cpu fan slows down as temps increase?

[brucek2]

Has anyone seen a situation where as the motherboard tries to drive the cpu fan higher, the fan actually slows down instead?

My CPU cooler is the Cooler Master V8. Its rated for 800 - 1800 RPM. My motherboard is the ASUS P6T Deluxe.

No matter what I do, I can't get my motherboard to drive the fan over 1250 RPM. Even worse, as the fan % increases past 50%, the fan starts spinning slower.

ASUS has a "fan xpert" calibration that shows this:
20% 434 rpm
30% 691 rpm
40% 958 rpm
50% 1205 rpm
60% 1054 rpm
70% 843 rpm
80% 767 rpm
90% 811 rpm
100% 958 rpm

I have tried every combination of motherboard settings (QFan disabled and enabled with silent/standard/turbo profile), as well as Fan Xpert. Result is the same for all of them -- once whatever control scheme is in effect tries to drive fan above 50%, it ends up spinning slower vs. faster.

What can I do to reach rated 1800rpm speed? The cooling I'm getting now is fine for regular use but if I try Prime95 it will rapidly get to 80c.

 

[SOFTengCOMPelec]

I've never seen this.
Have you tried the stock fan which came with the processor ?
Trying it, would help tell if it is your fan or motherboard, causing the problem.
It would be interesting to know what processor it was (not essential).

 

[brucek2]

Thanks. CPU is an i7 920. I have neither of a suitable fan or mobo on hand to test scenarios with, although that was my impulse for a next step too.

Part of me is thinking that as soon as spending money is required, I should probably be putting that into a newer platform instead. I had hoped for more from Haswell though so kind of in limbo atm.

 

[SOFTengCOMPelec]

Doing some quick searches on your cooler " Cooler Master V8", seem to mention an adjustable knob (fan speed ?), and some users don't think it is very good on I7's.

Did this use to work ok, and suddenly broke over time ? (hence bust fan or mobo possible)
Or is it a recent rebuild or something ? (hence incompatibility, 3/4 pin connector issue or something)

 

[tweakboy]

Ya I think it broke... gl

 

[brucek2]

There is a knob. I have it at max and the results above are from that setting. If I turn the knob down, fan speeds will decline across the board while still following the same pattern as above.

The system has been in use since the end of 2008. I think the fan has always run at the ~1200 speed and I just never noticed or cared before because the cooling is adequate for my not-that-strenuous regular workload.

I started poking around now because there's been a few so far unexplained crashes in the last couple weeks. I ran it OC'd at 3.5 for a long time but just set it back to stock a few days ago and have been "stable" since then, although I don't think I really would be at an artificial peak workload.

I really appreciate you having looked into this, thanks! Its got me scratching my head.

 

[SOFTengCOMPelec]

My spidey senses (shamelessly borrowed from Idontcare), are kind of thinking that your FAN is the problem, rather than the motherboard (pure GUT feeling, not based on tests/facts).

But, (out of my many past systems) I have had at least three computer systems in the past (probably more), at least two of which were overclocked, refuse to overclock, or some even run at stock (had to lower (underclock), to get to run reliably), as they got to about five years old.

If that "burnt out from overclocking syndrome" (if such a thing even exists, opinions/experiences differ) is what is really the bigger issue, then as you said in your previous posts, brand new haswell's (or AMD) might be your best option.

But a new fan would be a lot cheaper (but you may have to continue to run at stock and/or lower overclocks).

(The TIM, cpu thermal interface material can dry out as well, and fans can get very dirty and need de-dusting, on older computers).

 

[brucek2]

OK. Cooler Master offered to sell me a refurbished replacement fan for $5 (while also not being able to explain the lower RPMs.). I'm not so excited about taking the cooler on/off and switching out the fan (which doesn't look all that easy to change, its in the middle of the cooler), but maybe it'll be a good learning experience. I'm at least less hesitant than I was before because I thought it couldn't really be the fan, but I'll take your spidey senses over mine.

I agree there may be an underlying cpu degradation issue, but meanwhile I just got fixated on wanting to at least see what happens with the fan at its rated speed.

Thanks again.

 

[SOFTengCOMPelec]

OK. Cooler Master offered to sell me a refurbished replacement fan for $5 (while also not being able to explain the lower RPMs.). I'm not so excited about taking the cooler on/off and switching out the fan (which doesn't look all that easy to change, its in the middle of the cooler), but maybe it'll be a good learning experience. I'm at least less hesitant than I was before because I thought it couldn't really be the fan, but I'll take your spidey senses over mine.

I agree there may be an underlying cpu degradation issue, but meanwhile I just got fixated on wanting to at least see what happens with the fan at its rated speed.

Thanks again.

At $5, I'd be tempted to try it, just for the experience, and it may increase its trade-in/resale value as well.
Cpu fans are a pain in the neck, I relatively hate fitting the current Intel pin system, as I tend to mess up when it comes to the forth pin.

Good luck, anyway!

 

[Idontcare]

At $5, I'd be tempted to try it, just for the experience, and it may increase its trade-in/resale value as well.
Cpu fans are a pain in the neck, I relatively hate fitting the current Intel pin system, as I tend to mess up when it comes to the forth pin.

Good luck, anyway!

I'm with SOFTengCOMPelec here, $5 is cheap edutainment. You'll learn something in the process and it won't cost you an arm or a leg to learn whatever you are about to learn.

I too think the fan is probably just borked.

Consider how many times the fan is sold and resold until it finally is sold to you (the end-user) and that every business in that supply chain needs to mark-up the price by 50%, when you walk that back to the manufacturer you come to the conclusion they have to be making these things for maybe $0.50 each for the economics to work out on their end. (add in packaging, marketing, R&D, shipping, warranty/RMA overhead, etc)

What kind of QRA can we realistically expect a fan manufacturer to be able to afford for their production line of fans that is operating at a $0.50 BOM?

In that perspective, a borked fan or three is probably getting through to the end-user on occasion. And I'm betting the majority of them (the people with slightly borked fans) don't even know it because they see the fan spinning and just assume it is working as intended.

Only the rare individual such as the OP here goes to the effort of characterizing their fan to such a degree as to become suspicious that it might be borked.

 

[brucek2]

Turns out that "$5" fan comes with a $37 shipping charge. I do live in Hawaii but wow that's out of line. Fortunately Amazon Prime ships everything here at no charge beyond the yearly fee, so I just ordered a compatible looking fan from them instead. No separate mechanical fan speed knob but I think I'd prefer one less complication anyway.

Meanwhile in further digging, I realized that my front, side, rear, and cpu fans were all set to draw air in, leaving only the top fan as an exhaust. Probably been like that since the beginning. So I flipped the rear and cpu fans so they are exhausts now too. Bonus is that based on that experience I'm feeling much more comfortable about replacing that fan once it arrives.

That only got me a mild bump in fan speed, from ~1250 before to ~1300 now. But that same speed is now much more effective. My idle is the same at about 36c (ambient 27c), but now 10 minutes of prime95 will top out at 71c where before it would rocket to 80 and climbing within the first 30 seconds. What's key here is that 71 is just below the point where the bios failsafe will push the fan to "100%", which is actually a very slow speed, which is what cooked me before.

So my guess is I'm already in much better shape although I'm still curious about the fan speed issue. I also noticed the case fans follow the same pattern (peak speed at "50%" and slowing down beyond that), maybe indicating that they can't all have the same problem?

 

[SOFTengCOMPelec]

I also noticed the case fans follow the same pattern (peak speed at "50%" and slowing down beyond that), maybe indicating that they can't all have the same problem?

Now that you found that out (other fan(s) affected), it completely changes the diagnostic picture.

Did you perform the fan speed adjustments and/or measurements, using the bios (if so good, usually the most reliable way).

Or, using some kind of software (package) such as speedfan (usually one of the better ones) ?
(if so, these software tools are notoriously problematic, and tend to be very incompatible with some hardware, especially when trying to measure temperatures with them).

If you were using software, I would suggest using different software and/or the bios, to solve the problem.
Definitely DON'T think that anything is wrong with your hardware, until you have confirmed that the bios thinks that the fan speeds are wrong, and that the bios settings can't correct them (reach correct RPM above 50% demand).

If you have ruled out software/bios issues, then the blame would seem to be with the motherboard (assuming you are not using complicated external fan speed controllers and/or fan connector adaptors, and/or feeding all the fans via the same power connector lead on the power supply, which you are NOT doing, if I understand from what you have described so far). I.e. Your fans are simply plugged directly into the motherboard headers, for the respective fan type.

N.B. Motherboard being the culprit, is fairly unlikely in general. So there may be other things to eliminate first. Hence the above grilling (sorry) for unmentioned fan controllers, weird wiring arrangements etc.

 

[brucek2]

Yes, each fan is attached to its own fan header on the motherboard. The CPU fan & header are 4-pin while everything else is 3-pin. The case fans are the stock fans that came with my Cooler Master HAF 932 case.

BIOS-level control of CPU fan is very basic. It offers a real time readout of the fan speed, and four choices of speed profile: Silent, Standard, Turbo, and Disabled. Changes to speed are not applied until you save & restart.

Using BIOS alone, I can observe the following results (all at idle of course):
Silent = 508 RPM
Standard = 969 RPM
Turbo = 958 RPM
Disabled = 1068 RPM

I see no other options that would allow me for example to directly input a fan % or fan RPM.

From Windows, I can run ASUS's Fan Xpert. It offers these same four profiles plus now also "Intelligent" and "Stable" and also allows the user to directly create their own profile.

Fan Xpert has a GUI which displays the intended temperature to fan speed mapping for each of these profiles. So by looking at this I can see that, at idle temp:
Silent maps idle temp to 20%
Standard maps idle temp to 40%
Turbo maps idle temp to 66%
[Disabled just greys out the chart, but I believe is meant to apply a flat 100%]

When I apply one of the 4 basic profiles in Fan Xpert, it is reflected in the BIOS setting and vice versa. So I think both BIOS and Fan Xpert are ultimately changing the same setting. (Although the common statement I see on the net is that the two are "incompatible.")

Fan Xpert also has the "calibration" function where it will report the RPM fan speed for 20%, 30%, 40%, 50%, etc up to 100%. In this I still the same bell curve centered at 50%.

I looked briefly at SpeedFan and didn't see anything there that caught my eye as providing different results from BIOS or Xpert, but it wasn't an extensive trial.

 

[brucek2]

A few other quick notes:

1) I'm tempted to go get one of those basic electricity measuring devices (multi-meter?) and see if I can watch directly what voltages are being sent to the header pins.


2) I found this interesting rant by one of the Tom's Hardware editors:

from http://www.tomshardware.com/forum/259220-30-speed-control-working

"I've had NO LUCK in using Asus' P6T fan speed controls. Automatic controls for the CPU fan seem to do whatever it is they're programed to do, but changes don't seem to work. And the changes available are CRAP.

In fact, ASUS P6T FAN SPEED CONTROL IS GARBAGE and the company deserves a public flogging for putting such crap into a retail product."

This is disputed by other forum members but Crashman is not an inexperienced guy. Maybe just some of the boards are bad?


3) I just got this response from ASUS tech support:

"If your fan controller cannot get the fan past 1250 RPM, I would recommend contacting Cooler Master for a replacement fan as the fan motor may be dying."


4) I haven't really dug into the case fan situation yet. From the fan xpert view they look a lot like the cpu fan with the same bell shaped curve. Its a little ridiculous in that there's 3 fans attached to 3 headers but there's only one profile setting and calibration result, I guess maybe its an an average? The 3 fans aren't all the same size and I don't even know their RPM spec but I bet its not the same for all of them. There are other places where I can at least see the individual RPMs of each fan, but so far I have found no individual adjustments.

 

[SOFTengCOMPelec]

It does seem to be pointing to a faulty motherboard. Sometimes issues like this are cleared up by resetting the cmos, but I have the strong feeling it WON'T help here.

I was therefore going to suggest (which was also mentioned in the thread you just linked to), an independent fan speed controller. Such as "akasa 3-fan controller" etc.

I don't know much about them, so I will leave it for other members, who may have a lot more experience with them, for recommendations.

If you get a cheap multitester, it is probably best to concentrate on the simpler 3 pin fan headers, and see if the 0v (ground) and +12V lines (pinouts readily available on the internet) are present and ok. The third pin (sensing) is harder to test, but comes from the fans (I think) anyway, to let the motherboard know what the RPM is.
There could be test guides on the internet, I don't know.
I would set the mobo to full 100% fan on before testing (or the nearest equivalent setting).

 

[Markfw]

or put a aftermarket cpu fan on the 4 pin PSU connecters with an adaptor, and leave the cpu fan plugged into the cpu header just moving air around.

 

[brucek2]

This all started as a practical problem-solving exercise for me and has morphed into a bit of "edutainmnet" as Idontcare so well put it.

On the practical side, I think I've reached (or long passed) the point of diminishing returns. Between the return to stock, the aligning all the fans so they work together, moving the system so its less enclosed, de-dusting, and installing a fan profile that keeps the fan at ~1300rpm vs ~1000rpm before, I'm now pretty convinced my regular workload is not going to get anywhere near an over temp situation. I actually just shaved another 4c from prime95 peak (so down to 67c now) by disabling hyper threading, although not sure if that has any practical implications for real work or not.

Bottom line I think I'll be OK for at least a little while, hopefully long enough to reach a new chip or chipset introduction that makes me excited to build a whole new system around it.

On the edutainment side, I do still feel that I'm not getting all the cooling and therefore all the performance I could be. A new cpu fan plugged directly into a psu might produce more airflow although I'd lose both the rpm readout and the temp/speed profile that is part of the investigative experience. I looked into several fan controllers and there's potential issues there around some effectively converting you to a full manual experience (which I don't want), others with temperature sensors which I fear would be less effective than the ones in the motherboard (how am I going to stick one into the cpu?), and a fair amount of FUD in user comments which I'm sure I could eventually burn through but ran out of motivation.

If only I had a buddy in an electronics lab somewhere I'd love to have him send known volts / watts / whatever into my existing cpu fan and tell me what it took to get it to hit its rated 1800rpm speed, and then I could work from there to find out if my mobo was supposed to be able to do that and if so why it isn't. Hmm, maybe a project for an intern if only one was available for this purpose ;-)

For now I may let it rest until there's a new system to build. But I'm very glad for the learning experience and I will sure pay a lot more attention to all this then than I did while building the current system 4.5 years ago.

Thanks again guys and especially SOFTengCOMPelec for helping talk me through all this, I feel a lot better about my system than I did 48 hours ago.

 

[SOFTengCOMPelec]

I'm glad it has turned out well for you.
The extra experience you have just gained, may indeed (as you said) improve your next computer, and make it a smoother, easier process.
At least you have a better idea what is wrong, if it gets worse.
Anyway, good luck!