a MB-side sensor to adjust the fan, because that sensor isn't as accurate/responsive as it should be. 😊

I actually uninstalled the software control because every single time Windows updated some file, my case fan setting of constant 100% would be semi-lost and suddenly my case fans would only be going about 25% of maximum speed, yet would be reporting 100%, and this would be fixed by simply going into the software settings, fiddling with ANYTHING, and accepting the changes... but if this happens due to a Windows update, I could've been going minutes on end with the case fans at low speed. x_x;

The case fans, because of their size, are designed for low RPMs, so even at full speed they're very quiet, plus one fan is right over the CPU, the other is right in front of the HD. Good places for them. :B

CPU fan speed doesn't actually change during a reboot. It goes full speed on cold startup for only a moment, as well as when coming out of sleep mode.

Again, I'm not trying to debate my cooler choice or how bad or good the temps are or anything like that. I'm just trying to understand why my CPU fan speeds are not consistent with the temperature readings I'm seeing. :P

I'm using SpeedFan and it's great. 😀

I think you almost definitely have a thermistor inside the fan which is interfering. It's a dumb feature of AMDs stock fans that shouldn't be there, but it is, at least in the K8 and K10 fans I've seen.
The thermistor is adding a variable resistance to the 12v supply, so when the motherboard calls for 100%, it's not really happening.
If you bypass it, this behavior will be fixed. I've done that on 2 or 3 of these type of fans. They work much better when the power can reach the fan unimpeded. The motherboard will then have absolute control of the fan, as it should have, without the thermistor causing any lag.
Just goober some solder across the thermistor pins, or solder a tiny bit of wire if you can.

I recall one of the fans from a K10 required some disassembly before I could reach the thermistor, but the other kind was easy. Hopefully yours is the easy kind.