H150i Elite LCD coolant temps

[Jos Hideky]

1 minute ago, PAlvarez05 said:

The 13900k is a beast is not meant for a 360 AIO… 

This is i agree, what make me think its contact issue because coolant temp is not going up no matter what ...

[PAlvarez05]

Just now, Jos Hideky said:

stable at 100c  is not thermal throttle ?

if cooler cant handle the CPU, i think we will see high coolant temp right ?

From what it looks like your frequency isn’t dropping so no, you are not thermal throttling

like I said before your coolant temps won’t go up because your cold plate is unable to efficiently transfer the heat into the coolant, hence why they aren’t going up 

high idle temps are due to a high vcore and unlocked power limit 

[Jos Hideky]

1 minute ago, PAlvarez05 said:

From what it looks like your frequency isn’t dropping so no, you are not thermal throttling

like I said before your coolant temps won’t go up because your cold plate is unable to efficiently transfer the heat into the coolant, hence why they aren’t going up 

high idle temps are due to a high vcore and unlocked power limit 

Oh ok understand, even stable at 100c is not thermal throttle because the clock is still high ...
Cold plate = IHS ? So Delidding ?
Or maybe use better thermal paste or even the liquid metal one ?

[PAlvarez05]

You don’t know what a vcore is and you want to delid?

you need to crawl before you fly my brother 

you will end up damaging your hardware, if you have money and time to spare, start with a custom loop

[LeDoyen]

it is probably thermal throttling already because it's bumping flat on its thermal limit, as you can see on the curve. Maybe it's not throttling much but looks like it does.

For a balls-to-the-wall, no limits overclock as it is now, an AIO is insufficient. So you have to reduce its power limits.

The cold plate is the copper plate under the AIO head, that contacts the CPU. The IHS is the intergrated heat spreader. It's the metal piece you remove when delidding.

If you are not comfortable with tweaking an overclock, you shouldn't try delidding either ^^'

All you need is to go in bios and find the power limits to enter lower numbers; like 200 - 250W max, and see what it does under Cinebench.. then reduce to taste. It's easy and takes just a few minutes. no complex settings to tweak, the CPU will deal with it.

[Jos Hideky]

6 minutes ago, PAlvarez05 said:

You don’t know what a vcore is and you want to delid?

you need to crawl before you fly my brother 

you will end up damaging your hardware, if you have money and time to spare, start with a custom loop

Dont have to be me right, i can ask people that understand to do that for me ..
Is custom loop the solution ?
Is it  also going  meet with the same problem -> cold plate is unable to efficiently transfer the heat into the coolant

The custom loop is the answer if the coolant temp is so high and the radiator cannot dissipate the heat into the air right ?

CMIIW

Edited February 5 by Jos Hideky

[Jos Hideky]

4 minutes ago, LeDoyen said:

it is probably thermal throttling already because it's bumping flat on its thermal limit, as you can see on the curve. Maybe it's not throttling much but looks like it does.

For a balls-to-the-wall, no limits overclock as it is now, an AIO is insufficient. So you have to reduce its power limits.

The cold plate is the copper plate under the AIO head, that contacts the CPU. The IHS is the intergrated heat spreader. It's the metal piece you remove when delidding.

If you are not comfortable with tweaking an overclock, you shouldn't try delidding either ^^'

All you need is to go in bios and find the power limits to enter lower numbers; like 200 - 250W max, and see what it does under Cinebench.. then reduce to taste. It's easy and takes just a few minutes. no complex settings to tweak, the CPU will deal with it.

Yes i think so too, its constantly thermal throttling at that clock speed, but when i compare the cinebench result with other, its around there, 38k. 

I just feel really strange no matter what happen the coolant temp still at 40c ..
So the bottle neck in the heat transfer is from CPU to AIO
Not AIO to the surrounding.

Edited February 5 by Jos Hideky

[Jos Hideky]

Anyway thank you for Ledoyen and Palvarez for your knowledge and input, very much appreciated 👍

[LeDoyen]

yes exactly. but that bottleneck happens with CPU regardless of the cooler when Vcore is excessive.

There's nothing wrong with the cooler, it dissipates whatever it receives right now, but the CPU is just overstressed for nothing. You could probably hit the same speeds with less temperature.

And there's no saying under what conditions others hit that 38k score.

I'm ready to bet if you reduced power limits to 240 - 250W, the CPU wouldn't overheat and you may not even see a hit in cinebench perfs.

 

[Jos Hideky]

4 minutes ago, LeDoyen said:

yes exactly. but that bottleneck happens with CPU regardless of the cooler when Vcore is excessive.

There's nothing wrong with the cooler, it dissipates whatever it receives right now, but the CPU is just overstressed for nothing. You could probably hit the same speeds with less temperature.

And there's no saying under what conditions others hit that 38k score.

I'm ready to bet if you reduced power limits to 240 - 250W, the CPU wouldn't overheat and you may not even see a hit in cinebench perfs.

 

Ok where can you see the Vcore to know its excessive ?

Ow and the CPU can live longer ?
Just change it in the BIOS to 240W ?
Let me try ..

[PAlvarez05]

9 minutes ago, LeDoyen said:

yes exactly. but that bottleneck happens with CPU regardless of the cooler when Vcore is excessive.

There's nothing wrong with the cooler, it dissipates whatever it receives right now, but the CPU is just overstressed for nothing. You could probably hit the same speeds with less temperature.

And there's no saying under what conditions others hit that 38k score.

I'm ready to bet if you reduced power limits to 240 - 250W, the CPU wouldn't overheat and you may not even see a hit in cinebench perfs.

 

Well said 

[Jos Hideky]

1 minute ago, PAlvarez05 said:

Well said 

Do you also power limit your CPU ?

[PAlvarez05]

4 minutes ago, Jos Hideky said:

Do you also power limit your CPU ?

Nope im a balls to the walls overclocker 

[PAlvarez05]

6 minutes ago, Jos Hideky said:

Ok where can you see the Vcore to know its excessive ?

Ow and the CPU can live longer ?
Just change it in the BIOS to 240W ?
Let me try ..

What is it under load?

[LeDoyen]

3 minutes ago, Jos Hideky said:

Ok where can you see the Vcore to know its excessive ?

Vcore and power are directly related. the higher the Vcore, the higher the power draw. That's why people undervolt their GPUs for example, reduce power = reduce temps.

The problem is the ideal Vcore for a given CPU varies. if you were lucky to have good silicon, you may not need much to hit 5.7ghz. if you are unlucky you may need to pump so much voltage in that the thing will draw 50 - 60 W more than the good CPU, or it may not even reach the frequency you're after.

Because of that variability, motherboard manufacturers, when they unlock power limits, set the bios to use fairly high vcore to make sure even bad silicon will remain stable.

That's why leaving the thing fully unlocked is usually not a good idea.  At worst, you can reduce power limits, and as i said the CPU will conservatively reduce its vcore to adjust, or you can really fine tune the CPU by hand, to determine by trial and errror where its sweet spots are in terms of voltage, LLC, power and other good stuff, to get the absolute most of it with the cooler you have.

That's why i keep saying "reduce the power limits" 🙂 it's a fast way to get there if you don't know how to tweak a 13th gen CPU. (I don't , i am still on 10th gen!)

[Jos Hideky]

3 minutes ago, PAlvarez05 said:

What is it under load?

This one on cinebench just for like 5 sec, is enough ?

[PAlvarez05]

Just now, Jos Hideky said:

This one on cinebench just for like 5 sec, is enough ?

Yeap is pretty high 

[Jos Hideky]

7 minutes ago, LeDoyen said:

Vcore and power are directly related. the higher the Vcore, the higher the power draw. That's why people undervolt their GPUs for example, reduce power = reduce temps.

The problem is the ideal Vcore for a given CPU varies. if you were lucky to have good silicon, you may not need much to hit 5.7ghz. if you are unlucky you may need to pump so much voltage in that the thing will draw 50 - 60 W more than the good CPU, or it may not even reach the frequency you're after.

Because of that variability, motherboard manufacturers, when they unlock power limits, set the bios to use fairly high vcore to make sure even bad silicon will remain stable.

That's why leaving the thing fully unlocked is usually not a good idea.  At worst, you can reduce power limits, and as i said the CPU will conservatively reduce its vcore to adjust, or you can really fine tune the CPU by hand, to determine by trial and errror where its sweet spots are in terms of voltage, LLC, power and other good stuff, to get the absolute most of it with the cooler you have.

That's why i keep saying "reduce the power limits" 🙂 it's a fast way to get there if you don't know how to tweak a 13th gen CPU. (I don't , i am still on 10th gen!)

But isn't it reduce the power limit is also a form of tweaking the CPU ?😅

[LeDoyen]

it is, but it's just two fields to change 🙂 PL1 and PL2 limits.

tweaking an overclock is much more than that 😛 and you're kind of "un-tweaking" what those idiots at asus/MSI/gygabyte did to the intel limits to make sure everyone has a cooking CPU.

[c-attack]

You can use a 360mm radiator with a 13900K. The expected coolant rise for 300W on a 360mm radiator with fans at 1300rpm is around +10C to coolant, which means +10C to cpu. If it were actually possible to instantly remove all that heat in a single pass with super magical fans you would only drop 10C. I expect 90C is not what you had in mind either. All of that is voltage related and no cooler can do anything about that.  You have to address it on the power-voltage side. 
 

If I put an air cooler, a small 120mm AIO, and a massive 1 meter external cooling array on your CPU, they are all going to hit 90C+ in the first few seconds. The cooler gets rid of the heat conducted through the cpu. If you don’t get rid of that, things will slowly get worse. That is something that happens as the test continues for minutes/hours.  However, it can’t keep the cpu from getting hot in the first place. The air cooler has a lower heat capacity and will start to show a slow increase almost immediately. The little 120mm AIO can hang on a bit longer before slowly starting to increase. The theoretical 1m square external array may be able to blow off all the heat added on pickup on each pass through the radiator and it will not increase, but it will still hold at the 92C you started at. 

If you are pulling 310W running R23 your motherboard limits are off and the bios is liberally applying voltage. Both are responsible for current temps. Many motherboards come out of the box in this state in order to boost review scores on common metrics. You can’t run things like R23 in this state and you are likely hurting your score. I can run mine at 5.8x8 and the 253W limit on and beat your score. You need to turn it down to improve. 
 

First thing I would do is turn the Intel limits back on. The Raptor Lakes do not necessarily do better with unlimited power and you can maintain boosted clocks for longer with lower temps. The next step should to be rein in your Vcore, but it’s possible with the limits back in place that will take care of itself. 

Edited February 6 by c-attack

[Jos Hideky]

14 hours ago, LeDoyen said:

it is, but it's just two fields to change 🙂 PL1 and PL2 limits.

tweaking an overclock is much more than that 😛 and you're kind of "un-tweaking" what those idiots at asus/MSI/gygabyte did to the intel limits to make sure everyone has a cooking CPU.

Hahaa nice thnx again for the share, really appreciate it 👍

[Jos Hideky]

11 hours ago, c-attack said:

You can use a 360mm radiator with a 13900K. The expected coolant rise for 300W on a 360mm radiator with fans at 1300rpm is around +10C to coolant, which means +10C to cpu. If it were actually possible to instantly remove all that heat in a single pass with super magical fans you would only drop 10C. I expect 90C is not what you had in mind either. All of that is voltage related and no cooler can do anything about that.
 

If I put an air cooler, a small 120mm AIO, and a massive 1 meter external cooling array on your CPU, they are all going to hit 90C+ in the first few seconds. The cooler gets rid of the heat conducted through the cpu. If you don’t get rid of that, things will slowly get worse. That is something that happens as the test continues for minutes/hours.  However, it can’t keep the cpu from getting hot in the first place. The air cooler has a lower heat capacity and will start to show a slow increase almost immediately. The little 120mm AIO can hang on a bit longer before slowly starting to increase. The theoretical 1m square external array may be able to blow off all the heat added on pickup on each pass through the radiator and it will not increase, but it will still hold at the 92C you started at. 

If you are pulling 310W running R23 your motherboard limits are off and the bios is liberally applying voltage. Both are responsible for current temps. Many motherboards come out of the box in this state in order to boost review scores on common metrics. You can’t run things like R23 in this state and you are likely hurting your score. I can run mine at 5.8x8 and the 253W limit on and beat your score. You need to turn it down to improve. 
 

First thing I would do is turn the Intel limits back on. The Raptor Lakes do not necessarily do better with unlimited power and you can maintain boosted clicks for longer with lower temps. The next step should to be rein in your Vcore, but it’s possible with the limits back in place that will take care of itself. 

Im actually not really into the score competition, as im only casual user, the most demanding things i do is only games, which is not so much cpu demanding as GPU demanding
Other than that is youtube, office application and browsing

What im the most curious is, whatever the CPU temp is, whether is 40c idle, 60c light load or 100c full load
The coolant temp is still at 35~41c, meaning what ever the cpu temp, if the coolant temp is not changing, then there is no heat transfer.
Make me think there is contact issue.

Or the other explanation is the heat transfer from the CPU to the AIO is always much less then the heat transfer from the AIO (Via the radiator) to the surrounding.
So no matter how much the heat transfer to the AIO, the AIO can dissipate the heat faster hence the coolant temp remain stable.

But actually when i feel it with my hand, the radiator from my GPU (using 4090 suprim liquid with a radiator) is emanating much more heat, than the radiator H150i, meaning that suprim AIO is much more effective at transferring heat than H150i. Which is maybe again, because the H150i contact problem ?

What do you think ?

 

 

[LeDoyen]

2 hours ago, Jos Hideky said:

But actually when i feel it with my hand, the radiator from my GPU (using 4090 suprim liquid with a radiator) is emanating much more heat, than the radiator H150i, meaning that suprim AIO is much more effective at transferring heat than H150i. Which is maybe again, because the H150i contact problem ?

It's more related to the type of chip and contact with the cooler than on cooler efficiency actually.

GPUs have a lower transistor density (they are BIG silicon dies), and hence a lot easier to cool. Also the AIO on that card directly contacts the silicon, there is no IHS on a GPU, that's two big advantages it has over a CPU.

So, yes it transfers heat more efficiently because it's direct die, but apart from that, the power/temperature difference you see with your GPU is due to the silicon, and not to any contact issue you may have on the CPU.

If you delidded the CPU, you would drop 10 - 15C at best at the same power level. But you may still throttle because the motherboard would then let the CPU pull even more power (and get a few hundred more points on Cinebench).

 

Just to give you an idea about how GPUS and CPUs are cooled, when i run my GPU to 100% (doing some folding@home), it reaches 36°C with a water temp of 26°C . Mind you it's a 3090, at 370W. only 10°C difference

If i do the same to the CPU it will reach about 80°C when overclocked (~300W), yet its block isn't less efficient. it's just that the heat is a lot more focused on CPUs and the die temperature will always be hotter, even if you do direct die with liquid metal.

Oh, and dialing the CPU 100mhz slower gets me a CPU temp of 66°C and only 220W with almost no performance hit. It makes a tiny difference in Vcore voltage, but that translates to 80W saved, 

 

[c-attack]

6 hours ago, Jos Hideky said:

What im the most curious is, whatever the CPU temp is, whether is 40c idle, 60c light load or 100c full load
The coolant temp is still at 35~41c, meaning what ever the cpu temp, if the coolant temp is not changing, then there is no heat transfer.

You are assuming the temperature of the silicon should be the same as the water in the radiator. The water/coolant is a heat transport mechanism to take heat to the radiator for dispersal. Easiest analogy is a pot on the stove. You set the flame to high and 400F degrees. The water is never going to be 400F but is transmitting the heat into both the room and whatever food you toss in the pot. The flame, the bottom of the pot, and the water will not all be the same temperature. Your coolant is not going to match the cpu temp except when you turn off the flame (Cpu Vcore = 0v), but you will get pretty close at idle. 
 

Your coolant temperature change is a very mathematical watts in from the cpu, less watts dispersed through the radiator and fans. If want to see your coolant temp rise a whole bunch, turn off the fans. Now you are adding heat and not releasing any except for some natural radiated heat off the metal fins. It is a heat transport system and the change in coolant temperature reflects the amount of heat you’ve added to the water. There are clearly defined numbers for temperature change at a specific watts in, on radiator size A, and specific fan speeds.
 

Coolant temp is the lowest possible cpu temp with zero volts and thus +1 to coolant is +1 to cpu temp by raising the baseline. Same thing in reverse for cooling. As mentioned above, the expected coolant temp rise at 300W and 1300 rpm is going to be about 10C +-1C. You can make an assessment of your capabilities by looking at the difference between the coolant temp and the load cpu temp. In your case that’s +60C. That’s about the maximum sustainable differential. In order to get you to 80C in R23 at those settings, you would need to have a max coolant of 20C and thus a room temp of 10C. Obviously that’s not viable, but if you’re curious what extreme overclockers do, that’s it. They can’t get around the heat under the cpu problem either, so the way to deal with a +140C coolant to cpu differential is to lower the socket temperature to -60C with liquid nitrogen. Since you are not going to be doing that, you need to turn down the flame. 
 

R23 is not like anything you are going to use your PC for and you simply could forgo the heavy synthetic testing and do normal things. However, you would be better served putting the limits back on and fine tuning your voltage so you can maintain higher cpu clocks for longer by keeping the cpu temp down. You will get better performance this way and these CPUs have complex programmed behavior in specific temperature zones. 
 

Your gpu is different than a cpu as explained above, however the heat dissipation process is the same. The 4090 is outputting quite a bit more than your cpu and I’m sure you are regularly near the 450W limit. Then you are dispersing it through a 240mm radiator with less cooling surface area than your 360mm cpu radiator. 450W on 240mm means higher coolant temp than 300W on 360mm. Exhaust air temp from the radiator has a direct relationship to the coolant temp (small offset), so yes… your gpu exhaust feels hotter. It is because you are applying more heat to a smaller radiator that can’t get rid of it as well as the 360mm. 

Edited February 6 by c-attack

[Jos Hideky]

14 hours ago, c-attack said:

You are assuming the temperature of the silicon should be the same as the water in the radiator. The water/coolant is a heat transport mechanism to take heat to the radiator for dispersal. Easiest analogy is a pot on the stove. You set the flame to high and 400F degrees. The water is never going to be 400F but is transmitting the heat into both the room and whatever food you toss in the pot. The flame, the bottom of the pot, and the water will not all be the same temperature. Your coolant is not going to match the cpu temp except when you turn off the flame (Cpu Vcore = 0v), but you will get pretty close at idle. 
 

Your coolant temperature change is a very mathematical watts in from the cpu, less watts dispersed through the radiator and fans. If want to see your coolant temp rise a whole bunch, turn off the fans. Now you are adding heat and not releasing any except for some natural radiated heat off the metal fins. It is a heat transport system and the change in coolant temperature reflects the amount of heat you’ve added to the water. There are clearly defined numbers for temperature change at a specific watts in, on radiator size A, and specific fan speeds.
 

Coolant temp is the lowest possible cpu temp with zero volts and thus +1 to coolant is +1 to cpu temp by raising the baseline. Same thing in reverse for cooling. As mentioned above, the expected coolant temp rise at 300W and 1300 rpm is going to be about 10C +-1C. You can make an assessment of your capabilities by looking at the difference between the coolant temp and the load cpu temp. In your case that’s +60C. That’s about the maximum sustainable differential. In order to get you to 80C in R23 at those settings, you would need to have a max coolant of 20C and thus a room temp of 10C. Obviously that’s not viable, but if you’re curious what extreme overclockers do, that’s it. They can’t get around the heat under the cpu problem either, so the way to deal with a +140C coolant to cpu differential is to lower the socket temperature to -60C with liquid nitrogen. Since you are not going to be doing that, you need to turn down the flame. 
 

R23 is not like anything you are going to use your PC for and you simply could forgo the heavy synthetic testing and do normal things. However, you would be better served putting the limits back on and fine tuning your voltage so you can maintain higher cpu clocks for longer by keeping the cpu temp down. You will get better performance this way and these CPUs have complex programmed behavior in specific temperature zones. 
 

Your gpu is different than a cpu as explained above, however the heat dissipation process is the same. The 4090 is outputting quite a bit more than your cpu and I’m sure you are regularly near the 450W limit. Then you are dispersing it through a 240mm radiator with less cooling surface area than your 360mm cpu radiator. 450W on 240mm means higher coolant temp than 300W on 360mm. Exhaust air temp from the radiator has a direct relationship to the coolant temp (small offset), so yes… your gpu exhaust feels hotter. It is because you are applying more heat to a smaller radiator that can’t get rid of it as well as the 360mm. 

No not actually, im not assuming silicon temp will be same like the coolant temp.
But i actually assuming the coolant temp have bigger delta temp than only 4-5c, considering the delta for the CPU is 60-70c.

And actually, the temp on the GPU have way more controlled temp only reaching 70-80c even at full load with 240mm, as the CPU temp reaching 80-90c even at 50-60% load with 360mm. But worth also considering that suprim is actually is hybrid air and water aoi.

If the heat dissipation process the same then the suprim liquid here is better dissipating heat than H150i, unfortunately suprim doesnt have monitor for coolant temp in their software, so i dont know how big is the delta temp inside the radiator.

i dont know about how the heat concentration on the CPU comparing to the GPU going to affect the heat transfer to the coolant temp, but i think its going to be minimal ?

im assuming that even how the silicon got hot is different between CPU dan GPU, isnt it its going to heat the coolant temp all the same ?

Let me just try to reapply thermal in the H150i and see where it goes ...