[SOLVED] Cooling best practices with a fan curve and general case cooling

Bobbrugge

Honorable
May 2, 2014
82
4
10,535
I upgraded from a hyper evo 212 to an EVGA CLC280, which is a 280mm aio. The software for the RGB and fan control is, of course, busted. I have been using speed fan and am trying to figure out a fan curve. Default is 30% @ 30C up to 100% @ 60C.
I want it quiet until its time to do stuff. When I am not doing stuff, I am streaming tv.
So 15% ish CPU low load and 70% ish CPU high load. This AIO keeps me 20C cooler than the 212 did with everything on 100%, including load, which is awesome! But...
Can I run my CPU too cool? What is a good temp range to shoot for under heavier load and at low load?
I have been screwing with the fan curves and don't like anything better really than setting everything to 25% for low load applications, it gives me temps around 34C. I can drop my temps by 1C to 33C with 100% fans and pump and with 100% load, which is probably ok, but I am just not sure.
Keeping it cool isn't an issue anymore, but how cool? AMD recommends 62C as max. I have had my FX 6350 up to 79C with no thermal throttling, but only for a few minutes while testing. at any rate, overheating seems like it may be impossible with my 4.6Ghz O.C.
Now as far as case fans go, I have a bunch. 2 140mm front intakes, bottom and middle, except the bottom one is mostly blocked by the hdd cage.
I have a 120mm on the bottom of the case as intake.
2 x 140mm pulling air through the radiator and into the case from the top.
I have a 140mm fan blowing down on all of my VRM'S piggybacked off of the pump. It works sweet, when the pump ramps up, so does the vrm fan.
I have a 140mm blowing fresh air on the back of the cpu socket
and a 140mm rear top exhaust.
I am trying to create a high pressure case to fight dust.
Everything is quiet, but is it as optimized as it can be? Thats question 2.
Last question: Fan headers on the mother board. ASUS to be more precise. Using Speedfan, I can't control the PWM fan headers independently. That kinda sucks. Anyone know how? The water pump and cpu fan headers are PWM and I can control them just fine...
TIA!
 
Last edited:
Solution
Lol, it's already tomorrow, and yes I plan on it. 😊

The FX aren't Ryzens either. Too primitive for that. They aren't smart cpus that boost according to temp and voltage. To use a car analogy, they've got carburetors. They run idle at @ 5% cpu±, base speeds at somewhere around 25% or less cpu usage, but run flat out turbo boost at anything higher. In other words, they idle at idle, run secondaries for anything less than moving the mouse, but anything regarding stepping on the gas involves opening up all 4 barrels.

But you are correct about ranges. For a TM, it'll be in the 30ish to 10ish range, which is pretty much normal temps. It's only more extreme temps, such as right below 62°C or under 0°C where those cpus suffer. So if you can...
Yes you can, if your idle temp is 30c or 50c in low load it does does not matter, under load you want to keep it under 85-90c to stop thermal throttle. As long it is under that sort of threshold it is up to you what fan noise you can live with. I am someone that likes to keep my PC at around 55-60c while gaming but i wear haedphones so fan noise can be higher and not bother me.
 

Karadjgne

Titan
Ambassador
You have an FX cpu. Whatever temps you are reading, by whatever means, are wrong.

The FX cpus did not have thermal sensors in/on the cores. Don't ask me (or anyone else) why they adopted such a hair brained thing. The closest you can get is package temp (often called cpu temp) which is the temp of the IHS where it meets the cooler base. It's usually @ 10°C (ish) higher than actual core temps according to the engineering samples that had core sensors MacGyver'd into them.

The only programs able to accurately asses temps on an FX are AMD Overdrive and CoreTemp using Thermal Margins.

TM are backwards. The software uses complex algorithms made up of core usage, voltages, socket/package temps etc to arrive at a number. That number represents how much thermal room is left in the cpu, not an actual temp reading. A TM of 0 means you have 0 room left. A TM of 40 is great, all kinds of room. Anything under 10 and you are cooking, in the mid teens to mid 20's is about normal for loads.

It's similar to an oven, for the actual oven (Intel) you'll see °C numbers for the heat, but for the stove top (amd fx) all you get is warm-high with no actual numbers. And it starts at high and goes down in temp.

So you'll set SpeedFan according to what Overdrive and your ears tells you is decent, and don't worry about particular numbers, they will change according to multiple factors (time of day, how much coffee you've had, what cheek you are leaning on more) what you need to concentrate on is the Range the number represents. Anything in the 20's or better under a load is great, closer to 0 you get isn't so great. You'll set the fan speeds according to what you think you can get away with/live with.

The FX cpus are well documented for doing impossibly stupid stuff like idle of 8°C and loads of 18°C and overheating, having owners going nuts because SpeedFan or other software doesn't hit max fans until @ 70°C. FX cpus simply do not work with conventional Intel based temp monitoring software.
 
Last edited:

Bobbrugge

Honorable
May 2, 2014
82
4
10,535
I know about distance to TJMax.
I have used both amd overdrive and coretemp.
I can't figure something out and someone fairly well versed on FX cpus from this very forum either didn't understand, or didn't believe me.
Here is my hang up...
Speedfan says I am @ 50C
HWinfo says I am at 50C
CoreTemp with the Distance to TJMax set shows 40C with a max temp of 90C
AMD overdrive shows 20C...to TjMax. So 3 programs with the best monitoring agree on a temp, 50C is just an example, and Overdrive says something 20 * different.
What ever the thermal margin is moot as is the actual temp, or so it seems, with this level of cooling.
So if I understand right, no, it doesn't matter what the temp is as long as it isn't cooking. At this point it is about noise at load and what I prefer my operating temps to be?
There is no best practice for operating temp?
 

Karadjgne

Titan
Ambassador
With Overdrive it's a variable amount, not a temp as such, it's a number representing amount of space left, something closer to a % amount. That will change according to everything and anything. So it's not a finite number like a temp reading. It's like you lifting a heavy weight, you know exactly whether adding a little more would make the lift beyond your ability or if taking a little off would make it far easier, but the exact amount, be it 100lbs or 200lbs is a moot number. Handholds, size, awkwardness all change it.

So you'd set your fan curves according to your ears. The curve is what's important, not the actual number. If SpeedFan says the curve starts at 30°C and has a slow rise to 60°C and then a sharp rise to 70°C, so what. You verify with Overdrive that the 70°C that SpeedFan says is actually a viable TM, maybe 10. This way you know you are within the limits of the cpu, not within the limits of what a particular °C is. The curve could be the exact same, but have a range of 10°C to 40°C in SpeedFan, that doesn't matter, it's whatever the curve is, not the °C SpeedFan sets it as.

SpeedFan also reports 2 temps (as does HWInfo and Hwmonitor) on my old Z77 mobo as physically impossible. 255°C and -125°C (yes that's a negative number) so software as such isn't totally reliable or gospel.
 
Last edited:

Bobbrugge

Honorable
May 2, 2014
82
4
10,535
With Overdrive it's a variable amount, not a temp as such, it's a number representing amount of space left, something closer to a % amount. That will change according to everything and anything. So it's not a finite number like a temp reading.
I get that.
The curve is what's important, not the actual number.
I don't understand this, can you explain please? I think this is what I was asking...
You verify with Overdrive that the 70°C that SpeedFan says is actually a viable TM, maybe 10.
I did this, or so I thought by using 4 different programs to monitor and verify temps.
This way you know you are within the limits of the cpu, not within the limits of what a particular °C is. , it's whatever the curve is, not the °C SpeedFan sets it as.
I get this part of this too. It's the curve part that keeps loosing me. It seems like you are saying set the low and high temps and forget it?
SpeedFan also reports 2 temps (as does HWInfo and Hwmonitor) on my old Z77 mobo as physically impossible. 255°C and -125°C (yes that's a negative number) so software as such isn't totally reliable or gospel.
It does this for me too on one of my temperature probes. I know it is an impossible reading, so I don't even display it.
 

Karadjgne

Titan
Ambassador


Ok. This is something like the SpeedFan graph. If you look at the left numbers, they'd be the temp, the bottom the fan duty cycle. (yeah, I know, bad graph).

The temp numbers are set so at 20 the fan speed is 1000rpm. What I'm trying to say is you should transpose those in your head to TM values. So with a TM of 40, you'd be @ 400rpm, a TM of 20/1500 etc. Because whatever SpeedFan says the temp is is wrong. SpeedFan (or the rest) could say the temp is 50°C, but actual core temp could be 30°C or 60°C.

So you move the line on the graph according to what your ears and TM tells you. So when you start hitting a TM of @ 10, your fans should be hitting max speed. If SF says you are at 55 or 95 doesn't matter, you have a TM of 10.

That sets up the curve to respond to the TM values from Overdrive, using the °C values in SF. So you might have 400rpm at 40, 600rpm at 30, 800rpm at 20, 1200rpm at 10+. SpeedFan might show that as 40°, 50°, 60°, 70°± or for some FX users it's worse, 8°, 16°, 24°, 32°+. It's not the number that's important, but the fact the number represents cool, warm, hot, omg I'm cooking.
 

Bobbrugge

Honorable
May 2, 2014
82
4
10,535
I get what you are saying now.
Thats basically what I have been doing.
Assuming the only program that disagrees with all of the others is the only program that is actually reading right, I am able to transpose those figures. 60C on SF is 2C to TJMax in overdrive.
Basically right down the line.
You are saying I should set my fan curve to Distance to TJMax instead of the incorrect reading that SF offers as the FX probe is different than Intel.
Got it.
I am saying that overdrive tells me I have 0 to TJ max when SF reads 65C ish. Maybe I got lucky. So Since AMD says 62C is the hottest you can get the chip safely, I should be safe to think that what SF says is 62C is actually a couple of degrees below what overdrive says TJMax is. So at 62C on SF I am actually 1.5 or 2C below TJMax. Kind of a safety net.
By further setting the fan curve to 100% at 60C according to SF gives me another degree and a half or so from TJMax.
Unless the readings drastically change from air cooled to water cooled.
So bottom line is I get that SF is wrong. I get FX does it differently.
SF is really close, close enough for me in fact, that I am confident that setting the top of my curve is going to give me 100% at 60C and be totally cool as it will be about 8C below TJMax. I believe this because I can't get this over 40C at 100% load. Not at my current settings anyway.
Which leads me back to my first question, ie: What is the best operating temp at any given load? If it is just about acoustics, I could set everything at 100% keep idle temps under 10C and barely hear squat. I replaced the crap fans they send with some fans that cost me less than $7 each. I am down to about 20 db with fans and pump on 100% in a closed case.
But if I can use a car analogy, my old truck has a 318 in it. 195F is the normal operating temperature. Put a 180F thermostat in it and you get no heat in the winter. Put a 200F thermostat and it overheats and vapor locks.
They designed it to have a certain operating temperature where everything just jells. Is there a certain temp for my FX?
I appreciate your time and responses, call your mom tomorrow.
 
Last edited:

Karadjgne

Titan
Ambassador
Lol, it's already tomorrow, and yes I plan on it. 😊

The FX aren't Ryzens either. Too primitive for that. They aren't smart cpus that boost according to temp and voltage. To use a car analogy, they've got carburetors. They run idle at @ 5% cpu±, base speeds at somewhere around 25% or less cpu usage, but run flat out turbo boost at anything higher. In other words, they idle at idle, run secondaries for anything less than moving the mouse, but anything regarding stepping on the gas involves opening up all 4 barrels.

But you are correct about ranges. For a TM, it'll be in the 30ish to 10ish range, which is pretty much normal temps. It's only more extreme temps, such as right below 62°C or under 0°C where those cpus suffer. So if you can stay within SF 30-55ish, you'll have no worries but even if something pushes SF 60, you'll be ok for short periods of time because 62°C as measured by amd is when the cpu is reaching thermal throttle, not when it's reaching damage temps. That's somewhere in the 80's or if looking at cpu package temps, somewhere above 90ish.
 
Last edited:
Solution

Bobbrugge

Honorable
May 2, 2014
82
4
10,535
Lol, it's already tomorrow, and yes I plan on it. 😊

The FX aren't Ryzens either. Too primitive for that. They aren't smart cpus that boost according to temp and voltage. To use a car analogy, they've got carburetors. They run idle at @ 5% cpu±, base speeds at somewhere around 25% or less cpu usage, but run flat out turbo boost at anything higher. In other words, they idle at idle, run secondaries for anything less than moving the mouse, but anything regarding stepping on the gas involves opening up all 4 barrels.

But you are correct about ranges. For a TM, it'll be in the 30ish to 10ish range, which is pretty much normal temps. It's only more extreme temps, such as right below 62°C or under 0°C where those cpus suffer. So if you can stay within SF 30-55ish, you'll have no worries but even if something pushes SF 60, you'll be ok for short periods of time because 62°C as measured by amd is when the cpu is reaching thermal throttle, not when it's reaching damage temps. That's somewhere in the 80's or if looking at cpu package temps, somewhere above 90ish.
So it is a racing carb, sweet.