Hiya,

I build a HTPC and I'm using SpeedFan to control the fan speeds. I get a great range of speeds (standstill to max rpm) when manually changing the speeds between 0% and 100%. I followed this guide: http://unitstep.net/blog/2007/11/02/usi ... an-speeds/ and it partially works.

I set the desired cpu temp to a value, say 50c and the warning temperature to say 75c, with minimum fanspeed 25% and maximum 75%. In a perfect linear system that would mean 50% speed change over 25 degrees, i.e. 2% per degree.

However, as soon as temperature goes only a few degrees over the 'desired' temperature, fan speeds reach their maximum of 75%. When I set the warning temperature way up to 125c, the problem still remains. Fan control seems binary: lowest fan speeds when temperature is below 'desired' and maximum fan speeds when temperature is above desired..

Anyone experienced this?

For example, with Desired = 26c, Warning =120c, occuring temp = 27c
Minspeeds=0% Maxspeeds=100%
Actual speeds: 73% and 86%
Thats a bit steep when just being 1c over temp eh!

Any hints to make it less binary?

Thanks,
Niels

Go to Config/Options and change the "delta" value. The default is very large, resulting in aggressive and annoying fan speed changes triggered by small temperature changes.

I use 3% now, but have set this value as low as 1%.

Double check if the fan speeds are tied to any other temperatures. If you've installed a new hard drive recently, that could do it. The new HD temp, by default, is tied to all the speed controls.

Thanks.. The delta value doesn't change it; at 1% it just takes longer but eventually only being 1 degree above the desired temp will end up at near or full speed fans.

I've also kept it simple, speed 1 and 2 are just tied to one temperature (the cpu socket temp sensor).

Shame! Its got to be a software thing; it reads temperature properly, and manually adjusts fan speeds properly; why can't it linearly add fanspeed % as temperature grows a from a set desired to a set max..

If the delta is at 1% and the situation you describe happens you have one or both of two problems

1. Your fan at full speed is not fast enough to cool the monitored item therefore it will never cool off to the specified temp.

2. Your specified temp is too low for a realistic temp for the monitored device.

for example if you have a temp of 30c at idle (after minutes or hours of use) and you set the desired temp to 31c then any activity at all will put you in violation of #2. The realistic thing is to set the fan at a fixed high RPM, test at your worst load, then pick a desired temp between the idle measured previously and the load temp you get with a high fan speed.

"desired" in speedfan is not your ideal temp its the temp you want the fan to ramp up at. If you are seeing 26c at idle and 27c under normal use then set desired to 30c. Don't put your desired temp so darn close to the idle temp.

The desired temps I'm using right now in speedfan are:

cpu core 45
cpu core2 45
HD0 40
chipset 40
case 35

actual temps as I write this are

~40
~35
~32
~31
~25

Give speedfan higher desired temps and it won't be so reactive.

Thanks for the reply.



That is not the issue here, temperature can be kept nicely under control. That is the reason I set the desired temp to 26; because the actual temp was 27 and I wanted to see the effect of being 1 degree above 'desired'



Indeed that was, thing is though, I can load both cores and get them to about 50c. If I set my desired temp to 50c the same happens; the entered minimum fanspeed is active from 0c to 50c, and as soon as the temp hits 51, the fans will switch to the highest entered speed.

There is just no smooth transition from minspeed to max speed over a decent range of temps; whatever I do the effect is lowest / fastest fan speed. Desired temp, in my case, acts like a switch between slow and fast fan speeds. Sure I can say slow = 50% and fast = 60%, making the increase in RPM smaller, but I want speeds to gradually grow from the desired temp (min speed) to the warning temp (max speed)..

This is the first mobo I tried this on though I suspect the issue to be in Speedfan, not the mobo, as the speeds change very nicely between 0 and 100%, the temperatures don't jump all over the place to disturb the control.. Unless it also checks for RPM; the cpu fan spins at about 500rpm and the psu fan at about 650; if Speedfan reckons thats way too slow, that might explain the binary fan control. I could try a speed multiplier in the advanced settings but I have little hope!

Mobo is the Gigabyte GA-MA78GM-S2H, the 780 board reviewed by SPCR btw. It also comes with its own ugly software that I might try..

Actually, it is smooth, it just happens very very fast! This is on purpose. The idea is to have the fan speed at something other than minimum for the shortest amount of time possible (ie better to have 2 minutes @100%, than 20 minutes @60%). There is a way to get Speedfan to behave the way you want, but I think it is not desirable. The way is to use very high RPM fans. My experimentation has shown these fans are far more likely to hover at a value in between min and max. However, I find this behavior worse than the quick ramping up and down you get from lower RPM fans.

In my opinion, the correct way to use Speedfan is to set your desired temp very close to the highest temp you consider safe (eg something around 60C for modern CPU). This way, your fans almost never ramp up and they also begin to ramp down quite quickly, so you are mostly at your idle fan speed -- where you want to be. Yes, it is quite noticeable when the fan speeds up, but I think this is an acceptable trade off for it almost never happening.

I don't think it's desirable. What ends up happening is that when you have an intermittent high CPU load, the temperature flips between low and high, and the fan along with it. Such rapid temperature changes don't do it any good, and the rapid fan speed changes are annoying.

It should instead try to maintain a balance between temperature and noise. Ideally you would be able to set a min and max desired temperature with the fan speed linearly interpolating between those two temperatures. That way you could have the current behaviour by setting the two temperatures close, or you could have a more gradual change by spreading them apart.

This doesn't happen if you set the desired temperature relatively high, unless you have ridiculously over powered fans. There is a hysteresis built in, so your fan will not begin to slow down until you are a couple of degrees under the desired temp. Also, the fans ramp down much more slowly than they speed up. In over two years of using Speedfan, I have never seen the fan speed flipping you are describing. The behavior is always the same -- fan ramps up very quickly, stays at max for several minutes, then slowly ramps down. If running at something like 70% CPU utilization for a long while, the fan will only ramp down part way, until the load actually drops back to idle for a while.


Well, it doesn't. However, it is pretty easy to get Speedfan to approximate most of this behavior. Set the desired temp very close to the hottest you want things to get (what you set the warning temp to is pretty irrelevent). You will find this method gives you the gradual ramping you want on the downstroke and, while the upstroke may be abrupt, most of the time things will never get warm enough to trigger it. If you can't live with this, either get a T-Balancer or write your own software . . .

OK, so you are saying that if the min fan speed is 50% and the max is 75% and the delta is 3% you don't see it go 50, 53, 56, 59, 62, 65, 68, 71, 74, 75 over a space of tens of seconds? It literally switches from 50 to 75 in one jump?

If so, is the fan a 4 pin PWM or a 3 pin?

Its pretty hard to keep the temperature exactly at 1 degree above or below the desired temp, it tends to go down as the fans ramp up so what I describe is likely to happen with just being 1c higher than desired but it might also be 2c.

It doesn't jump from 50 to 75 but it takes steps as set in the options equal to the delta so with delta=10 its 50,60,70,75% speed and with delta =1 it is 50..51..52..75% speed, taking longer to reach 75% of course.

Its a regular 3 pin Nexus fan.

I'm quite sure it is just how the software works. I could wait until summer and see the amount of cooling required for a very hot day, and set those required fanspeeds as the 'max' speed, so speefan will either put the fans in 'slow cool days' mode or, when busy, ramp up to 'hot day' mode.

Its a lot better than nothing but I can't see why it can't apply a linear fanspeed curve between desired and warning temps, that can't be more than an hour of programming!

Because this not how the author (Alfredo Comparetti) wanted his program to work. He wanted to the fan to spend as much time at its idle speed as possible. The idea of the current algorithm is to maximize time at idle speed, not to (as you want to do) minimize time at top speed. The Warning Temperature isn't there so much to help you set a temperature range for fan behavior as it is to give you a warning that your cooling is inadequate or has failed.


You don't need to do this at all. I think you are missing the (admittedly undocumented) fact that reaching the Warning Temp will cause your fan speed to go to 100%, regardless of what you set Max Speed to. So, you could year round have a setting like Min=50% and Max=70% and set Warning Temp = Desired Temp +5. This way anytime it gets especially hot your fan will go up to 100%, but the rest of the time they will not go above 70%. Again, the idea is not to use Warning Temp to try and create some kind of response curve, but instead have it as a failsafe.

Fair enough, if Alfredo had the time he could probably add a few methods of fan speed control.

Basically then there are 3 speeds we can use with the current Speedfan
lowest speed: below desired temp
medium speed: above desired temp
max speed: at and above warning temp

That is probably flexible enough and pretty safe too.

No, if you Play with the delta, desired, and min temps you can get the fans to be way less reactive. On my system its more like:

lowest speed below desired temp for some long period of time
variable speeds below/at/above desired temp
max desired speed above desired temp for some long period of time
100% speed way above desired temp (never happens unless I load something like RTHDRIBL and Prime95 for several minutes.

Seriously I often see my system above min fan speed but below the max desired speed for extended periods. I always set my min speed within a few percent of the stall speed for the fan and it isn't uncommon for me to see the speed vary to something like 10 or 20% above the stall speed and still be 40 or 50% below the max.

When you have it balance that delicately the drop off to min speed still takes a while and the spin up to max still takes a while so it just varies around what it needs to be without hitting either limit. You just have to play with the settings enough to see that behaviour some of the time and be willing to stop looking at the fan speeds once you do.

Your ability to do such things has more to do with the capabilities of your HSF than your tweaking of Speedfan. Many people can't get away with fan speeds <<50%, even at idle.

Thanks but I'm just using the stock HSF on a 89W 90nm AM2 4200+. No heat pipes, no special thermal grease, no magic fans. Just a plain block of aluminum with a 70mm fan on top.

I've been able to find similar sweet spots of settings on a 939 130nm 4200+ again 89W with a stupid aluminum block 70mm no heat pipes stock HSF.

It's not the processor or the HSF, its knowing how to set ALL the settings in speedfan together and being persistent enough to tweak them over months of use without overdoing the adjustments.

I was actually looking forward to buying a lower TDP processor so I wouldn't have to hear the fan when speedfan actually does ramp it up. There is one game I play that is not multicpu aware and pegs out at 100% cpu on one core doing simple stuff that you could do on a Pentium III.

FWIW, my stock HSF is set to run between 29% and 85% but spends the vast majority of the time pegged at 29% which is about 715 RPM.

Well, what works for you works for you. I think you are simplifying too much to think it will work well for others. Just by switching apartments, both air-conditioned but with different airflow patterns, I've had to increase my idle PWM from 50% to 70%. Mind you this is only buying me idle temps averaging around 40C (on-die sensor) on a stock voltage E6300 with an aftermarket heatpipe cooler that uses a 120mm fan. The HSF is tower style and is parallel to the case exhaust which is also 120mm and PWM controlled by the same header! Again, this is with an ambient temperature of around 22C. Now I've noticed that even 90nm X2 tend to idle a lot cooler than early model C2D, so that probably accounts for a lot of the difference in our idle fan speeds, but that is what I was trying to say in the first place -- not everyone can run their fans at <50% even at idle!

That said, I'm sure you are very good at configuring Speedfan. I didn't mean to question your skills, I just wanted to point out that even skills are not enough for all varieties of ambient conditions and equipment configurations.

I think that is a key point. The original poster in this thread has a HTPC presumably with little or no case fans and possibly poor airflow as regards to exhausting heat from the CPU area.

Part of what makes it work for me is I tend to put more case fans in than the most vocal SPCR members and then run them at low voltages.

So with a SX835 case that uses 80mm case fans I put 3 front and 2 rear and run them all at low voltages.

With a Solo I put a 120mm in the rear and two 92mm in the front and again run them at low voltages.

Yes I could get away with less fans in either build but the extra front to back airflow helps prevent stale air/heat buildup around components. Besides I can't hear the difference between a solo with a 120mm fan, a solo with 120mm + 92mm@5v, or a solo with 120mm + 2x 92mm@5v. Adding low speed fans to the front don't make a significant difference to the noise level or tone as far as I can tell.

Quiet is good but going for minimal fans in some case designs will leave you with poor airflow and for speedfan users that will show in a controlled fan ramping up more than you would like.

I'm not saying that changing settings in speedfan is going to make a bad build suddenly better. I'm just saying if everything is right then speedfan works better than if something is on the edge thermally.

Maybe I came across as too optimistic or condescending or something (pick a label so much of non face to face communication is inferred). If so, I'm sorry. If you didn't know that you could hit balance in your speedfan settings you might give up and never try. I'm just saying don't give up on speedfan in general because the default settings don't work on a specific build.

The 'funny' thing is, there probably isn't even a NEED for using Speedfan as the fans even when inaudible (psu 650rpm, case/cpu fan at 500rpm) temperatures struggle to hit 50c, only when both cores are stressed a while with Prime95.

I'm still happy that it (speedfan) exists to have some failsafe, should it get hot the fans will spin up. I was just surprised that the temperature monitoring and fan speed is controllable, yet the implementation, though really easy to program, is fairly simplistic. The base is there, why not allow a temp vs speed curve etc etc..

Still though, its a usefull tool that could've been a lot more impressive from a fan control pov.

I'm actually talking about airflow outside the case. I think this is a badly overlooked aspect of cooling. Lot's of people focus on room temperature, but forget that ambient airflow can make a significant impact (5 degrees or more). Basically, good airflow right around your case intakes/exhaust makes it as if you are running higher speed fans.

fwiw I caught your implication I just went a different way with my post. It's very true that threads on SPCR often get caught in minutia when there is the whole:

component temp/airflow < subarea temp/airflow < whole case temp/airflow < room temp/ariflow < outside temp

I have the exact same problem. Has anyone found a solution to this?

I have my temperatures set a Desired: 50C and Warning: 70C and RPM limits at 40% and 100%. So if the core temperature is 60C, I want the fans to be at 70% PWM. However, they go to max right above 50C (e.g. 52C).

Have your read the relevant SPCR article? In case, post comprehensive screenshots of your configuration tabs.

Then your second speed value should be 70%, not 100%.
(100% speed happens at the "warning" temperature anyway, regardless of other settings.)

But if the second value is 70%, then it will hit 70% at 52C, which is not what I want either. It drives the speed to the set maximum from the minimum the moment that you cross the desired threshold - this is not what I want. I want it the RPM to scale in proportion as the temperatuer between 50C and 70C. This is what the ASUS MB BIOS control does, which is what I want, except that the minimum is 60%, which is too high.

yeah, in my current config I'm using 25% min, 55% max and that's only because I'm overclocking the video card. I had the max fan set to 40% before I started overclocking the video card.

If you set the max fan speed value or the warning temp value too low you'll know it when it jumps straight to 100%, if it never jumps to 100% then you know you are close to dialed in on where it should be limited.

Anyone know of an open source alternative to SpeedFan? Does ASUS provide any sort of motherboard SDK? The goal is to be able to customize the software to do exactly what I want.

In that case, you can use the "Advanced fan control" feature to set up the temperature/speed curves

Thanks that seems to be working!