H150i temps

Hello,

I have another problems now with my brand new H150i cooler.

I have slightly overclocked system, R7 2700X@4.1Ghz@1.35V so should not consume more than 150 watts. And before on idle I was getting all balanced profiles - around 900rpm and sub 30 degree coolant temps.

 

Now my fans are at 1300 rpm, coolant temps get to 31 degree - seems like something is not working properly ? It is 360mm radiator, so with ML fans should works almost inaudible like it was at start, now I have almost same noise levels as I had with 280 mm corsair AIO half working... Putting pump to extreme mode is also not helping at all...

 

Yes CPU has a lot lower temps, maby -15~20 degrees. But it is still noisy...

 

I also attach screenshot of cpu load and temperatures ...

 

EDIT: According to HWINFO64 average cpu load is 40W, so for 40W with top of the line cooler I can not have almost inaudible system ? Wtf seriously ...

Edited January 2, 2020 by AndrejusD

You need to take control of the fan curve and design your own. All CPU systems are subject to the effect of room or environmental temperature. If I pick up your system and drop it in 12C room, the fans are never going to move off their minimum because the coolant won't even get back to standard room temp. If I put you in a tropical bungalow with no air conditioning and a 32C room temp, the fans will blast right from the start. Everything has to have a baseline temp vs acceptable fan noise level. The three presets were designed around a 20-23C room temp, but there are some intervening factors, most notably case design. Most users will have a baseline coolant temp of of 4-6C above their room temperature. Hardware matters too and a 16 core whatever will have a higher idle TDP than inline 4 core from 5 years ago. Essentially, every set-up is unique and that does not lend itself to standardized presets.

 

The solution is to slide the curve up or down based on your environment. What you really want is for the fans to sit quietly on the desktop and ramp to moderate speeds when continued load is present or more specifically, when the coolant temp actually has some measurable rise. My normal suggestion is you set the 120mm fans to a tolerable level at your normal idle coolant temp. That is probably between 700-900 rpm for about 30C. The sweet spot for most 120mm fans for work vs noise is between 1000-1300 rpm. About 3C down the line, start to ramp the fans up to 1000 and then you want the 1300 rpm to be at the maximum coolant temperature you normal see doing whatever it is you do. Don't use a stress test to find this. You are trying to create a useful fan curve for daily life. I will estimate that is going to be about +6C without accounting for GPU heat and the effect on the entire case. Most people will long load coolant temps of around +6-10C. Somewhere around 45C, put the fans to max. You should not reach this point, but you will get a clear and audible warming if you do.

 

The supper quick option is to shift the presets up or down by a specific amount. You could take the Quiet preset and shift each point upwards 7C. That would likely replicate the intended curve for your environment.

 

*Performance Tab for the H150i. Hit the + to create a new "cooling mode" (you can rename this). Tools down below. Use the boxes (lower left) to enter X, Y coordinates for temp and RPM. Much easier than dragging the dot. The three presets are in the upper right hand corner. You can click on them to transform your curve, then drag each point up by the desired offset.

 

Pump speeds are fixed and still work off the Quiet, Balanced, Extreme presets. Those presets have a very different meaning here and correspond to 1100, 2160, and 2850 rpm. Leave it on balanced. You can use the 1100 speed if you need it on the desktop, but you will want to put it back before doing extended load work. There is a temp penalty at the lowest speed.

Edited January 2, 2020 by c-attack

Finally found that custom curve and now they spin only @ 360-380 while maintaining 33 degree coolant temp.

 

Next question - can I spin down iCue 465 stock fans to the same rpm ? They are now used for air exhaust. I use nzxt grid 2 to control them and even on smallest values they spin @ 900 rpm. Can corsair commander fix that ?

LL120 fans should have a 600 rpm minimum and that is very quiet on that model. Since they are PWM, the controller shouldn’t matter. Most likely your motherboard is using CPU temp as a control variable and that makes it rather hard to maintain any kind of stable or graduated speed. Not sure what other choices you have for a control variable.

 

A Commander Pro will allow you to set the fans with almost every available metric, including coolant so you can match front 3x120 with the top and rear 120s. Gpu temp is also available as a choice. This could either replace the Lighting Core in the 465x or you can plug the LNC’s usb 2 cable into the two on the Commander Pro. All lighting and fan speed control will be through iCUE.

Ahh, so this is why they won't spin down normally :) nzxt controller is not pwm so it shoud hive smth like 7V or so when put to minimum 20 percent of load. I think I need to try put some resistors to lower their speed and bump percentage a bit. I have few resistors from my previous bought fans, must try ir later :) Just don't want to buy whole commander because of 3 fans :eek:

So lets go futher.

Now I was trying to overclock my cpu a bit more than 4.1 Ghz 1.35V, lets say to 4.175Ghz and 1.4V and...

Something is wrong here. Max power draw from the CPU was 170W, and it hit 90 degree at 36~37 degree coolant temperature. So delta temp is 53 degree ? :eek::eek: Which resulted cpu temp to spike to 90 degrees. Wtf ? Really ... See attached printscreen.

 

To be honest ... I was expecting that 180 euro AIO would outperform some air coolers like Noctua or BeQuiet at least in few degrees, as theese confirm to perfectly dissipate 250W chips at under temperature limits, and now I see it might be doing even worse :mad::(::[pouts::sigh!: ...

 

Will applying liquid metal from thermal grizzly void my warranty for H150i ?? Wanna give it a try, what if cpu surface is not flat enough...

The Vcore to CPU temp graph is never linear. Nearly all CPUs have a tipping point where instead of getting +1-2C for each 0.01v of Vcore, you get +5-7C. This typically becomes the effective end of the line without going to extreme measures. All CPUs, whether Intel or AMD, are voltage limited. I can give an external radiator array the size of a wall and you still can't run 1.60v on a CPU. Voltage creates the heat and it must pass through the CPU first, before the radiator can dispose of it. If you take a cheap air blower, a 360mm AIO, and the aforementioned wall array, then load up to 1.40v, all three will give you the exact same CPU temp at 100% load -- at least for the first 15 seconds or so.

 

What happens after that is a function of the coolers performance. You conduct heat in. Cooler must dump it elsewhere or conduct the retained heat back to the CPU. Little air blower can hang on for 15-30 seconds before the temp creep up starts. 360 AIO can handle every CPU wattage out there, but regardless some coolant rise is inevitable. The massive wall panel will be the exact same CPU temp from start to finish as it dumps all heat in each pass.

 

What you are getting from your radiator size is the ability to continue to run higher settings without additional penalty, beyond the unavoidable one from voltage. It is a small numbers game in comparison to the voltage penalty, but it is the only one you can really play. To address the conducted heat between CPU/cold plate, you have to change the CPU is some way. Different model CPUs behave differently. Every once in a while you can get massive results from delidding (Intel Coffee Lake -20C), but this is not common. I can't speak to the voltage tipping point on Ryzen 2 series, but surely that discussion is out there. Ryzen 3 doesn't have much headroom either. Sadly, none of these things are solvable through cooling mechanisms. It would be nice if it were all that easy. Then we could buy big cases and pack in 4 radiators and all run 6 GHz. That's not how it works. What you get for increasing your radiator surface area (number or fan slots) is the ability to get maximum possible cooling with much lower fan speeds. You can get really good cooling from a thick 120mm radiator. You just have to keep the fans at 1800 rpm all the time. At acceptable voltage levels, you can keep your fans quiet without penalizing yourself. However, it can't stop the pins from heating up the CPU.

I understand your point, but in my case something is blocking thermal movement from cpu to waterblock, this is why I think I will need to lap my 2700x or use liquid metal. I've been lapping cpu's for a long time. My previous LGA775 and 1366 x5675 xeons, both were lapped to ideally flat surface, and it helped to achieve +50% OC at least :) I know for ryzen it won't help much in OC, but in thermals it must.

 

Found theese two videos

[ame]

[/ame]

[ame]

[/ame]

 

They just confirming my thoughts that maby ryzen lid is not flat enough so thermal interface just can not transfer heat that fast.

 

What do you think ?

I don’t know that I would describe it as “blocking” thermal transfer. The silicon just retains more than you would like. As for lapping, I don’t know. I haven’t had an AMD chip in 12 years. That question needs to be answered by someone who has done both. Liquid metal will begin the alter the appearance of the cold plate almost immediately. It will discolor it slightly. That should have no impact on performance.

I don’t know that I would describe it as “blocking” thermal transfer. The silicon just retains more than you would like. As for lapping, I don’t know. I haven’t had an AMD chip in 12 years. That question needs to be answered by someone who has done both. Liquid metal will begin the alter the appearance of the cold plate almost immediately. It will discolor it slightly. That should have no impact on performance.

 

I mean that if there is thick layer of thermal grease, because the lid is not perfectly flat, it might be that it takes time for the heat to transfer from cpu to watercooler, and if that is the case, then liquid metal will remove this limiting factor and it heat will move much faster, what will result in better cooling performance. Simple physix.

The entire point of TIM is address imperfections in the surface. I would not get your hopes too high for the temp difference between using liquid metal vs normal TIM. On a CPU, 3C would be about the best possible result. I also don't really like what it does to the CPU lid, although perhaps AMD makes uses a different compound. You don't use Liquid Metal if you have future resale hopes for your gear. I do like it on GPUs and Nvidia's chips are not affected long term. However, like always you need to be extremely careful. No one wants to make a mistake on a $1500 GPU.