Question AIO + 4 Fans - How to find the sweet spot under 100%?

[ldehaes]

I recently watched this video and it made me think about finding the sweet spot myself:

View: https://www.youtube.com/watch?v=F7U9QZhulXk


How do I go about finding the right constant percentage for myself? The noise doesn't matter in my case, all I want is to keep the average CPU temp as low as possible with my current equipment.

AIO is positioned for exhaust, the four fans are set for intake.

 

[bill001g]

If you don't care about noise you run the fans at 100% to get the maximum air movement.

Just be aware lower CPU temp does not in many cases increase your performance. You are basically just getting a lower number on your display with no other advantage.

Key thing is if your temp is below the point where the cpu starts to degrade the clock speed it really doesn't matter if it is 10 degrees below or 15 degrees below the cpu performance will be the same.

 

[ldehaes]

If you don't care about noise you run the fans at 100% to get the maximum air movement.

Just be aware lower CPU temp does not in many cases increase your performance. You are basically just getting a lower number on your display with no other advantage.

Key thing is if your temp is below the point where the cpu starts to degrade the clock speed it really doesn't matter if it is 10 degrees below or 15 degrees below the cpu performance will be the same.

What temps do I look for in this case? I have AMD Ryzen 5 7600x3d.

Is my setup even capable of cooling the CPU to the effect you described?

 

[bill001g]

Load hwinfo64 and see if you get red numbers. It tends to know the limits on almost all cpu. I think it is 95c for your cpu but I forget.

It will actually tell you if the cpu is hitting thermal throttle limits.

All depends on the AIO you are using. Almost all 360 aio can cool any cpu. 240 tend to do a pretty good job also. 120 are stupid since a good $50 air cooler can exceed the cooling and have less parts to fail. There are bunch of $50 or less air coolers that have no issues cooling anything other than overclocked 14900k.

 

[Paperdoc]

Basically you want to be sure that the fans CAN cool your CPU properly (NOT reach the thermal limit of the CPU chip) even at the very highest workloads you ever get to. If your highest workload lets the rad fans run at significantly less than their max fan speed, you can afford to set the PUMP speed lower and make the fans run faster. Once you figure out what pump speed that is, leave it set that way always unless you change your general usage pattern to include new even higher workloads sometimes.

 

[ldehaes]

Update:

I went with 80% capacity of my AIO after testing loads in the 75-90% range. The fans are all at 100% capacity, since I don't worry about the noise.

Used HWinfo to log temps and GenericLogViewer to visually compare the results after using OCCT for "stability tests" on my CPU.

CPU: AMD Ryzen 5 7600X3D
Cooler: Arctic Liquid Freezer III 360 A-RGB

 

[Paperdoc]

If I understand your last post, BOTH the pump and the rad fans are set to fixed speeds. Pump at 80%, rad fans at 100% always and you are happy with temperatures of the CPU. You might try setting the rad fans' header to use the "Standard" or "Normal" Profile and let the mobo adjust he fans to slower speeds when your workload is light. That would reduce noise (you are not worried about that) and possibly give longer fan lifetime. Check what that does for CPU temps.

 

[ldehaes]

No, the fans are just regular PC fans, which are all at 100%. I don't think there's an option to set my AIO pump speed directly, at least I don't remember seeing it anywhere. That leaves the fans attached to the radiator, which I ended up setting at 80%.

 

[Paperdoc]

You have not told us what mobo you have - maker and exact model - so I can't suggest exactly how to do this. But if you are not really interested in making the rad fans adjusted automatically by the mobo, you don't need advice anyway.

 

[JayGau]

This Jaytwocents test was awful. He used only one AIO model for the whole thing, then measured the CPU temperature instead of the liquid and totally ignored statistical fluctuations. After I watched his video I thought "ok, so we should definitely run our pump at 100%". He wanted to show that "we are all doing it wrong" to get views but ended up showing the opposite.

AIOs models are different from each others and have different pumps. This 80% he found might be the best for the pump he used for his test, but could be bad for another one (another model at 80% could fall on his 70% curve, for example). Does he really think we should all run the same test every time we get a new AIO? Of course not, and since he showed that there was this sharp drop between 70 and 80%, that just proves the importance of running it at max speed to make sure you are not on the wrong side of this critical point (unless you are willing to do this elaborated test every time you buy an AIO).

And statistically speaking, there is likely no difference between his three curves from 80% to 100%. The CPU temperature always fluctuates a lot depending on a many factors (ambiant temperature, background load, voltage fluctuations, etc.). A CPU temp of 72.5c after 11 minutes and 32 seconds under load will likely not be 72.5c after 11 minutes and 32 seconds if you redo the exact same test a second time. He should have done his test with the liquid temperature instead, which is the most important value when you work with AIOs. This temp is very stable, changes very slowly, and it's what matters since it's the liquid that cools down the CPU.

So, according to me, you should just run your pump at max speed and ignore this video.

By the way, the Artic Freezer coolers usually come with two options for the fan and pump control (both included in the box): an all-in-one cable that connects to the CPU fan header and controls everything with only one fan curve, and another triple cable for the rad fans, the pump and the VRM fan that connects to three fan headers on the mobo so you can control them separately.