Custom Fan Curve only responds to Commander Pro Temp

[Dondomingo]

Been trying to get my Fan curves set and this ICUE program is giving me a run.

The custom fan curve Ive set up to use, only seems to change RPM if the sensor is set to Commander Pro Temp.  Which is typically 10 degree's cooler than the GPU Sensor.  When I change it to the GPU sensor 1, 2 or 3 ( hot spot  ) or the CPU Sensor in the drop down, all fans drop to 600rpm. Then when i switch over to the Commander pro temp sensor, the fans react according to the current temp of the commander pro.

The fans are set to the 4pin setting for the coolermaster masterfan Halo's.  So on the Dashboard i can see the temps for the CPU, GPU and Commander pro.  When I switch over to the Commander pro it's only listing the fans, voltage and temp for the commander Pro. 

Now the crazy thing is , i created a zero RPM to see which fans are which in sequence and that works no matter the temp sensor's selection.

 

What am I missing?

 

[Dondomingo]

So I rolled it back to ICU 4.33.138 and now GPU Sensor #3 doesn't even register, only GPU 1 & 2....  how the hell did that drop off lol.

[Dondomingo]

Reinstalled V5 and GPU temp sensor #3 is back and so is the lack of controls for the RPM other than the commander pro temp sensor.

[c-attack]

CUE 4.33 and older did not have the extra temp data for some models, so that part is expected. 
 

There have been several user posts complaining about malfunctioning fan curves on this 5.7 release, but most had zero detail.  This does makes it clear.  I’ll see if I can replicate tonight. 
 

The Commander Pro Temp probe(s) is the right control source to use in most applications. It’s also native to the controller, so it works with or without CUE running and no max fan blast when you boot up and shutdown. The case fans help control the ambient case temperature and the probe measures that in a specific place. The fans don’t directly cool the CPU or GPU. You’ll have to figure out the normal range for that exact sensor position, but it is a more consistent control method. You don’t want the case fans to drop off for 3 seconds because you opened a game map and the GPU load/temp dropped to a fraction. 
 

Regardless of the above, this is a problem and is likely going to need a software correction. It’s probably going to be linked to how the data is fetched/presented for the fan curve which has been something of an ongoing issue for a while. Possible there was an unintended casualty in trying to correct that. I don’t think AMD owners have gpu temps at all right now. 

[Dondomingo]

Hey There C-Attack,  I've seen you reply to thousands of posts as i was researching the problem, so I know I'm in good hands lol.

First let me be honest, I didn't build the system, my condition with my hands doesn't warrant me getting my hands into tight cabinets , with tiny screws unless I'm trying to check my level of PSTD.  So I had Digital Storm build it, I'm just trying to figure out the reasoning they used.

So Commander pro has the two temp leads in slot 1 and 2. I can follow those leads to one being behind the CPU MB with a fan directly on it (Picture 1)  and the 2nd looks to go to the front and taped under the GPU. (picture 2 & 3) .

Why does Commander pro only have one temp registering then?  As you can see in the pictures above (Sensor labeled Commander Pro) I checked to see if i can add another Commander Pro Sensor but there is only the one.

As a side question, where does the commander pro pick up the GPU temps from? (GPU Temp sensor 1,2,3)  Is it picking up the rest of this data from the Bios/MB?

After digging around, looks like the Bios is controlling the Coolermaster AIO fan curve and one chassis fan (I'm betting it's the one on the back of the MB)  The other 6 fans are controlled by the Commander Pro. I've verified this.

Regardless, you're right though, seems after a update this started happening and seems from other posts to be only affecting AMD users go figure lol... Told my friends I was gonna give the AMD 7 7800X3D a go, I've been a Intel person since the 80's but the reviews and had some bad experiences with some way older AMD chips, but the stellar reviews on the chip convinced me otherwise. Go figure , first week I'm already running into AMD only issues. That'll teach me.

Here's the core of the build in case you need:

AMD Ryzen 7 7800X3D (8-Core) 5.0 GHz Turbo
ASUS ROG Strix X670E-E Gaming (Wi-Fi) (AMD X670 Chipset)
 64GB DDR5 6000MHz Kingston FURY Beast RGB
1500W Corsair HX1500i
PNY GeForce RTX 4090 24GB
Operating System: Microsoft Windows 11 Home (64-Bit)

 

[c-attack]

No doubt there is an issue with the fan curve on the Commander Pro when the variable is CPU or GPU temp.  I did not go through the rest of the variables, but it is near certain they will the same.  I think this will affect the Commander Core the same way, but the CUE Link Hub and it's connected devices appear to work normally when using CPU package temp.

 

As for the temp probes, you can put them anywhere you want for informational purposes.  However, the key location with the most utility is the rear exhaust -- either inside or outside the case.  Regardless of whether the heat comes from the GPU, RAM+VRM, CPU air cooler, or anything else, a rise in exit air temp tells you interior heat it up and you can mitigate that by bring in more cool air from outside while trying to evacuate the other.  It's somewhat difficult to directly cool a component like the GPU, CPU, RAM without putting a specialized fan right on it.  The GPU's own fans always will be the key to managing it's temp.  The case fans are on the "don't make this worse" side of the table.  

 

The following is an easy way to set things up.  Everyone has a max fan speed limit they are willing to tolerate for normal  use.  Whatever that value, you want it to be just above your load case temp.  So if you put a temp probe near the rear exhaust and it generally tops out at 55C when gaming, set your max allowed fan speed to 58-60C to start.  Other than that, it's really about noise management.  Nobody drops 5C on their GPU because they increase the fan speed by 200 rpm.  It's not anywhere close to being that sensitive to fan speed changes.  You also likely have normal desktop exit air temp for the probe in the mid to upper 30s and that should be set to your preferred quiet fan speed.  Everything in between can be a straight line.  In practical purposes, this makes the case fans really quiet except when the GPU starts using a lot of power for an extended time.  

 

The radiator fans for the AIO present a choice.  Easy thing to do is move them back to the motherboard so it can properly get a CPU temp.  Quick and mostly self-explanatory.  The other option is to put a Commander Pro temp probe on the exhaust side of the CPU radiator.  There is a direct relationship between the liquid temp and the exhaust air temp, typically as a small offset like +-2C from the actual.  This makes a very effective way to "run from coolant temp" even when you don't have a real sensor.  When you heat up the inside with GPU heat, it also will reflect that and increase speed.  That can be a plus of minus depending on use and case layout, but it's the same as when using any Corsair AIO and the coolant temp within CUE.  

 

1 hour ago, Dondomingo said:

Why does Commander pro only have one temp registering then?

These should generate automatically with no user input required.  The most probable is the sensor wire is not fully inserted or defective in some way.  These 10K thermistors are cheap and not too difficult to damage.  It could be the senor on the C-Pro as well, so try moving it to another register like 3 or 4.  

 

1 hour ago, Dondomingo said:

As a side question, where does the commander pro pick up the GPU temps from? (GPU Temp sensor 1,2,3)  Is it picking up the rest of this data from the Bios/MB?

I believe it gets it from the GPU driver, much like the rest of the system.  CUE uses the CPUID developer kit to fetch data from other devices.  My CUE is properly obtaining these control variables and I have data, but something is preventing the fan curve from utilizing it.  I do not believe this was a known condition and it's going to be a bit before we get a fix.  There are some separate ongoing issues with AMD GPU temps and past issues with AMD CPU temps.  Obviously if CUE can't get the AMD GPU temp, it won't work for  control regardless so there are some  overlapping issues.  Nevertheless, this fan curve one is not limited to AMD.  It is affecting my Intel CPU and Nvidia GPU.  

Edited October 13, 2023 by c-attack

[Dondomingo]

C-Attack, you are a damn national treasure.  Thanks much.

Gonna check those leads to determine which one is not connected/broken.  I'll switch their locations as per your suggestion. Unless i find the Commander Pro unit is to blame, then I'll switch that out.

Other than that, until they fix this update,  will have to look into something to watch the GPU as its probably the hottest unit, reaching 86c one time during a long play session I went ahead and logged the sensors for a snap shot. I think I read PNY's website saying 95C is their cut off before it shuts down but thats a 4090 RTX for you, they run pretty smooth under that heat. I should of looked into a closed loop and did the GPU as well, but that hindsight for yah.

So, thanks for all the tips.  I'll say one thing, even at max RPMs these Coolermaster Masterfan Halos on the case are dead quite. 1800rpm and I barely notice. So I'm not too concerned as far as noise vs performance.  Rather have the fans working doing their job vs what I've seen one degree can do to a capacitor or connector. a day or so ago, I was in the bios, testing the fan curve on the Watercooler, now at 100% those fans can make some noise for sure.

Take care!

 

[Flyprdu2]

9 hours ago, c-attack said:

No doubt there is an issue with the fan curve on the Commander Pro when the variable is CPU or GPU temp.  I did not go through the rest of the variables, but it is near certain they will the same.  I think this will affect the Commander Core the same way, but the CUE Link Hub and it's connected devices appear to work normally when using CPU package temp.

 

As for the temp probes, you can put them anywhere you want for informational purposes.  However, the key location with the most utility is the rear exhaust -- either inside or outside the case.  Regardless of whether the heat comes from the GPU, RAM+VRM, CPU air cooler, or anything else, a rise in exit air temp tells you interior heat it up and you can mitigate that by bring in more cool air from outside while trying to evacuate the other.  It's somewhat difficult to directly cool a component like the GPU, CPU, RAM without putting a specialized fan right on it.  The GPU's own fans always will be the key to managing it's temp.  The case fans are on the "don't make this worse" side of the table.  

 

The following is an easy way to set things up.  Everyone has a max fan speed limit they are willing to tolerate for normal  use.  Whatever that value, you want it to be just above your load case temp.  So if you put a temp probe near the rear exhaust and it generally tops out at 55C when gaming, set your max allowed fan speed to 58-60C to start.  Other than that, it's really about noise management.  Nobody drops 5C on their GPU because they increase the fan speed by 200 rpm.  It's not anywhere close to being that sensitive to fan speed changes.  You also likely have normal desktop exit air temp for the probe in the mid to upper 30s and that should be set to your preferred quiet fan speed.  Everything in between can be a straight line.  In practical purposes, this makes the case fans really quiet except when the GPU starts using a lot of power for an extended time.  

 

The radiator fans for the AIO present a choice.  Easy thing to do is move them back to the motherboard so it can properly get a CPU temp.  Quick and mostly self-explanatory.  The other option is to put a Commander Pro temp probe on the exhaust side of the CPU radiator.  There is a direct relationship between the liquid temp and the exhaust air temp, typically as a small offset like +-2C from the actual.  This makes a very effective way to "run from coolant temp" even when you don't have a real sensor.  When you heat up the inside with GPU heat, it also will reflect that and increase speed.  That can be a plus of minus depending on use and case layout, but it's the same as when using any Corsair AIO and the coolant temp within CUE.  

 

These should generate automatically with no user input required.  The most probable is the sensor wire is not fully inserted or defective in some way.  These 10K thermistors are cheap and not too difficult to damage.  It could be the senor on the C-Pro as well, so try moving it to another register like 3 or 4.  

 

I believe it gets it from the GPU driver, much like the rest of the system.  CUE uses the CPUID developer kit to fetch data from other devices.  My CUE is properly obtaining these control variables and I have data, but something is preventing the fan curve from utilizing it.  I do not believe this was a known condition and it's going to be a bit before we get a fix.  There are some separate ongoing issues with AMD GPU temps and past issues with AMD CPU temps.  Obviously if CUE can't get the AMD GPU temp, it won't work for  control regardless so there are some  overlapping issues.  Nevertheless, this fan curve one is not limited to AMD.  It is affecting my Intel CPU and Nvidia GPU.  

In that same vein, I was really hoping that CUE could pull the temp from the water temp sensor on my EKWB distro.  It's nowhere to be found.  HWinfo sees it, but HWMonitor does not.  ASUS Crosshair Hero X670E.   This is using the CUE Link with Hub.

Ideally, I'd like to have my fan curve run on water temp.

Edited October 14, 2023 by Flyprdu2

[c-attack]

3 hours ago, Flyprdu2 said:

was really hoping that CUE could pull the temp from the water temp sensor on my EKWB distro.

Where is the water temp sensor from the EK plate connected?  The motherboard?  Typically these are 2 pin 10K wires just like the Commander Pro.  If so, it can connected directly to the Commander in the same place.  

[Flyprdu2]

I'm not using the Commander Pro/XT.  I've got the new Link Hub.  It has no pins for temp sensor, so it has to be relayed through the motherboard.  Which CUE does not recognize that temp probe.

Is there a way to force an update to recognize T_SENSOR on a Asus Crosshair Mobo?

[c-attack]

1 hour ago, Flyprdu2 said:

I've got the new Link Hub.

No temp probe options there.  The new Hydro X stuff has internal sensors but makes it largely incompatible with most other gear for normal options. On the bright side, you don't have the fan control issue described above.

 

So the T_Sensor, W_Temp, or any of the other names for the MB 2 pin temp connector should show up in CUE under the motherboard, but with Asus and CUE it's going to have a very unhelpful name like ASUSTek CO... ...RO Temp #1-27.  No way to know which temp # belongs to the T sensor except by side by side comparison.  Open up something like HWInfo, HWMonitor, or AI Suite if you're using it and get eyes on that value.  Then comb through the CUE Dashboard presentation of the motherboard to find it's number.  There will be a lot of duplicates and garbage values, which is typical.  Once you find the matching value, you can pull it out of the motherboard stack so it can be it's own widget.  That may keep it from renumerating on subsequent boots.  Now that you know the number, you should be able to find it in the drop down sensor list for any custom fan curve, Commander Pro or XT.  

[Flyprdu2]

I've looked.  I'm pretty sure that it's not there.  As there isn't one that matches exactly. 

You said that CUE pulls from CPUID,  well HWMonitor also does not show the T_Sensor...  but HWInfo does.

This can be solved by software, but it hasn't been resolved yet.

[c-attack]

Hmmm....  yeah that could be the problem.  If the CPUID based HWMon doesn't see it, then CUE is not likely to see it either.  Strange HWMon doesn't pick it up.  It should be on the same level of importance as a fan speed.  

[Flyprdu2]

38 minutes ago, c-attack said:

Hmmm....  yeah that could be the problem.  If the CPUID based HWMon doesn't see it, then CUE is not likely to see it either.  Strange HWMon doesn't pick it up.  It should be on the same level of importance as a fan speed.  

I agree.  HWMonitor has always been "iffy" when it comes to temp sensors.  For Link Hub to be solely reliant on CPUID for temp reporting is not a premium solution for a premium product.

[schmak01]

So you all are aware, there is also a reddit thread on the topic regarding the update not using non-Corsair thermal sensors for fan curves (defaulting to the lowest RPM). Same issue here and sameish solution, just used the thermistor I had placed inside the fins of my 3080ti Hybrid Radiator to control the fan speed for now. 

 

 

[Dondomingo]

So in the meanwhile as I await a fix for the software issue. I did manage to get sensor #1 working (not fully plugged in) and added 2 more sensors, one for the AIO fan exhaust and one for the rear exhaust.