H150i - Change pump speed on trigger

I dont think this is currently possible but I figured I would ask.

Is there a way to change the pump speed based on some/any temperature sensor ? For example, for fans I can create a custom curve based on coolant/other temperatures. Would be nice to change the pump from say balanced to extreme when a temperature starts increasing.

I have a solution (very crude workaround) for this problem. I have a 9900k, and the difference between quiet and performance of the pump when the cpu is on full load is around 15 degrees celcius.

This is not an ideal solution - by far, but it works for me. Warning: It requires some light scripting/programming

To sum up my solution: I use LINK to log the temperature of the CPU package to a file, and I use a small Python script to read the file every second. If the temperature reading is above a given threshold -> the python script replaces the profile-file of LINK with a profile that has the pump set to High, then it restarts the link service which causes the profile-file to be reloaded. If the temperature goes down, and stays down for a given amount of time, the script replaces the profile-file with a profile-file with the pump set to quiet. This python script is started automatically when I start my computer.

A bit more details if anyone actually wants to do this:

Start by identifying which profile-file you are using. This is within C:/programdata/Clink if I remember correctly. Try changing the profile in LINK a bit, and look for which file was changed by looking at the "modified date" in windows explorer.

Make a folder named "quiet". Set your profile to Quiet pump speed. Copy the profile-file into the folder.

Make another folder named "high", change your current profile to "high" pump speed, and copy the file into the folder named high.

Now activate logging in corsair link, and choose CPU package temperature with the lowest possible interval (1 minute at the time of writing). Specify the location of the file, and remember it.

Make a script/program that reads this file every second, and check if the temperature on the last line of the file (the last reading) is above the threshold you want, if it is: Copy the profile from the "High" folder into the original profile-folder. Now use OS command calls to stop the corsair link service and restart it.

Notes:

A: Sometimes corsair makes a new log file, so make sure you check which log file is most recent, and read the last line from that file.

B: Sometimes LINK fails stops working when you restart the service, I check if the logger outputs the same value for 60 or so sample, indicating that the program has failed. then I restart. do this until the program starts reporting temperatures that aren't always the same.

C: The temperature logging/reading/restarting service is a bit slow, and my 9900k heats up really fast (seconds), so i extended my python program so that it also checks CPU usage, and sets high pump speed if the processor is working very hard.

D: As far as I can tell, the pump and fans keep working while the service is restarting

I do realize this is probably a lot more work than most people are willing to do. but I figured there may be one more person like me who doesn't mind tinkering with this kind of stuff to get a quiet pump when idle and low temperatures when on full load.

I'm not currently at home and don't have access to my python script, but if anyone wants it I can upload it at a later date.

CPU temp, not coolant temp. And it was maybe a bit over exaggerated.

At idle, pump speed does nothing but make more noise, no change in CPU temp. - so a quiet pump is great

When gaming pump speed quiet -> performance lowers CPU temp about 5-6 degrees.

when using Handbrake with AVX instructions the difference between quiet and performance of the pump is 10-13 degrees

You can never accurately assess pump or fan efficiency with CPU temperature. It's too dynamic and there are too many other variables to slice an instant in time data point and declare meaningful change. You need to use coolant temperature. That is the variable directly affected by fan and pump alterations. There isn't any reason to program the software to respond to CPU temperature. That's now how this cooler or any other water system works.

I am looking back now but cannot find your cooler model listed above. For some reason I thought you had a 2016 era H115i and the above would represent a common issue on that model that needs attention. Quiet and Performance sounds like a 2 speed model and we are probably talking about the same family of products. Can you add in some detail?

"You can never accurately assess pump or fan efficiency with CPU temperature." - sure I can. I have a quite high threshold value for changing to a high pump speed, and I require multiple readings of low values before a change back to quiet - to eliminate these "dynamic" changes you refer to. I have tested with both coolant temp and cpu temp. For me, with some tweaking, cpu temp is the best choice.

"There isn't any reason to program the software to respond to CPU temperature." Yes, there is. The benefits of using CPU temp is that you can react to thermal changes MUCH faster. Of course there is a tradeoff. Once in a while my pump goes to high when it didn't really need to, but when it needs to, it does it way quicker than if i listen to coolant temp. Within 3-4 seconds of AVX instuctions my CPU gets very hot, changing the pump speed imidiately drops the temperature drastically, but if I use the coolant temperature this doesn't happen before the coolant has had time to warm up - essentialy making me run a hotter cpu for a longer time. I have done quite a lot of testing, and my current setup works wonders.

I've used this setup for a couple of months now. When I'm idling or doing light desktop work, my computer is nearly silent, and my temps are great. When I'm rendering heavy multi-threaded workloads, my script changes the pump speed within a second or two.

"Quiet and Performance sounds like a 2 speed model and we are probably talking about the same family of products."

- I have the same cooler as the original poster "H150i". It also has a "balanced" pump speed, but I find the noise from balanced almost as bad as performance, so I'd rather only use quiet and performance.

I do not have a problem in need of solving. - I simply wanted to share my solution to have a quiet pump when idle, and an high performing pump when doing heavy processing. - feel free to use my solution with coolant temps instead of cpu temps, it would still work. But your cpu will run hotter for a longer time before the pump speeds increases

"You can never accurately assess pump or fan efficiency with CPU temperature." - sure I can. I have a quite high threshold value for changing to a high pump speed, and I require multiple readings of low values before a change back to quiet - to eliminate these "dynamic" changes you refer to. I have tested with both coolant temp and cpu temp. For me, with some tweaking, cpu temp is the best choice.

I could write a multi paragraph explanation of the cooling process, but frankly I don't think you really care. You like doing things you own way. That's fine. There is a difference between CPU temperature on the socket side versus the amount of heat that is transferred into the cooling system and needs to be disposed of. The fact that you can make a cooling system work with less than optimal tools does not represent a valid scientific process or analysis, let alone a conclusion.

"There isn't any reason to program the software to respond to CPU temperature." Yes, there is. The benefits of using CPU temp is that you can react to thermal changes MUCH faster. Of course there is a tradeoff. Once in a while my pump goes to high when it didn't really need to, but when it needs to, it does it way quicker than if i listen to coolant temp. Within 3-4 seconds of AVX instuctions my CPU gets very hot, changing the pump speed imidiately drops the temperature drastically, but if I use the coolant temperature this doesn't happen before the coolant has had time to warm up - essentialy making me run a hotter cpu for a longer time. I have done quite a lot of testing, and my current setup works wonders.

This argument fails at the start. If maximum performance was your prime goal, then you would initiate all pumps and fans to maximum prior to initiating the load. You've instead made it reactive and this again goes back to not understanding how the cooler works and difference between the two temperatures. You can continue to operate as you are doing and nothing bad will happen to your CPU, but you are spinning your wheels without purpose undermining the unstated and obvious secondary goal.

"Quiet and Performance sounds like a 2 speed model and we are probably talking about the same family of products."

- I have the same cooler as the original poster "H150i". It also has a "balanced" pump speed, but I find the noise from balanced almost as bad as performance, so I'd rather only use quiet and performance.

 

I do not have a problem in need of solving. - I simply wanted to share my solution to have a quiet pump when idle, and an high performing pump when doing heavy processing. - feel free to use my solution with coolant temps instead of cpu temps, it would still work. But your cpu will run hotter for a longer time before the pump speeds increases

I always like it when people make me guess what products they have. It really adds a fun, game show like dimension to the forum. If you had included your cooler model in any shape or form in your post or in the specifications, this conversation never would have occurred. You are jumping back and forth between the ultra low 1100 rpm pump speed meant for quiet desktop work and the maximum. Yes, you will see a difference in performance both at the CPU and coolant temperature level when doing this. Most CPUs will register a noticeable coolant change between the 1100 RPM and either of the two higher speeds on the PRO models. I do not recommend the 1100 rpm pump speed for moderate CPU load, let alone more taxing things like rendering, encoding, etc. However, the words quiet, balanced, and performance are device specific and Quiet on a H150i is not the same as Quiet on H115i or H115i Platinum or a H80i GT. This 1100 RPM quiet pump mode only exists on the three PRO series coolers. This is significantly different than someone using a H100i/H115i witnessing a +10-15C change between their 2000 rpm Quiet mode and 2950 rpm Performance Mode, which is unfortunately a commonly reported problem.

So we are back to using a measure of common sense and listing what product you are talking about. If you want to write a script to auto change you pump, that's great. If you want to share it, that's great. However, your science is soft and you wasted a lot of your time and mine by not doing something very basic and obvious across multiple posts. Don't berate me for trying to make sure you don't have a failing unit. Given your lack of understanding of the cooling mechanics, it would also be reasonable to assume you may not recognize an appropriate coolant delta as well.

Also, Link was replaced with iCUE coming up on a year ago. If you prefer it, that's fine, but this may not have much utility for others.

Alright. I see your points. Thanks for taking an interest.

Regarding me not disclosing my device; This is a thread specifically about the pump of a H150i, I thought it was implied that i had the same device as I did not specify anything to indicate otherwise. But note taken, I'll say so next time.

and, no. Maximum performance is not my prime goal. my prime goal is quiet desktop use and good performance when gaming or rendering, without having to do any manual switching. - that's literally what this thread is about.

Thanks for looking after the posters here. I see that I should have clarified much earlier that i use the absolute lowest and highest pump settings. But I must admit that i'm slightly upset that you call my science "soft" when there is, as far as i can tell, no better way to solve this specific scenario. Also claiming that i have a "lack of understanding of the cooling mechanics" based solely on the fact that I use CPU temp instead of coolant temp is a straw man argument

Regarding me not disclosing my device; This is a thread specifically about the pump of a H150i, I thought it was implied that i had the same device as I did not specify anything to indicate otherwise.

That makes perfect sense and is quite logical. Unfortunately, 90% of the planet's population does not follow that in practice. They keyword search, find one commonality in the thread, and then jump in. The "my device doesn't work thread" usually has 10 different devices in it and unique hardware configurations for each one. What's worse is the spam bots now write "helpful" posts and tack on the back end long after the original conversation. They go something like, "Hey, I have that problem too and this is how I fixed it. I wrote a little script to make it work. You can download it here: http:http://www.dropbox.htheh.LOVE-ME.fhhryu.com". Good luck! Now the spam bots don't talk back so that did not remain an issue, but the one thing they all usually have is a 1-2 post history and no device specifications in their profile or in the post. They spam the thing across every tech site and can't/don't bother to try and go around the security measures for altering your user data. It's a weird world.

I am going to try and explain the relationship between CPU temperature and coolant temperature and why it matters in this discussion. However, let me get this out of the way first. I am not trying to convince you to re-write the script for coolant temperature, suggesting you are endangering your hardware, or that there is anything wrong with what you are doing. For your usage, which seems likely to jump between idle-lowest possible load and a sustained maximum for the duration of the event, it really doesn't matter much and probably works well. It might be less optimal for the average user's roller coaster ride of cycling CPU temps when gaming or other highly variable loads. If you want/need to keep the pump on Quiet 1100 rpm, you most certainly cannot use that for rendering/encoding and would otherwise have to manually switch it each time. Some people are OK with ticking a few boxes to max before a run. You wrote a script to do it. Use it. I would want the pump to shift to maximum for a render immediately and I suspect that is what your program does.

Now that said, this is why I was hammering on the coolant temperature aspect of the discussion. Your CPU temperature is voltage induced, less the amount of heat the CPU can conduct through the cold plate to the whatever is on the other side (air cooler or water cooler). The heat must travel through the CPU to get to the cooler and heat is always created on the pin side. If I set my Vcore to 2.00v and boot up, no cooler on earth will save me, because the CPU with heat up to maximum and shut down almost instantly. This CPU side of the equation is critical and the CPU temperature matters greatly. However, nothing you do with the pump or fans will affect this pin to CPU heat exchange. You have a limit regardless of cooler type and it is a combination of voltage, CPU type/design, and instruction/load type. You can see this clearly if you initiate a maximum load. In that first instant, your CPU temp will jump from its 25-40C idle level to +35-50C depending on voltage settings and load. That "instant 100%" change is all voltage. You will not see the cooler have an effect until 15-20 seconds down the line where it may tick up at +1C every 30 seconds or so. That is the coolant temp rise described below.

The cooler's real job is not cooling, but transporting heat elsewhere. Whether its an air tower or a water cooler, it takes the conducted heat and blows it somewhere else. I have started referring to this as the second stage of cooling to make the point. The cooler is a dumping ground for CPU heat, but if you don't get rid of it, it will heat up the coolant. This matters because the cold plate between the two stages is bi-directional and heat goes both ways. This means two things: 1) Each degree of coolant temp change equals +-1C on the CPU. Coolant temp goes up +6C, then this is effectively a +6C penalty to CPU temperature and it also goes up +6C. 2) Your water coolant temperature is the minimum possible CPU temperature. You coolant temperature + the above 100% on CPU delta = maximum CPU temperature when under load. Pump speed obviously affects the cycle rate or the number of trips each unit of water can make in a given interval and fan speed affects how much heat the radiator can dissipate as the water passes through on each trip. However, these things affect the coolant temperature only, which then indirectly affects CPU temperature. Except for something synthetic like a Linpack test, most loads are variable even at near maximum and the CPU temps jump quite a bit making it hard to assess change in real time. The coolant temp is (should be) much easier to take regular measurements without having to strain out outlier CPU peak data.

Most people have single digit coolant temperature deltas, so the possible reduction is the same single digit value. You can never get the coolant delta back to ambient room temp, except in that first moment after waking from sleep, etc. Most people have a baseline coolant temperature about +4-7C over their ambient room temperature. Lots of variables including case design, location, pump/fans, CPU type, and OS power level settings (EIST, etc). Coolant temperature is not a critical value in the same way as CPU temperature. Get too hot on the CPU, you throttle then shutdown. Get too hot on the coolant and you only shut down if the CPU temp is bumped sufficiently into the same danger zone. This makes coolant temperature an "efficiency variable" regarding fan and pump speeds. Since those things only affect coolant temperature, that is what you need to look at to accurately see if fan speed 1400 is better than 1200. Because it is additive to CPU temperature, it is possible to measure it from CPU temperature and coolers without a temp probe still are assessed this way. However, it is very difficult to create identical testing situations and even some professional reviewers make mistakes in their controls (like lettings the fans follow a pre-programmed curve). For a normal user, no gaming experience and most professional app runs are not exactly the same each time. Why the CPU temp spiked at 4:53 could be programming, glitch, or any number of things. Scientifically, you have to look at coolant deltas to make value judgments about pump or fan speed changes -- at least without a whole lot of background testing condition data.

Now that was a very long winded explanation of why people in general should look at coolant temp data when assessing cooler performance and we didn't even get into fun stuff like environmental changes. However, my reason for asking you originally was very limited in scope. Coolant temperature is the single easiest and clearest indicator of a problem with the unit. There are a multitude of reasons why your CPU might be at 85C. Extreme load, high Vcore, you left it on Auto settings and like to run Prime95 for no reason, etc. If your coolant temperature change is out of bounds, there are only a few reasons.

1) Your case environment is hot and thus your temps are worse than everyone else. Even in this situation, the coolant delta would be the same as someone else with the same hardware. You just go +6C from 38 to 44C baseline, instead of 21 to 27C and so your temps are +17C warmer than person A because of the local temperature penalty. The cooler is fine, your room is hot.

2) Your radiator or fans are partially blocked or in some way are prevented from getting rid of the heat at the normal rate (case design/dust filter). Coolant temperature is thus higher and so are CPU temps as a result.

3) There is something wrong with the cooler. Flow rate is inhibited in some way and this is causing poor coolant temps at both idle and load. In the early stages before it becomes a complete standstill affair, this can often be recognized when you see a large difference in coolant temperature between differing pump speeds. The slower speed is noticeably affected by the back pressure while the faster speed is still able to get some fluid through to the radiator. Most of the coolers are two speed models (2000/3000) and you should not see more than 1 or 2C of difference (or any). Flow rate is not usually a limiting factor on AIO coolers with their low restriction radiators and short circulatory pathways, so whenever you do see a large difference, there is reason for concern. The one exception to this is the 1100 rpm speed on the Pro coolers. This seems to be below the threshold for optimal cooling. I have used 1500 rpm versions of similar design and the fall off is nowhere close, so there must be a mathematical reason somewhere in that 1100-1300 rpm zone, but regardless there it is.

When you only referenced two speeds and expressed a significant change in temperature between them, and in combination with using Link, I presumed you were on an older cooler where this should not happen. Unfortunately this does happen to one specific series quite regularly and I was worried you were sitting on a dying cooler without realizing it was in the early stages of failure. Knowing it is a H150i and you were cycling between 1100 and 3000, I do not think there is a problem. However, if you have time, please share your coolant temp change on the high speed when doing your professional work. The 9900K is relatively new and we try to collect data about where things should be for varying processor and cooler combinations to help identify when there is a problem. I can mathematically work out a watts created, watts dispersed kind of calculation, but it's messy and often does not represent the real world mechanics of heat exchange. My guess is you see +8C during a 30-45 minute run, but would appreciate any data you have.

Here's some data!

I have tested the difference between pump speeds of 1100rpm vs 2800rpm, as well as fan speeds of 400rpm vs 1200rpm vs 1600rpm

I did not test the balanced mode of the pump, as I don't use it because i percieve the noise as more irritating than the 2800rpm mode.

Hardware: 9900K 5GHz all cores ~1.32V, H150i pump

All temperature values are Celsius. The workload was encoding an h265 video with AVX instructions. The system ran for 15 minutes before the temperature readings to stabalize temperatures. The ambient temperature was about 23-24 Degrees. The temperature values are averaged over 5 minutes after the initial warmup

Results:

Pump Speed 2800RPM fan speeds 1600RPM: Liquid temp 36, CPU package temp 81

Pump Speed 2800RPM fan speeds 1200RPM: Liquid temp 37, CPU package temp 83

*** invalid run Pump Speed 2800RPM fan speeds 400 RPM: Liquid temp 39, CPU package temp 85 *** invalid run

Pump Speed 1100RPM fan speeds 1600RPM: Liquid temp 42, CPU package temp 93

Pump Speed 1100RPM fan speeds 1200RPM: Liquid temp 43, CPU package temp 94

*** invalid run Pump Speed 1100RPM fan speeds 400 RPM: Liquid temp 46, CPU package temp 96*** invalid run

Notes:

I did not notice any thermal throttling, but as the limit is 100C, its very possible that the highest temperature runs had some minor throttling.

My case is a define r5 with dust filters, some cable cluttering and HDD cages. The airflow is below average.

edit: The 400rpm are invalid, I'm not deleting them because they are referenced later, but do not consider them to be valid results. the actual 400RPM runs were so hot I had to quit the tests before the temperature stabalized

Ugh... that 1100 rpm pump speed is brutal and kind of sells the point that it would be good from a hardware perspective for the pump to kick up automatically when you cross a certain threshold - like 40C coolant. I'm glad it exists, but you definitely have to remember to set it back.

That is a pretty impressive and tight coolant spread between the 400 and 1600 rpm speeds. The 1600 vs 1200 is exactly what I expected, but the only dropping 2C at 400 rpm is not. That is a pretty nice flat fan speed vs coolant delta curve and confirms most people do not need to blast their fans on the H150i without an additional reason.

Do you remember (more or less) what the starting coolant temp was for the 2800 rpm pump speed runs? Ambient 23-34 so perhaps 28C?

Do you remember (more or less) what the starting coolant temp was for the 2800 rpm pump speed runs? Ambient 23-34 so perhaps 28C?

28 sounds about right.

The 1600 vs 1200 is exactly what I expected, but the only dropping 2C at 400 rpm is not.

I went back and redid the 400rpm test as you commented it was unexpected, and it turns out I must have misconfigured it.

I believe it was probably set to quiet instead of a fixed rpm. The new results are nowhere near the previous temps. Coolant temp skyrocketed to 47 degrees even with the 2800rpm pump speed. I quit the test because the temperature just kept on increasing.

The 1600 and 1200 fan speeds were consistent with the previous results, but the 400rpm runs are invalid. - my bad