Set Custom Fan Curve to H110i or CPU Package?

Hi All,

 

I run a 6700K @ 4.4 GHz (1.28V) - ROG Maximus VIII Hero.

 

After 4 hours of stress testing (100% CPU load) Corsair Link reported the following temperatures (ambient room temperature 30c):

 

CPU Package: 76c

H110i (coolant): 38c

 

My default fans are particularly noisy and I have a custom fan curve set to respond to the CPU package temperatures as follows: 40-59c = 700rpm, 60-74c = 850 rpm, 75-79c = 1000 rpm and 80c and above = 1500rpm.

 

With these settings the fans do not ramp up and down and are essentially inaudible (quieter than Quiet Mode). However, should I rather set them to respond to the coolant temperature?

Yes, technically the fans should be set to coolant temperature (H110i Temp) and that is also how the device physically works. Fan speed doesn’t do anything for the temperature in the CPU socket. If you are happy with your fan behavior, then fine. The worst that can happen is they constantly ramp up and down at the desktop level. However, one drawback maybe what happens when iCUE/Link is not running. The H110i automatically saves that curve to the device, but cannot use CPU temp without active software. It will interpret it as coolant temp and fan speeds will be very low. Not really a problem, unless you try to run a stress test like Aida64 or something else with iCUE not running.

Thank you for your explanation.

 

I'm going to change the group to H110i Temp today and run Realbench 2.56 again to see what difference it makes.

 

I'm very new to AIO liquid cooling - can you please explain the following statement to me.

 

Fan speed doesn’t do anything for the temperature in the CPU socket.

While it may vary a little between CPU models and even the program giving you the information, CPU Package temp is a specific temperature sensor, usually on the socket side of the CPU. So while this is important in regard to maximum safe temperatures, the actual package temperature is a result of the voltage applied at the pins and the CPU material design. More Vcore, more heat. No cooler can work around that and its why extreme overclockers cool the entire socket area with liquid Nitrogen rather than use 10ft tall radiators. This has a been a popular question lately and I am going to snip one of my earlier responses.

 

Your CPU temp is the result of the electricity in the socket and the CPU load and materials, less the amount of heat conducted away through the cold plate. The fans do not directly factor into this at all. The radiator and liquid loop is really the waste disposal part of the cooling system. Pick up waste heat at the cold plate, dump it at the radiator. Now the fans come into use. They aid in the rate of heat dissipation from the radiator. Obviously this only effects the coolant temperature and not the CPU directly. You put 1.60 Vcore on your CPU and no cooler can save you, no matter how many fans or how large the radiator. All of heat must be conducted through the CPU to get rid of it. You can't bypass and that is how the coolant temp fits back into CPU temperature. The conductor pathway between coolant stream and CPU across the cold plate is a two way street. Heat goes both ways, so while it will conduct out, it conducts back to the CPU. Your coolant temperature (H110i Temp) is effectively the lowest possible CPU temperature at idle (or theoretically with zero voltage). Coolant temperature goes up +5C. That makes the baseline CPU temp +5C and everything shifts upward by that value. It also means fan speed does not have a huge effect on end CPU temperature wit the larger 280mm and 360mm coolers. If the coolant change is only 4C, then that is the most you can reduce CPU temp at any fan speed.

 

So by running accordance with a very dynamic variable like CPU temperature, the fans will change speed frequently. That isn't necessary in a water cooling system. Slow to moderate fan speed will effectively "dump the heat elsewhere". You only get into trouble when you don't take out the trash and the heat builds up. Still, with a 280mm cooler you are never going to overheat even at minimal speeds. It is effective enough all in its own to prevent this and the water has a decent capacity to hold heat, unlike a small fan blower. So for your system I would expect coolant to go up about +6C or so at 100% with the fans in the 1000-1200 rpm range. If you were to deliberately cut them down to a fixed 500, the coolant might go up +12C, but that is only a 6C difference so you would have to already be in big CPU temp temp trouble (85C+) for this to get you into the throttle zone. That means the only people who really need to be a little aggressive with their fan curves are those running right against their preferred CPU temp limit already. So if 80C is your limit and you are constantly in the 75-80C range at 1.xx Vcore, then you either have to lower Vcore or be aggressive with the fans. In that circumstance, the 1-3C difference may be of value. On the other hand, my delided 8700K peaks out at 55-60C in Summer, so I don't really care if my coolant goes up +2C today since that would only shift my peak CPU temps up by +2C.

 

Now that all sounds simple enough, but there are two big monkey wrenches that get thrown into the works.

 

1) Ambient Room/Case Temperature - If you have a room on the sunny side of the building, you may experience definite changes in room temperature throughout the day. No matter what I do with the air conditioning, my office is going from 22C to 28C in the afternoon sun as the building bakes in the Summer. That +6C shift is enough to throw my curve out of balance. Now I am content to leave it set for the afternoon temp scale and that results in lower morning speeds, but there are circumstances where this might be annoying. This would still be a problem if using CPU temp since the +6C is applied to everything in the room, but since your typical load delta is also only +6C, it feels more dramatic on coolant curves that may only have a +-10C range.

 

2) GPU waste heat - It is not uncommon for people to see their highest coolant temperature when gaming or doing CPU/GPU combo renders. While your CPU only load might be +6C at 100%, you may see +10C when gaming with a high CPU/GPU usage game. This is not because the cooler is less efficient, but the GPU is heating up the inside of the case and everything else as well. Same problem as above, but internal to the case. Most people do not sit around running prime95 all day, so it doesn't make much sense to set your curves for a stress test environment (you should probably use fixed speeds when doing that anyway).

 

You should set your curves for what you actually do and do it from a noise perspective. Figure out your typical idle/desktop coolant temperature and set that to a comfortable low fan speed (probably 500-600rpm or whatever you like). About 3-4C further down the line, start to increase the speed. Take note of your typical use peak coolant value during real use. If it's +6C, then set a tolerable medium speed to that position (650-900, whatever). Save the 1500+ rpm blast for around the 45-50C mark. You should never get there under normal conditions. This will provide an audible alarm if you cooler is out of its normal range and you should stop and investigate. Unless you are right against your thermal threshold, you don't need more than 1000 rpm for anything. I do my benchmarks at 1250 rpm on a 280mm cooler. Anything faster I wouldn't use and has no value to me (and is LOUD). You can make your own determinations about what is loud or not, but also note the differences in temperature between these speeds. It is typically hard to see and may only be +-1C change. That also means it is only +-1C at the CPU level as well. Choose for listening comfort.

Thank you for your very detailed response, I really appreciate it.

 

I changed my group to H110i and set my fans at a fixed rpm of 700. My particular set of fans start to whine at anything above 700 rpm. I repeated the Realbench 2.56 stress test for 4 hours. The ambient temperature was 1 degree lower at 29c. LINK reported the following peak results:

 

CPU Package: 71

H110i: 37

 

Thus my package temperature went down by 5c and my coolant temperature by 1c.

 

When I installed my H110i at the top of my case I accidently fitted the two fans to intake air instead of exhausting it. At the moment I have 4 Corsair ML140 Pro fans in my case, two front intakes (blowing air over the HDD cage and GPU), one bottom intake fan (blowing air over the GPU) and one rear exhaust. Do you think I will gain anything by flipping the fans on the H110i around to exhaust air?

Not in terms of CPU temp, but potentially in terms of lower case temperature which affects everything- coolant, GPU, RAM, VRM, etc. If you can give me an idea of your typical usage, I can probably make a recommendation. For cpu only rendering intake might work best. For mixed load activity, exhaust is probably better. You might also think about 2 more ML140 for the radiator. I don’t like the stock fans much and your description of the noise is consistent with most users. You could run the ML140 at 1000-1100 for the same noise as 700.

I use the PC only for gaming, nothing else. In games such as Battlefield 1 and Assassin's Creed Odyssey my GPU temperature peaks at 72c in an ambient environment of 30c.

 

I agree with your recommendation and have ordered another 2 ML140 Pro fans for the radiator. I am very happy with the performance of the existing 4 ML140 Pro fans in my case. I will make use of the opportunity to install them as exhaust fans this time round, seeing that my load is mixed. I have also ordered some Cooler Master Nano Thermal Paste, which I understand is excellent, and will apply this to the CPU to see if it also makes a difference. Hopefully I will learn something in the process.

Keep in mind that, if your GPU is at 72C, it's dumping a good bit of heat into the case. If your cooler is configured as exhaust, this will have an impact on your CPU cooling. You can confirm this by running a load test with the case open (more airflow/cooler air).

I think I would try top exhaust for your case. The top of the case is typically warmer than anywhere else and this has a more substantive effect on the coolant temp than the air passing through. Regardless, getting air out of the box is still the priority and that would seem to be better balance. If it does not turn out better, than you know for sure and can flip them back in 15 minutes.

Keep in mind that, if your GPU is at 72C, it's dumping a good bit of heat into the case. If your cooler is configured as exhaust, this will have an impact on your CPU cooling. You can confirm this by running a load test with the case open (more airflow/cooler air).

 

I played Battlefield 1 yesterday for about 90 minutes with the side panel on and thereafter for about the same amount of time with the side panel off. I thought Battlefield 1 would be a good test of both the CPU and GPU (maximum settings, 1080p, dx12). The ambient temperature in my study, thankfully, was a bit cooler at 26c. My peak temperatures according to LINK were:

 

Panel on: CPU - 63c, GPU - 68c

Panel off: CPU - 61c, GPU - 65c

 

Is this a telling statistic to establish whether to intake or exhaust? If not I'll try the exhaust route first over the weekend (my fans arrive on Friday), and if that fails, flip the fans around as c-attack suggested.

More telling would be the coolant temperature.

More telling would be the coolant temperature.

 

Luckily I wrote it down.

 

Panel on: 34.3c

Panel off: 34.0c

What that shows you is the case is +2-4C warmer with the door on. Theoretically that means you need more exhaust flow to remove GPU heat, but probably everyone is a few degrees warmer with the case closed. Hard to have perfect flow. Was this with the current fans in the intake position?

Was this with the current fans in the intake position?

 

Yes (push configuration), the new ML140 Pro fans I ordered are being delivered on Friday.

I installed the new ML140 Pro fans over the weekend and also applied new Cooler Master Nano thermal paste. I tested the fans in an intake (push) and exhaust (pull) configuration. The push/pull configurations were purely based on my aesthetic preference (fans hidden out of sight above the radiator).

 

To be consistent I ran Realbench 2.56 with both configurations for 4 hours each, running the two new fans at 700 RPM fixed also. The ambient room temperature was 28c. The difference in CPU temperature as reported by the peak temperatures in LINK was:

 

Intake (push): CPU package 60c, coolant 34.9c.

Exhaust (pull): CPU package 65c, coolant 35.5c.

 

However, the intake result is a whopping 16c better than my initial test! I assume it's a combination of the thermal paste and me maybe not fitting the water block properly the first time.

 

To test the effect on the GPU I played Black Ops 4 MP for about 2 hours with both configurations. According to LINK my GPU displayed the following peaks:

 

Intake (push): 71c

Exhaust (pull): 69c

 

On the balance I decided to keep the intake (push) configuration.

What was 16C better? Peak CPU temp? Yes, that would be TIM, contact, or some voltage change in the testing. No 16C change in the cooler. You seem consistent both ways can choose whichever you prefer and it shouldn’t matter.

What was 16C better? Peak CPU temp? Yes, that would be TIM, contact, or some voltage change in the testing.

 

Yes, CPU package temperature. I agree must have been TIM or contact, voltage and all other settings have remained unchanged.