Question High temps with 7700k and NZXT Kraken X72 Push/Pu

[xHawaiiToast]

Hello everyone,

this is my first post and I will try to present my problem as good as I can.

My Specs:

Intel i7 7700k @4,6/4,8 Ghz // 1,25/1,3V
NZXT Kraken X72 Push/Pull
32Gb G.Skill RipJaws V @ 3200Mhz // 1,353V
MSI Geforce Gtx 1060 Gaming X 6Gb
Asus ROG Strix Z270E Gaming
500Gb SanDisk Extreme Pro
500Gb Samsung 850 Evo
256Gb Samsung 850 Evo
4Tb Seagate Barracuda
650 Watt be quiet! Dark Power Pro 11
Creative Soundblaster Z
Corsair Crystal 570X

Fans:
1x Corsair LL120 @ the back of the case
2x Corsair LL120 @ the top of the case
3x Corsair LL120 @ the front of the case, pushing air into the radiator
3x NZXT Aer P120 behind the radiator, pulling air out of the radiator

The air is getting pulled from the front, through the radiator and the top and is getting exhausted out of the back.

So my main problem is about thermals and how they interact with my fan speed. I can only run aida64 or prime95 @4,2Ghz all the time without a boost clock if I want the temps to stay lower than 80°C or if I want prime95 to work at all.

I wanted to overclock my i7 to at least 4,8Ghz, but I'm getting temps of 100°C max on all cores after running aida64 for a few minutes.
I can't even use prime95 because it always gives out that a Fatal Error has occured where a 0,4 was expected, but a 0,5 or someting came out. I just don't understand why. I tried paying around with the CPU voltage settings, but it didn't change anything.

After I saw this high temps I decided to reduce the clockspeed to 4,6 Ghz at 1,25 Volts but the temps were still above 90°C at all cores. A small improvent, but not enough.

There's actually a pretty big difference between the temps shown by NZXT CAM and HWMonitor. I use both at the same time and the difference is more than 10°C sometimes. Could it be some sort of sensor problem?

Now the strangest thing: It almost doesnt't matter at all at which fan speeds I run my system. When I stress tested my CPU with Aida64 at around 40% fan speed and at 100% fan speed, the results were more or less the same. I also ran a Cinebench R20 Benchmark and got a Singecore score of 489 and a Multicore score of 2419 with about 40% fan speed. But when I ran the same test with 100% fan speed both scores were a little bit worse instead of getting better. I don't get it.

Edit: After a user told me that the case limits my front airflow I removed the glass panel and did all the tests again. It didn´t change anything for the temps.
I also monitored my cpu temps with AiSuite from Asus and now I´m more confused than I was before. AiSuite gives out a max temperatur of 73°C while letting aida64 work and there are no big spikes anymore as with HWMonitor or Aida64.
What are the real temps now? Aida64 with 95°C max or AiSuite with 73°C max or NZXT CAM with 80°C max. I´m just confused.

Fluid Temp: https://drive.google.com/file/d/1_WUA2WlXVHIfcwkEN8dxYU5cX8herPmn/view?usp=sharing

AIDA64 run: https://drive.google.com/file/d/14LZpknr5nyge-Y48ts7giNiqs_OOEQ8Q/view?usp=sharing

CAM Temps: https://drive.google.com/file/d/1B0R1mNEgeKS6IGvgDXlgxL9NudFrYXgg/view?usp=sharing

AiSuite CPU Temps:https://drive.google.com/file/d/1yLmDQMAJ1aS1Tj0n_BAfi_FL4MJj-fBg/view?usp=sharing

 

[bmockeg]

Hello everyone,

this is my first post and I will try to present my problem as good as I can.

My Specs:

Intel i7 7700k @4,6/4,8 Ghz // 1,2/1,3V
NZXT Kraken X72 Push/Pull
32Gb G.Skill RipJaws V @ 3200Mhz // 1,353V
MSI Geforce Gtx 1060 Gaming X 6Gb
Asus ROG Strix Z270E Gaming
500Gb SanDisk Extreme Pro
500Gb Samsung 850 Evo
256Gb Samsung 850 Evo
4Tb Seagate Barracuda
650 Watt be quiet! Dark Power Pro 11
Creative Soundblaster Z
Corsair Crystal 570X

Fans:
1x Corsair LL120 @ the back of the case
2x Corsair LL120 @ the top of the case
3x Corsair LL120 @ the front of the case, pushing air into the radiator
3x NZXT Aer P120 behind the radiator, pulling air out of the radiator

The air is getting pulled from the front, through the radiator and the top and is getting exhausted out of the back.

So my main problem is about thermals and how they interact with my fan speed. I can only run aida64 or prime95 @4,2Ghz all the time without a boost clock if I want the temps to stay lower than 80°C or if I want prime95 to work at all.

I wanted to overclock my i7 to at least 4,8Ghz, but I'm getting temps of 100°C max on all cores after running aida64 for a few minutes.
I can't even use prime95 because it always gives out that a Fatal Error has occured where a 0,4 was expected, but a 0,5 or someting came out. I just don't understand why. I tried paying around with the CPU voltage settings, but it didn't change anything.

After I saw this high temps I decided to reduce the clockspeed to 4,6 Ghz at 1,2 Volts but the temps were still above 90°C at all cores. A small improvent, but not enough.

There's actually a pretty big difference between the temps shown by NZXT CAM and HWMonitor. I use both at the same time and the difference is more than 10°C sometimes. Could it be some sort of sensor problem?

Now the strangest thing: It almost doesnt't matter at all at which fan speeds I run my system. When I stress tested my CPU with Aida64 at around 40% fan speed and at 100% fan speed, the results were more or less the same. I also ran a Cinebench R20 Benchmark and got a Singecore score of 489 and a Multicore score of 2419 with about 40% fan speed. But when I ran the same test with 100% fan speed both scores were a little bit worse instead of getting better. I don't get it.

Did you just install the x72? Sounds like you need to check the installation as it can’t be getting contact correctly for those temps. Is the pump set to full speed?

 

[xHawaiiToast]

Did you just install the x72? Sounds like you need to check the installation as it can’t be getting contact correctly for those temps. Is the pump set to full speed?

No I installed it about half a year ago and didn't have any trouble until I started to overclock a bit more. The cooler is sitting on top of the cpu with no movement or something like that at all.
The pump is at 100% almost all the time and I can definetly hear it work

I even tried to loosen the cpu block up a bit and run the same tests to see if there is any difference, but there is almost none. After that I screwed everything back in tightly and the temps were the same.

 

[xHawaiiToast]

I just received my new thermal paste (Arctic MX-4) and reinstalled the pump block.

I took some pictures of the already applied thermal paste:
https://drive.google.com/file/d/1eRPnN6vgT12XmopIgvnorQzGoMW-SV2u/view?usp=sharing
https://drive.google.com/file/d/1eNzzopZXx2T2AL9InpeJI3Gx-RhM1JGa/view?usp=sharing

What do you think about the distribution of that?

After I was done I did a stress test without putting the glass panels back on. Here is the result:
https://drive.google.com/file/d/1iqYPGqA6ImFDC7tnEcZqgigKGIrKul4Z/view?usp=sharing

Then I closed all panels and did another one. This is the result:
https://drive.google.com/file/d/13hgl1MnLFSEFJHkMg_i07LqsqxwrXNAg/view?usp=sharing

I also checked the connection and mounting without finding any problems. And I changed my top fans to be exhaust.

The temps are still way too high for my opinion. May you have some other ideas now?

 

[bmockeg]

I just received my new thermal paste (Arctic MX-4) and reinstalled the pump block.

I took some pictures of the already applied thermal paste:
https://drive.google.com/file/d/1eRPnN6vgT12XmopIgvnorQzGoMW-SV2u/view?usp=sharing
https://drive.google.com/file/d/1eNzzopZXx2T2AL9InpeJI3Gx-RhM1JGa/view?usp=sharing

What do you think about the distribution of that?

After I was done I did a stress test without putting the glass panels back on. Here is the result:
https://drive.google.com/file/d/1iqYPGqA6ImFDC7tnEcZqgigKGIrKul4Z/view?usp=sharing

Then I closed all panels and did another one. This is the result:
https://drive.google.com/file/d/13hgl1MnLFSEFJHkMg_i07LqsqxwrXNAg/view?usp=sharing

I also checked the connection and mounting without finding any problems. And I changed my top fans to be exhaust.

The temps are still way too high for my opinion. May you have some other ideas now?

What are your temps after removing the front panel and changing the top fans to exhaust? The 7700k is a pretty hot chip which is why most delid it before overclocking.

 

[xHawaiiToast]

As I said I already moved the top fans so the temps you see are the temps you ask for. Removing the front panel doesn´t change anything. The case is getting plenty of air

 

[Phaaze88]

-trying to OC 7700K
-using Kraken X72, push-pull setup
-temps are high across multiple stress tests
-fan setup looks to be fine
Hmm... we could do with some more info from what you've already posted thus far.

A)I don't think you're using enough voltage for the OC - at least according to Silicon Lottery's binning statistics - which is likely why you get the fatal error.

CPU	All Core SSE Frequency	All Core AVX2 Frequency	BIOS Vcore	% Capable
7700K	4.80GHz	4.60GHz	1.400V	100%
7700K	4.90GHz	4.70GHz	1.412V	Top 96%
7700K	5.00GHz	4.80GHz	1.425V	Top 78%
7700K	5.10GHz	4.90GHz	1.437V	Top 36%
7700K	5.20GHz	5.00GHz	1.450V	Top 12%
7700K	5.30GHz	5.10GHz	1.450V	Top 1%

Even the 'bronze sample' on that list needs 1.40v for 4.8ghz... perhaps the sample you've got is more refined than the ones above?

B)Prime95
Did you run Small FFT, and TURN OFF the AVX options at the bottom of the torture test options? They tend to be on by default.
AVX is too much for just about anything in P95.

C)What LLC(Load Line Calibration) setting did you use? The higher the level, the higher the thermals will be, due to voltage overshoot from counteracting Vdroop.
The low-medium levels are best with air cooled motherboard VRMs. Avoid high and extreme levels unless the VRMs are liquid cooled, or using LN2.

D)When you tightened the cpu block to the cpu, you didn't focus down one corner at a time, did you?

E)Seeing as the Kraken X72 is front mounted, the tubing is likely entering the rad from the top - I doubt it's long enough for you to 180 the rad.
This has me wondering if some air is getting pushed through the pump during testing. Take the rad out and move it around, turn it 180 if possible - see if you can't get those air pockets to get stuck in the rad, which is the best place for it.
They are not topped-off, so there will be some air present - air tends to rise to the highest point in the loop.
With vertically mounted rads in AIOs, tubing entering the rad from the bottom is the better orientation. The pump ends up pushing the air pockets into the rad, where the air then rises and gets trapped there.

F)There's this relation with cpu cooler effectiveness and the IHS, but I'll save that for another time...

That's about all that comes to mind at the moment.

 

[xHawaiiToast]

-trying to OC 7700K
-using Kraken X72, push-pull setup
-temps are high across multiple stress tests
-fan setup looks to be fine
Hmm... we could do with some more info from what you've already posted thus far.

A)I don't think you're using enough voltage for the OC - at least according to Silicon Lottery's binning statistics - which is likely why you get the fatal error.

CPU	All Core SSE Frequency	All Core AVX2 Frequency	BIOS Vcore	% Capable
7700K	4.80GHz	4.60GHz	1.400V	100%
7700K	4.90GHz	4.70GHz	1.412V	Top 96%
7700K	5.00GHz	4.80GHz	1.425V	Top 78%
7700K	5.10GHz	4.90GHz	1.437V	Top 36%
7700K	5.20GHz	5.00GHz	1.450V	Top 12%
7700K	5.30GHz	5.10GHz	1.450V	Top 1%

Even the 'bronze sample' on that list needs 1.40v for 4.8ghz... perhaps the sample you've got is more refined than the ones above?

B)Prime95
Did you run Small FFT, and TURN OFF the AVX options at the bottom of the torture test options? They tend to be on by default.
AVX is too much for just about anything in P95.

C)What LLC(Load Line Calibration) setting did you use? The higher the level, the higher the thermals will be, due to voltage overshoot from counteracting Vdroop.
The low-medium levels are best with air cooled motherboard VRMs. Avoid high and extreme levels unless the VRMs are liquid cooled, or using LN2.

D)When you tightened the cpu block to the cpu, you didn't focus down one corner at a time, did you?

E)Seeing as the Kraken X72 is front mounted, the tubing is likely entering the rad from the top - I doubt it's long enough for you to 180 the rad.
This has me wondering if some air is getting pushed through the pump during testing. Take the rad out and move it around, turn it 180 if possible - see if you can't get those air pockets to get stuck in the rad, which is the best place for it.
They are not topped-off, so there will be some air present - air tends to rise to the highest point in the loop.
With vertically mounted rads in AIOs, tubing entering the rad from the bottom is the better orientation. The pump ends up pushing the air pockets into the rad, where the air then rises and gets trapped there.

F)There's this relation with cpu cooler effectiveness and the IHS, but I'll save that for another time...

That's about all that comes to mind at the moment.

A) I did the same testing with 4,8 Ghz while running Adaptive Mode, which always supplies the CPU with more voltage than needed in most cases. But I don´t get the function of the adaptive mode 100%, because it doesn´t matter if I set the Offset sign to "+" or"-" or if I type in 1,2V or 1,25V as the boost voltage while running at 4,6Ghz for example. Vcore Max (measured with HWMonitor) is always 1,264V.
With that number in mind, I can´t imagine that you need at least 1,4V to run an 7700k with 4,8Ghz. At least I never had a single problem while running at 4,8Ghz even though the temps weren´t good either. It´s just confusing that my system was stable while running aida64 while prime95 didn´t even start up completely.

B) Most of the times I did a Blend test, but I did some Small FFT´s as well. I´m currently using Version 26.6 because I was reading about the new version causing some problems with some type of processors. After I changed from 29.8 to 26.6 my temps were a little better while running the same tests as well. I don´t have any AVX Setting where you mentioned it to be. Maybe I´m blind?

C) I let this setting on auto for a long time but after I saw my high temperatures I set it to level 3. After some time I tried Level 5 and went back to auto again. So my last tests were on an auto setting. I just set it to level 2. Maybe I can see some difference.

D) I´m doing an apprenticeship as an mechatronics engineer so no, I didn´t just tighten them down one by one. I tightened them down really carefully in an X pattern. But did you see the marks on the IHS and the pump block left from the thermal paste in my reply above? It really looks like some places aren´t even covered. I just installed coolers and never had to troubleshoot any other one so I don´t have much references at all.

E) Yeah I was thinking about doing that but my fan setup is really in my way there. I tried to move the hoses around while stress testing but that didn´t do anything. I will test that out tomorrow.

F) I know that you can go on and on with almost every topic in the pc world, BUT there are plenty of people using the same cpu + cooler that don´t have such extreme temps so I can confidently say that it´s 100% not the combination of both.




Other Question: Is it normal for the liquid temperature of an AIO to be at around 29°C (reading off of CAM right now) while only using chrome and go up to only about 35-36°C after 5-10 min of testing? I´ve seen many people write about liquid temps of 40/50°C or so...

 

[Phaaze88]

A)Use manual override - it'll make the whole process easier. Once you've found a stable frequency/Vcore, then slide over to Adaptive/Offset if you want.

B)P95 versions 26.6 and earlier don't have the AVX options embedded.

OQ: 10mins isn't a long enough sample for liquid cooler. It has a far longer TTS(Time To Soak) compared to air coolers. 15-20mins is good enough sample on air. For liquid, you're looking at closer to an hour.

 

[xHawaiiToast]

A)Use manual override - it'll make the whole process easier. Once you've found a stable frequency/Vcore, then slide over to Adaptive/Offset if you want.

B)P95 versions 26.6 and earlier don't have the AVX options embedded.

OQ: 10mins isn't a long enough sample for liquid cooler. It has a far longer TTS(Time To Soak) compared to air coolers. 15-20mins is good enough sample on air. For liquid, you're looking at closer to an hour.

A) I did it exactly like that. I first set the Vcore in Manual Mode, test it out a little and then set it to adaptive mode when I found a stable voltage. I found 1,25V at 4,6Ghz stable after a few tests, but the system gives out 1,28V at max even though I've set the Vcore to 1,25V with a negative offset sign, which should offset the Vcore by a negative number. That's what's strange to me. It's just useless to waste more of my fime finding the perfect Vcore with that method because the adaptive mode just does whatever it wants and not whatever I set. And the automatic increase of the Vcore of about 0,1V at AVX loads when using adaptive mode or offset mode isn't helpful at all. All of the tests worked with the Vcore manually set to 1,22V and everything worked flawlessly so I realy don't understand why my Mb forces so much Voltage on it.

B) What do you mean by not embedded? Is P95 V26.6 not using AVX loads at all or are there just no settings for it?

OQ) I know that 5-10 mins isn't enough. I was running the tests for way longer at the beginning but I realy don't want to stress my Cpu that hard all the time. It hurts me to see my cpu at about or over 90°C all the time and I don't think that it's a good idea to let the 7700k sit at 80/90/100°C over a long periode of time. I will start testing longer after delidding

 

[Phaaze88]

A)That's LLC at work. The different levels will cause varying degrees of under/over shoot.
Whatever Vcore you found stable at manual, expect to need a little more with Adaptive/Offset.

B)Correct. It doesn't use AVX at all.