Computer cooling philosophy/technique.

Bluefront · Post by **Bluefront** » Sun Nov 07, 2004 12:18 pm

I'll limit this to CPU cooling, since it's the most critical temp in most setups.... and since exact temps are up to the individual, they don't matter for the sake of this discussion. I've tried different methods, but haven't really decided which suits me best. Here's a few different techniques.

1...Don't worry about it...concern myself with noise only.

2...Set fan speed manually, to a point I think gives a safe temp wide open.

3...Use temp controlled fans, hope they speed up enough.

4...Use manual fan control, adjust to faster speeds with heavy usage.

5...Regulate fan speeds with some sort of automatic control, either by software or a hardware controller.

6... Run fans slow as possible and have some sort of manual/automatic fan control that turns on more fans with high temps.

There's probably a few methods I missed (water cooling aside). They all have their good points. I like #3 but it's hard to find the best fan to do this. #6 would/does work well, but can be difficult, maybe costly to setup. #5 is nice also, but the right controller to do this is hard to find and setup properly.

So how do you do it?

Tigr · Post by **Tigr** » Sun Nov 07, 2004 1:17 pm

As for my personal favorite, I prefer to have no fan controllers. Why bother? If you have a fan controller, you probably live sometimes with higher noise and sometimes - with lower. And some time, sitting in a movie, you'll remember that you did not switch off your PC and you did not turn up the fans, so you'll fret about your computer burning up.

I prefer a system that will keep it quiet and remain within the range of acceptable temperatures no matter what I do.

This may sound like the option number 2 but this is even different. You don't use any fan controllers, you run all your fans straight from 12 or 5 volts to avoid the desire to fiddle with them. Makes your life so much more ... well, difficult

sthayashi · Post by **sthayashi** » Sun Nov 07, 2004 1:34 pm

I subscribe to #1 & 2, and I need to learn to move completely back to number 1.

Tigr, I prefer fan controllers, because I get more of an option to play with the settings. If it turns out that 5v is too low, I can always turn it up. No rewiring to deal with.

fanerman91 · Post by **fanerman91** » Sun Nov 07, 2004 1:41 pm

I tend to use the first 3. I use the fan control to lower the fan's speeds to well, the minimum, except for the CPU Panaflo, which I can't hear through the case and through my desk, so it's at 12V. After the fan's speeds are at minimums, I just forget about them, never adjusting them after that. My Seasonic Power Supply has an automatic control so I never worry about that.

wim · Post by **wim** » Sun Nov 07, 2004 5:17 pm

#1 here..
i use amd processors almost exclusively so i don't worry about thermal throttling, and i've upgraded cpu frequently enough that any shortening of lifetime has never been relevant. as long as it's p95 stable, i don't care what the temps are.

Post by **Rusty075** » Sun Nov 07, 2004 7:07 pm

My take:

You really can't literally just do #1. There's always some concern over the temp as well, if not you would just unplug the fans altogether and just accept the fact that you'll only be able to use your machine for 5 minutes before it shuts down.

#2 is what I consider to be the simplest and easiest method to achieve the goal. You can usually get a machine to be quiet enough that way that the other methods are almost academic. (if I can have a machine be below the ambient dBa at full load, who cares if it's imperceptibly quieter at idle?)

#3. I like it in concept, and tried it for a long time, but never achieved the results needed. There's just too much variability in application for a factory preset temp curve to work properly for each situation. You need a way to program your own voltage/temp curves, and I haven't found a way to do that....yet.

#4. Nah. What happens if your machine decides to lock itself at 100% load while you're gone?

#5. Software...no. Crashes happen, regardless of your OS. I wouldn't want to trust my baby to a piece of software written by the lowest-bidder outsourced programmer.

(that's a joke, btw) Hardware, maybe, but see #2 above.

#6 Is just an advanced form of #5, right?

Post by **MikeC** » Sun Nov 07, 2004 7:59 pm

Focusing only on the methods I use myself...

2) Fixed fan speed: As others have already said, this is the simplest method. You can usually get a CPU HSF, video card cooler, PSU & case fans in a good case, set for very low noise & airflow, that will give perfectly safe temps over a wide range of applications. The trick is to try and give yourself hot weather headroom. IE, if you set up in winter when your room temp never exceeds 20C, then make sure there is 10C headroom for summer peaks at 30C. It's usually the best choice if building a system for someone else.

4) Manual fan control: Basically lets you tweak the above more easily when necessary. It's what I use in 90% of my own systems either with fanmates or multi-fan voltage controllers.

5) Automatic fan control: Actually, I like using the better on-board hardware controllers at fixed speed setting when available. Not thermal controlled, but a fixed speed, as with AOpen's SilentBIOS. It is unaffected by vagaries of OS (windows) stability & easily controlled.

6) Automated thermal fan control: In the middle of investigating how well this works -- with the mCubed T-Balancer. I think this will work quite well for a super powerful system to run very quietly most of the time and ramp up / down precisely as needed in accordance with specific temperatures at selected measurement points. The full review to come...

Post by **Ralf Hutter** » Mon Nov 08, 2004 7:05 am

#2 for all the systems I build. It's the simplest, most trouble-free way to go. I'm a firm believer in the "KISS" principle.

I just set the cooling to be sufficient under the worst possible conditions (ambient at 95+°F, full CPU load) and leave it there. That way the system is guaranteed to run within a safe range at anything less than that.

burcakb · Post by **burcakb** » Mon Nov 08, 2004 9:00 am

Same here as Ralf.

And since I run folding all the time, for me, there's no distinction between light & heavy load. It'd better be quiet & cool all the time.

Bluefront · Post by **Bluefront** » Mon Nov 08, 2004 4:58 pm

Yeah I agree that #2 is the easiest to setup, and probably the best for someone else's computer. But it certainly is not the quietest. It insures your computer will be noisier than necessary a large part of the time.

I'm looking forward to the second generation T-Balancer fan control unit, and more info about the thing...it looks very promising.

ddrueding1 · Post by **ddrueding1** » Mon Nov 08, 2004 6:19 pm

I use some funky combination of 2, 5/6, and 1.

The only fan that runs 24/7 is the 120mm in the PS. It is manually controlled via a fanmate and I use the temp of my HDD to determine the appropriate speed. The only other fan in the system is on the CPU, and is off 90% of the time (controlled by the BIOS of my ASUS board). As to video, northbridge, and other hotspots? Who cares? Their failure is highly unlikely to affect my data.

josephclemente · Post by **josephclemente** » Mon Nov 08, 2004 8:34 pm

I like #2 as well (although I use fanmates, would that be #4? I leave them set at my desired setting and leave them alone) My computer noise stays the same no matter what it is doing. While the PSU may be thermally controlled, it won't ramp up in any ambient temperature I've subjected it to.

If I had more control over my fans (automatic slight increases in speed as needed without PWM noise), I probably wouldn't have made my system as quiet as I have for full-load. In the end, I'm glad I didn't have any help from smarter fan control - my system sounds great all the time.

wim · Post by **wim** » Mon Nov 08, 2004 8:40 pm

Rusty075 wrote:My take:

You really can't literally just do #1. There's always some concern over the temp as well...

hrm.. i think you can, literally, but maybe we interpreted #1 differently. what i like is #1 with a prerequisite that system is stable - it doesn't matter whether the stable temp is 300 K or 3000 K. temp and stability are correlated, that's for sure, but I'm only concerned for stability - not whatever number the mobo gives as temp.

the problem with constraining yourself to a target temp: unless you know the system very well, you could be chopping off an interval of safe headroom there. not to mention the temp reporting from board to board is probably not accurate/consistent to nearest degree.

if you can get below ambient with 2 it's a moot point. and for building other peoples systems where you don't have a chance to stress test stability in a variety of conditions/environments and worst case scenario, #1 is stupid. but for your own system - why do we worry about temps? because temps correlate with stability. so why not just cut out the middleman and worry about stability?

Schroinx · Post by **Schroinx** » Tue Nov 09, 2004 3:40 am

As of now I use number 2 and 5 for my two machines. The older one does not support Speedfan, and thats why it is set up by the numer 2 philosofy. My workstation/gameing machine is setup with speedfan, so when having a mild load, it runs fairly silent. When I need the speed, the fans ramp up, making sure that even under heavy load and with some 3D going on, it won't fry. Since it is an Intel cpu I don't have to worry about frying the cpu, but the vidcard is still an issue.

1) Well, that'll fry your PC or lower it's lifespan.

2) If executed properly, this is a good solution, because when the machine runs at full speed, it will be cooled adequately, I assume. However if your machine cannot be silentliy cooled at max load, you have to live with the noise all the time.

3) I have no confidence in the solutions I have seen. Also there is difference in when the CPU should ramp up, and the casefans.

4) So I forget to alter the speed, or some process goes wrong and run at full load when I'm not at home, leaving me a fried pc. Not a good solution.

5) This should be split into two. Software control and hardware control. The software method is ok, but unstable, as windows is unstable. If you have a cpu without thermal protection, I would not recommend it. The hardware besed control is the best, as this run independently from the system it cools. You get the benefits from a silent running system and still have a lot of headroom to keep it cool when running at full load. I believe this is the best way, but then again, I'm an engineer, so I like complicated solutions. However I haven't found any controller that does what I want it to do, and haven't had the time to build one myself.

6) If you have a fan in your system thats not running, it will alter the airflow. Two fans beside each other, and only one is running, air from the running fan will go out (or in) trough the not-running fan, creating poor airflow. When running with a minimum of fans, one need good airflow, so this does not really sound like it will work.

The best cooling solution for a computer I have seen is the one made for the G5. Here you have different channels for each of the main heat producers, and fan control.

When talking about heat, there is one issue to remeber. Most of the power used in a PC is turned into heat in your box. First you have to remove the heat from the spots where most of it is generated, like CPU, GPU, PSU and HD's. Then you have to tranfer this heat out of the box. The amount of heat is between 100 and 250 watts, which is significant. However many power amps can generate as much heat and still be cooled passively.

In the end what this boils down to, is how much noise are you ready to accept, both running full load, and idle, and how hot will you let your system run? Heat shortens some components lifetime, as HD's.

My 5c...
/Schroinx

Bluefront · Post by **Bluefront** » Tue Nov 09, 2004 4:25 am

Schroinx....as far as #3 is concerned, I have successfully used Thermal-controlled "case" fans directly on the CPU. With proper placement of a remote sensor, and a suitable heat/rpm curve, this setup works well....but like Rusty said, it's hard to find a fan that meets your own requirements.

And #6....How about a "blow hole" fan on the very top of the case, that only turns on at high temps. Shouldn't affect airflow very much, and would be silent most of the time.

And how about a tower type heatsink that uses two fans....the second fan only turns on at high temps. I've used this setup before and it works good, but you need a specific MB that can to do this or an external control like a DigitalDoc 5. I like this technique.....

Schroinx · Post by **Schroinx** » Tue Nov 09, 2004 5:31 am

just a short comment. I did not say that these solutions won't work. just that they are not the most optimal.

#3 If the fanspeed is based on an temp sensor on the HS, the fan will first respond when the temp of the heatsink rises. The temp of the core can rise much faster than the HS, and therefore the tempprobe, this introduce a delay in the reaction of the cooling system. Most often this will not be an issue, but the solutions is not optimal. Also since price is an issue, the tempprobes used is of the NTC type. These analog probes comes in various qualities, but in generel they are not that precise. So all in all, I don't think this is a good solution. If you have a CPU/System with low heat dissapation (sry for the spelling) then it does not mean much, but I would be worried with OC'ed Prescotss and AMD64 FX cpus.

#6 Still you will have air coming in or out of the top vent, when the fan does not spin, making a difference in airflow. For the latter HS question, the second fan is restricting airflow, which in turn means that the other fan have to run at a higher speed, than without a second fan attached. Running both at 40% than one at let's say 60%, will be more efficient.

I'm not saying it can't be done, just that it is not the optimal solution. I mean we're discussing cooling concepts. I have seen people run their PC in a bowl with oil, or using nitrogene, so I assume almost everything can be done. Question is, what is the best solution?

Also you state in the OP that it is about CPU alone. This is a wrong way of thinking. In a PC you basically have two systems. Spot cooling that removes heat from CPU/GPU and so on, to the air inside the case. Then you have a system that removes heat from the inside case. Since you cannot change the heat output from the components, the first system have to cool the components adequately depending on how much heat the produce at any given time. The efficiency of this system in greatly impacted by the heat already in the case, and therfore the efficiency of the second system. The difference in temp from the HS to the air inside the case is the single most important factor, for the efficency of the system as a whole. In other words, if you have a system with good airfolw trough the case, then spotcooling can be held at a minimum, and if the airflow is poor, the spotcooling have to work harder.

Since we're talking about philosofy here, there is one other solution. The mobo could have a fan-controller onboard, being set up in bios, or software, but able to function without software control, and with like 5 or 6 channels for fans, and able to read temps of GPU, CPU, ambient case, PSU and HD's. That would give endless possibillities on how to set it up, and also make it possible for the system to test it self, and find the optimal values, if your not interested in doing so yourself. Well, it could also be a USB card with a microprocessor. The springing point is getting the temp readings from the CPU diode and GPU temp sensor and pass these on to the fancontroller. Speedfan have most of the nessacary features, but you cannot set how the fan ramps up and at what temp is should run at max, so the noisecharateristics is not the best. And it is softwarebased, so if soemthing goes wrong, it might fry the computer.

Edit: The T-balancer seems to have most of these features, but lacking the possibility of getting temps directly from SMART-data, PSU, CPU and GPU diode. The thing that often sucks is that the PSU is acting as both spot cooling and case cooling, evidently influencing both systems. PSU's should be made with an SMBus interface to the mobo for temp and fan control, and so should GPU's. In an optimal world that is.

I hope it makes sense. It is difficult to explain when english is not my native language. Oh, and this got a bit longer than I promised.

/Schroinx

ddrueding1 · Post by **ddrueding1** » Sat Nov 13, 2004 5:27 pm

My philosophy is pretty straightforward when I use the terminology Schroinx used.

"Case Ventilation": Use just enough ventilation to keep the case air near ambient temps. Optomise case design to produce a very slow airflow through most of the case.
"Spot Cooling": Install a large enough heatsink/pipe to allow the slow case airflow to cool hotspots sufficiently. This may require relocating components into the airflow (hard drives), mounting exessively large heatsinks on components that are out of the airflow (Zalman on an ATI 9200), or using a custom heatpipe solution to move the heat into the airflow (currently working on something for my NB).

Computer cooling philosophy/technique.

Computer cooling philosophy/technique.

Re: Computer cooling philosophy/technique.

Re: Computer cooling philosophy/technique.