H150i Elite LCD coolant temps

[LeDoyen]

the difference in ease of cooling a cpu and gpu is massive.  With the same power draw my cpu is twice as hot as the gpu.

they are fundamentally different so you can't really compare the H150i with the GPU AIO.

They work the same way but they cool two completely different types of silicon.

[Jos Hideky]

1 hour ago, LeDoyen said:

the difference in ease of cooling a cpu and gpu is massive.  With the same power draw my cpu is twice as hot as the gpu.

they are fundamentally different so you can't really compare the H150i with the GPU AIO.

They work the same way but they cool two completely different types of silicon.

For argument sake, lets make an analogy

stove fire maybe 700-800c comparing with camp fire let assuming 500-600c.

You said CPU die heat area is much more concentrated that GPU die ?

Sorry if misunderstood, but basically this is what u mean in the last post right ?

But is it safe to say both of type of the fire, stove fire or camp fire can boil water up to 100c right ?

Well in this analogy, the CPU cannot boil the water 

Edited February 7, 2023 by Jos Hideky

[LeDoyen]

The CPU can red 100°C diode temp, but the IHS will not be at that temperature. That's in part why people who delid will gail about 15 - 20°C.

Even the thickness of silicon on the back of the chip will have a gradient of temperature.

Let me use your analogy and twist it a bit 🙂

Imagine you have a big *ss cauldron you want to boil. your GPU would be that bonfire, but your CPU would be an acetylene torch.

The bonfire would heat it up pretty quick because it's larger, despite being cooler. The acetylene torch is over 3000°C, yet it's hopeless at heating it up. If anything you'd punch a hole in it, yet it's a lot hotter. It's not the cauldron that doesn't transfer heat well, it's just that you are trying to boil tens of gallons with a small heat source, despite it being hotter.

That's basically what your AIO is doing. you are blasting a small area of the CPU die with over 300W of electrical power, but the AIO is cooling the CPU IHS as much as it can already, so your CPU temperature just runaways up to throttling temperature.

You can use a custom loop waterblock with as many radiators as you like, it won't make much difference. It's not contact based, it's just that the CPU heats up in a lot smaller area compared to a GPU.

Not the whole silicon surface of the CPU is.. CPU. you have a good lot that is just cache memory and IO, that don't heat up much. but the active computing cores are what generate the heat.

under your IHS, the silicon may be 1/3rd of the surface. of that silicon maybe 2/3rd are CPU cores, and you have almost 300W of heat generated there.

On your GPU, it may be 4 or 5 times larger, for the same heat output.

[c-attack]

9 hours ago, Jos Hideky said:

But i actually assuming the coolant temp have bigger delta temp than only 4-5c, considering the delta for the CPU is 60-70c.

You keep viewing this as a negative rather than a positive. Turn off your fans. Watch the coolant temp rise. See what happens. Incidentally coolant to cpu temp differential is constant for a specific load/instruction. If want to analyze, use something with a constant load like the CPU-z bench or OCCT linpack. It might be a good idea to start using the non-AVX cpu-Z mark anyway since you are clearly hitting the limit. If you have a contact issue, the milder cpu-z mark will also instantly hit the max and throttle. If it is substantially lower, you are looking at settings issues. 
 

9 hours ago, Jos Hideky said:

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

Apples and oranges as described above. You can’t compare your cpu temp to your gpu temp. Watts into the radiator and watts dissipated can be compared. 
 

 

If you start a cpu test and you instantly hit 100C or some other unacceptable temperature, either you don’t have good conduct (poor temperatures in non-load conditions as well) or you have too much voltage. Even if you had a wall sized radiator you can’t run 1.50v+ on the cpu. The ability to move the heat elsewhere is not relevant when the instant load temperature is critical. This part of the cooling is always conductive. 

Edited February 7, 2023 by c-attack

[Jos Hideky]

2 hours ago, LeDoyen said:

The CPU can red 100°C diode temp, but the IHS will not be at that temperature. That's in part why people who delid will gail about 15 - 20°C.

Even the thickness of silicon on the back of the chip will have a gradient of temperature.

Let me use your analogy and twist it a bit 🙂

Imagine you have a big *ss cauldron you want to boil. your GPU would be that bonfire, but your CPU would be an acetylene torch.

The bonfire would heat it up pretty quick because it's larger, despite being cooler. The acetylene torch is over 3000°C, yet it's hopeless at heating it up. If anything you'd punch a hole in it, yet it's a lot hotter. It's not the cauldron that doesn't transfer heat well, it's just that you are trying to boil tens of gallons with a small heat source, despite it being hotter.

That's basically what your AIO is doing. you are blasting a small area of the CPU die with over 300W of electrical power, but the AIO is cooling the CPU IHS as much as it can already, so your CPU temperature just runaways up to throttling temperature.

You can use a custom loop waterblock with as many radiators as you like, it won't make much difference. It's not contact based, it's just that the CPU heats up in a lot smaller area compared to a GPU.

Not the whole silicon surface of the CPU is.. CPU. you have a good lot that is just cache memory and IO, that don't heat up much. but the active computing cores are what generate the heat.

under your IHS, the silicon may be 1/3rd of the surface. of that silicon maybe 2/3rd are CPU cores, and you have almost 300W of heat generated there.

On your GPU, it may be 4 or 5 times larger, for the same heat output.

Nice and better analogies, yes, i agree, now how to turn that torch light so can transfer heat more efficiently ?

[Jos Hideky]

7 minutes ago, c-attack said:

You keep viewing this as a negative rather than a positive. Turn off your fans. Watch the coolant temp rise. See what happens. Incidentally coolant to cpu temp differential is constant for a specific load/instruction. If want to analyze, use something with a constant load like the CPU-z bench or OCCT linpack. It might be a good idea to start using the non-AVX cpu-Z mark anyway since you are clearly hitting the limit. If you have a contact issue, the milder cpu-z mark will also instantly hit the max and throttle. If it is substantially lower, you are looking at settings issues. 

Apples and oranges as described above. You can’t compare your cpu temp to your gpu temp. Watts into the radiator and watts dissipated can be compared. 

If you start a cpu test and you instantly hit 100C or some other unacceptable temperature, either you don’t have good conduct (poor temperatures in non-load conditions as well) or you have too much voltage. Even if you had a wall sized radiator you can’t run 1.50v+ on the cpu. The ability to move the heat elsewhere is not relevant when the instant load temperature is critical. This part of the cooling is always conductive. 

Yes its negative thing, the bottleneck should be from the radiator to the surrounding, if the bottleneck is the IHS to the AIO, then its useless to have big radiator, maybe 2x120mm is enough, or even maybe 1x120mm is enough. 

This is why i kind of disappointed when the CPU is still throttling when buying H150i, now i know ...
But from i see from the other youtuber when they stress test the 13900k, its not throttling, and they dont mention about controlling the voltage like you recommending me.

 

[c-attack]

3 minutes ago, Jos Hideky said:

Yes its negative thing, the bottleneck should be from the radiator to the surrounding

You’re welcome to tell the universe it does not understand physics. As discussed earlier, the radiator size does not matter if your instant load cpu temp is too high. It does matter for continued use when under the cpu temp limit. While three vastly different cooler sizes will all start at the same temp, the smaller ones will continue to show a slow and steady climb as watts in exceeds watts dissipated. The cpu temp will slowly increase. The massive wall sized cooling system might never increase from the initial cpu temp and it as able to dump all the heat in one pass. A 360mm is decent sized radiator. It could take 10min or more of continual load for it to find a max coolant temp rise for a specific watt load. Not relevant when you hit 100C 5 seconds in and the cpu throttles. 

[Jos Hideky]

1 hour ago, c-attack said:

You’re welcome to tell the universe it does not understand physics. As discussed earlier, the radiator size does not matter if your instant load cpu temp is too high. It does matter for continued use when under the cpu temp limit. While three vastly different cooler sizes will all start at the same temp, the smaller ones will continue to show a slow and steady climb as watts in exceeds watts dissipated. The cpu temp will slowly increase. The massive wall sized cooling system might never increase from the initial cpu temp and it as able to dump all the heat in one pass. A 360mm is decent sized radiator. It could take 10min or more of continual load for it to find a max coolant temp rise for a specific watt load. Not relevant when you hit 100C 5 seconds in and the cpu throttles. 

You misunderstand me, i told u its a negative thing because 3x120mm is useless, Corsair shouldn't design something that is transfer heat IHS->AIO is smaller than the AIO->Surrounding air, because as u said, its useless.

And i already tried 10 min or more at full load and the coolant temp delta is only 5c at most comparing idle.

even at 10-20% load the cpu temp already 70-80c and coolant temp is still at delta 3c at most

 

Edited February 7, 2023 by Jos Hideky

[LeDoyen]

1 hour ago, Jos Hideky said:

Yes its negative thing, the bottleneck should be from the radiator to the surrounding, if the bottleneck is the IHS to the AIO, then its useless to have big radiator, maybe 2x120mm is enough, or even maybe 1x120mm is enough. 

This is why i kind of disappointed when the CPU is still throttling when buying H150i, now i know ...
But from i see from the other youtuber when they stress test the 13900k, its not throttling, and they dont mention about controlling the voltage like you recommending me.

 

Normally, youtubers test the CPU under Intel specs.

When fully unlocke,d they label it as overclocked, to see its balls to the wall performance, but that is usually at the verge of throttling, and manually tweaked.

The problem you have is some motherboard manufacturers do follow Intel guidance, some do not and unlock power limits by default.

What you are facing right now is not a H150i that fails to cool a 13900k, it's a H150i that fails to cool a heavily overclocked 13900k. It just can't do that unless you manually tweak the overclock. It's just a shame that you have your CPU overclocked without having chosen to do so, it was basically forced by the motherboard bios settings they have out of the box. Then people come thinking their cooler is damaged, or that CPU is insanely inefficient. Turns out neither are true, it's just marketing tricks from motherboards manufacturers to sell "better performing" boards.

Now you know, it's at the cost of heat.

 

Now, to transform that torch to transfer heat more efficiently, again ^^ you hacve to reduce power limits. It will run much cooler, and may even perform better.

[c-attack]

I am at an end with this.  If you don't understand the science at this point in the conversation, it is not likely I can teach it to you in a few more posts.  However, that should not stop you from addressing this in a more scientific fashion

 

1) If you get a test result that seems strange or out of bounds, repeat to confirm and then try a different test.  One was suggested previously with non-AVX instructions.  Poor contact will be evident in all uses.  It is not likely someone sold you a piece of non-conductive metal block.  Look at the other two areas mentioned.

 

2) Search for other sources of information.  You are probably the 10,000,000th person to ask these questions and they are not limited to this CPU and AIO series.  You don't need to take our word for it.  Seek out more information from other sources.  

Edited February 7, 2023 by c-attack

[Jos Hideky]

4 hours ago, c-attack said:

I am at an end with this.  If you don't understand the science at this point in the conversation, it is not likely I can teach it to you in a few more posts.  However, that should not stop you from addressing this in a more scientific fashion

 

1) If you get a test result that seems strange or out of bounds, repeat to confirm and then try a different test.  One was suggested previously with non-AVX instructions.  Poor contact will be evident in all uses.  It is not likely someone sold you a piece of non-conductive metal block.  Look at the other two areas mentioned.

 

2) Search for other sources of information.  You are probably the 10,000,000th person to ask these questions and they are not limited to this CPU and AIO series.  You don't need to take our word for it.  Seek out more information from other sources.  

Im sorry English is not my mother tounge, and even more with this topics.

I just want to convey to you what i mean, i somehow understand what you mean, but you misunderstood my concern.

1. i want to confirm if my AIO installation is correct, now i know its ok. The thermal throttle is not in the AIO but in the CPU die itself.

2. If the bottleneck is heat transfer between IHS to AIO then its useless to have bigger radiator, or even maybe custom water-cooling.

I don't care about the score, i presented the score just as something to consider if the AIO installation is incorrect.

I want to find a way how to efficiently transfer the heat from IHS to any CPU Cooler so it wouldn't thermal throttle, its seems in 13900k case, it wouldn't be possible without lower the voltage.

 

[Jos Hideky]

12 hours ago, LeDoyen said:

Normally, youtubers test the CPU under Intel specs.

When fully unlocke,d they label it as overclocked, to see its balls to the wall performance, but that is usually at the verge of throttling, and manually tweaked.

The problem you have is some motherboard manufacturers do follow Intel guidance, some do not and unlock power limits by default.

What you are facing right now is not a H150i that fails to cool a 13900k, it's a H150i that fails to cool a heavily overclocked 13900k. It just can't do that unless you manually tweak the overclock. It's just a shame that you have your CPU overclocked without having chosen to do so, it was basically forced by the motherboard bios settings they have out of the box. Then people come thinking their cooler is damaged, or that CPU is insanely inefficient. Turns out neither are true, it's just marketing tricks from motherboards manufacturers to sell "better performing" boards.

Now you know, it's at the cost of heat.

Now, to transform that torch to transfer heat more efficiently, again ^^ you hacve to reduce power limits. It will run much cooler, and may even perform better.

Haha the 13900k coming in from the box power to the walls then ?
So is there any CPU cooler that can tame 13900 without reducing power limits ?

[LeDoyen]

it's not the cpu which is the problem but the bios settings.

apart from that, maybe custom loop cooling would get you there but if the cpu gets colder, guess what.. it may boost even higher and thermal throttle at 330 or 340W instead of 310 because it had more cooling headroom.

slapping massive cooling on a freewheeling 13900k is like bandaid on a wooden leg.

Don't blame the cpu or the IHS or anything else. ALL CPUs behave the exact same way when it comes to cooling. it is purely the fault of your motherboard manufacturer thinking it would be a good idea to let a cpu boost freely against its thermal limits instead of following specs.

You can solve your problem in literally 30 seconds by just reducing two power limits in the bios

[Jos Hideky]

5 hours ago, LeDoyen said:

it's not the cpu which is the problem but the bios settings.

apart from that, maybe custom loop cooling would get you there but if the cpu gets colder, guess what.. it may boost even higher and thermal throttle at 330 or 340W instead of 310 because it had more cooling headroom.

slapping massive cooling on a freewheeling 13900k is like bandaid on a wooden leg.

Don't blame the cpu or the IHS or anything else. ALL CPUs behave the exact same way when it comes to cooling. it is purely the fault of your motherboard manufacturer thinking it would be a good idea to let a cpu boost freely against its thermal limits instead of following specs.

You can solve your problem in literally 30 seconds by just reducing two power limits in the bios

I dont want to put out the flame, i want to tame the flame 😁

[LeDoyen]

Then you have only one option :  Chillers, or liquid nitrogen 🙂

There are very recent posts here about people with H170i AIOs with the same problem you have. The only option to deal with excessive VCore is basically sub-ambient cooling.

That's the principle of runaway thermals. You go so far out of spec that it can't be controlled through normal means.

I'm tempted to say your GPU experiences the same thing 😛

80°C on watercooling is pretty bad. that's worse than what an aircooled GPU normally experiences. And Nvidia GPUs thermal throttle above 83°C

[c-attack]

8 hours ago, Jos Hideky said:

is there any CPU cooler that can tame 13900 without reducing power limits ?

No, that’s what we are trying to explain. If you start R23 or any other test and 2-5 seconds later you are at 100C, then you are over your voltage-conductivity limit. Even if I built you that wall sized cooling system you would get the exact same results you are seeing now. The radiator and fans part of the cooling system is like taking out the trash. If you don’t get rid of it, that heat will start to pile up and increase the total cpu temp and ruin your day. But the real cooling on the cpu is conductive — no matter your cooling type. Instantly bad cpu temps are not a cooler problem — unless it is contact related. 
 

8 hours ago, Jos Hideky said:

. If the bottleneck is heat transfer between IHS to AIO then its useless to have bigger radiator, or even maybe custom water-cooling.

Sometimes. A bigger radiator allows you to expel the same amount of wattage with less effort. So in a less critical state, someone running a small 120mm AIO on a 13900K will need to keep the fan in an extremely aggressive state at any load condition to try and keep the coolant down. For a long continuous load like renders, encoding, folding, or stress testing the little 120mm will lose the heat in vs heat out battle and the cpu temp will continue to climb. Most of us can’t manage a +25C increase to cpu temp on a 13900K and that would be on top of the initial load cpu temp. So if your initial load temp was 80C, the temp would slowly increase to 105C over time. Your coolant delta is only +4-5C so that means your AIO is doing a good job on heat in vs heat out. But it also means you can only reduce the cpu temp by 4C with more fan speed. I don’t think you would enjoy having a single fan radiator running at 2400 rpm all the time, so the 360mm is the right choice especially for real world use. If you are gaming, browsing, watching, whatever the fans can stay in a relaxed state with negligible effect on cpu temp. When you get a larger radiator you gain cooling maximum heat capacity (watts), but mostly you get a less noisy system. I run 2x480mm radiators in my combination cpu/gpu system. Most of the time I can’t hear my fans because they only need to run at 1000 rpm to remove most of the heat. However, I can’t run 1.50v either. It will be instantly too hot. 
 

8 hours ago, Jos Hideky said:

want to confirm if my AIO installation is correct, now i know its ok. The thermal throttle is not in the AIO but in the CPU die itself

This is worth looking into. Standard thing to do is take the CPU block off and look at the TIM spread. If you didn’t have good contact, there will be little clumps or lines of paste that did not spread out. Without taking the block off, you can try and assess this by looking at cpu temps. 
 

1) Run a milder stress test like the bench test in cpu-z. It does not have AVX instructions and should produce a lot less wattage than R23. I would expect it to run 15C lower even at your current settings. If it hits 100C right away, then I think there is a contact problem. 
 

2) A contact problem affects heat transfer in all load conditions. Fully quit iCUE and load up something like HWinfo or HWMon that keeps track of all Core temps, current, min, max. If you see temp spikes into the 60-70c range opening browsers, sifting through folders, looking at pictures, then you likely do have a contact issue. Be aware this is not a fool proof strategy. Some games do use AVX instructions to assess your system when you load. I see my peak cpu temp as the games loads, not during the 3 hours of actual play. 
 

3) Lots of talk on Raptor/Kaby Lake about using contact frames instead of the motherboards socket loader. I used from the start so I can’t do a before and after, but my temps are good. However, I will repeat that my R23 temps were in the low 90s straight out of the box with power limits off and 2x480mm radiators. You simply can’t run a lot synthetic tests these days without manual tweaking. 

[PAlvarez05]

Just now, LeDoyen said:

Then you have only one option :  Chillers, or liquid nitrogen 🙂

There are very recent posts here about people with H170i AIOs with the same problem you have. The only option to deal with excessive VCore is basically sub-ambient cooling.

That's the principle of runaway thermals. You go so far out of spec that it can't be controlled through normal means.

I'm tempted to say your GPU experiences the same thing 😛

80°C on watercooling is pretty bad. that's worse than what an aircooled GPU normally experiences. And Nvidia GPUs thermal throttle above 83°C

Wow this post is still going

I have a 13900KS running at 5.9ghz is able to pass OCCT 12hr stress test without issues on an H170i, idles around 26-28c and depending on the load it can go up to 100c, under normal use it doesn’t go pass 70c, sounds like he has unrealistic expectations of the hardware he is currently using and lack knowledge of how to tune his hardware.

4090 water cooled via an AIO, also OC to around 3100mhz that’s never reached 70c regardless of load

I am still trying to figure out what it is he wants? 

[Jos Hideky]

14 hours ago, PAlvarez05 said:

Wow this post is still going

I have a 13900KS running at 5.9ghz is able to pass OCCT 12hr stress test without issues on an H170i, idles around 26-28c and depending on the load it can go up to 100c, under normal use it doesn’t go pass 70c, sounds like he has unrealistic expectations of the hardware he is currently using and lack knowledge of how to tune his hardware.

4090 water cooled via an AIO, also OC to around 3100mhz that’s never reached 70c regardless of load

I am still trying to figure out what it is he wants? 

Nice, how bout your coolant temp temperature ?
What is your power draw ?

When doing Cinebench R23 is the CPU still at 100c ?

[Jos Hideky]

14 hours ago, LeDoyen said:

Then you have only one option :  Chillers, or liquid nitrogen 🙂

There are very recent posts here about people with H170i AIOs with the same problem you have. The only option to deal with excessive VCore is basically sub-ambient cooling.

That's the principle of runaway thermals. You go so far out of spec that it can't be controlled through normal means.

I'm tempted to say your GPU experiences the same thing 😛

80°C on watercooling is pretty bad. that's worse than what an aircooled GPU normally experiences. And Nvidia GPUs thermal throttle above 83°C

Looks like Palvares is doing a lot better than me ..
Im idling at 40c, even opening an app, momentarily load spiking can also spike the CPU temp to 80c.

but still the coolant temp is at 305c

GPU actually even when at 100% load the temp is at 70~80ish.

[Jos Hideky]

14 hours ago, c-attack said:

No, that’s what we are trying to explain. If you start R23 or any other test and 2-5 seconds later you are at 100C, then you are over your voltage-conductivity limit. Even if I built you that wall sized cooling system you would get the exact same results you are seeing now. The radiator and fans part of the cooling system is like taking out the trash. If you don’t get rid of it, that heat will start to pile up and increase the total cpu temp and ruin your day. But the real cooling on the cpu is conductive — no matter your cooling type. Instantly bad cpu temps are not a cooler problem — unless it is contact related. 
 

Sometimes. A bigger radiator allows you to expel the same amount of wattage with less effort. So in a less critical state, someone running a small 120mm AIO on a 13900K will need to keep the fan in an extremely aggressive state at any load condition to try and keep the coolant down. For a long continuous load like renders, encoding, folding, or stress testing the little 120mm will lose the heat in vs heat out battle and the cpu temp will continue to climb. Most of us can’t manage a +25C increase to cpu temp on a 13900K and that would be on top of the initial load cpu temp. So if your initial load temp was 80C, the temp would slowly increase to 105C over time. Your coolant delta is only +4-5C so that means your AIO is doing a good job on heat in vs heat out. But it also means you can only reduce the cpu temp by 4C with more fan speed. I don’t think you would enjoy having a single fan radiator running at 2400 rpm all the time, so the 360mm is the right choice especially for real world use. If you are gaming, browsing, watching, whatever the fans can stay in a relaxed state with negligible effect on cpu temp. When you get a larger radiator you gain cooling maximum heat capacity (watts), but mostly you get a less noisy system. I run 2x480mm radiators in my combination cpu/gpu system. Most of the time I can’t hear my fans because they only need to run at 1000 rpm to remove most of the heat. However, I can’t run 1.50v either. It will be instantly too hot. 
 

This is worth looking into. Standard thing to do is take the CPU block off and look at the TIM spread. If you didn’t have good contact, there will be little clumps or lines of paste that did not spread out. Without taking the block off, you can try and assess this by looking at cpu temps. 
 

1) Run a milder stress test like the bench test in cpu-z. It does not have AVX instructions and should produce a lot less wattage than R23. I would expect it to run 15C lower even at your current settings. If it hits 100C right away, then I think there is a contact problem. 
 

2) A contact problem affects heat transfer in all load conditions. Fully quit iCUE and load up something like HWinfo or HWMon that keeps track of all Core temps, current, min, max. If you see temp spikes into the 60-70c range opening browsers, sifting through folders, looking at pictures, then you likely do have a contact issue. Be aware this is not a fool proof strategy. Some games do use AVX instructions to assess your system when you load. I see my peak cpu temp as the games loads, not during the 3 hours of actual play. 
 

3) Lots of talk on Raptor/Kaby Lake about using contact frames instead of the motherboards socket loader. I used from the start so I can’t do a before and after, but my temps are good. However, I will repeat that my R23 temps were in the low 90s straight out of the box with power limits off and 2x480mm radiators. You simply can’t run a lot synthetic tests these days without manual tweaking. 

Oh ok2, so actually what my received settings in here is for like the over clocker settings  in the liquid nitrogen setups.

Already look into, wipe the stock heat paste and try to use better one, Kryonaut, but actually the result is bit worse after that 😅

Do you recommend using like liquid metal type paste ?

Yes there is a brief spike to 60-70c range when opening applications, even when the cpu usage to tp 20-30% for like 2-3sec the temp also quickly go up to 60-70c but also quickly go down afterthat.

[LeDoyen]

that's courtesy of your 1.45v VID you were showing on HWmonitor

[LeDoyen]

If you've applied a ton of paste, the result will be worse during the time it takes for it to squeeze out 😛

Kryonaut should get you better results, but you probably wouldn't notice it at that stage. Maybe the CPU power will go slightly higher with the added 2 or 3° headroom you would have gained.

the settings you have are not for LN2 overclockers, it's for... nothing, they are just ridiculous 😛

[PAlvarez05]

7 hours ago, Jos Hideky said:

Nice, how bout your coolant temp temperature ?
What is your power draw ?

When doing Cinebench R23 is the CPU still at 100c ?

You want better temps, buy better hardware, simple as that.