[SOLVED] i7 8700k running too hot with proper cooling during stress test?

thewalleprod

Distinguished
Feb 28, 2016
165
0
18,690
I'm running a stock 8700k with Corsair H100i AIO cooler on a Prime Z370-a mobo. I reapplied Thermal Grizzly Kryonaut thermal paste on the CPu and I have three 120mm fans on the bottom for intake, one 120mm exhaust fan in the back, and three 120mm exhaust fans on top. the 240mm AIO is mounted in the front as an intake. Case is a Lian Li Dynamic XL.

Ran Aida64 with HWinfo64 running sensors for temp testing. I checked off "Stress CPU", "Stress FPU", and "Stress Cache". My max temp for CPU package was 90c after 10 minutes of stress testing in Aida64. I'm not sure why this is. I have MCE disabled in BIOS, XMP is enabled for 3200mhz RAM. Fans are set to standard curve. I feel like temps are pretty high for this.

Any suggestions what I should do?
 
Solution
I have three 120mm fans on the bottom for intake, one 120mm exhaust fan in the back, and three 120mm exhaust fans on top. the 240mm AIO is mounted in the front as an intake. Case is a Lian Li Dynamic XL.
You should probably put radiator in top position (as exhaust).
I don't think currently it has very much access to fresh air. And it's not front intake. More like obstructed side intake.

Can you show any photos of your system?
(upload to imgur.com and post link)

It's this case? Correct?
81fbjKwKgsL._AC_SL1500_.jpg
I'm running a stock 8700k with Corsair H100i AIO cooler on a Prime Z370-a mobo. I reapplied Thermal Grizzly Kryonaut thermal paste on the CPu and I have three 120mm fans on the bottom for intake, one 120mm exhaust fan in the back, and three 120mm exhaust fans on top. the 240mm AIO is mounted in the front as an intake. Case is a Lian Li Dynamic XL.

Ran Aida64 with HWinfo64 running sensors for temp testing. I checked off "Stress CPU", "Stress FPU", and "Stress Cache". My max temp for CPU package was 90c after 10 minutes of stress testing in Aida64. I'm not sure why this is. I have MCE disabled in BIOS, XMP is enabled for 3200mhz RAM. Fans are set to standard curve. I feel like temps are pretty high for this.

Any suggestions what I should do?
Manually set all of your fans (including the AIO fans and pump) to 100 percent Maximum and rerun the tests. If the temps are lower then you would have to reevaluate your standard curves and adjust them to increase airflow. Or if still too high you could consider a 360mm radiator.
 

thewalleprod

Distinguished
Feb 28, 2016
165
0
18,690
Manually set all of your fans (including the AIO fans and pump) to 100 percent Maximum and rerun the tests. If the temps are lower then you would have to reevaluate your standard curves and adjust them to increase airflow. Or if still too high you could consider a 360mm radiator.

AIO is connected to CPU_FAN. I only have 1 fan that isn't PWM and it's connected to the hub. Is that fan automatically at full speed if being powered by SATA?
 
AIO is connected to CPU_FAN. I only have 1 fan that isn't PWM and it's connected to the hub. Is that fan automatically at full speed if being powered by SATA?
A sata to fan adapter hub should provide the full 12 volts for the fan to run at maximum. You can check your hub specifications to verify. The main thing is that the AIO pump and AIO fans are both running at 100 percent maximum during testing.
 

thewalleprod

Distinguished
Feb 28, 2016
165
0
18,690
A sata to fan adapter hub should provide the full 12 volts for the fan to run at maximum. You can check your hub specifications to verify. The main thing is that the AIO pump and AIO fans are both running at 100 percent maximum during testing.

Just ran all fans that are connected to the hub at full speed, and ran the AIO fans/pump at Extreme from iCue software, ran Aida and still got 88-90c max CPU package temps. I noticed that my max VID volts hits 1.394 which seems way too high for a non-oc 8700k..
 
You might consider installing Intel's XTU, and applying a negative offset?

(Also, per HWmonitor, what clock speeds are being sustained on all-core tests? If all cores are 4.7 GHz, then it sounds as if MCE is enabled in BIOS, which will contribute substantially to heat, as 'stock' clock speeds for all core operation would be only 4.4 GHz or so, at best...)
 

thewalleprod

Distinguished
Feb 28, 2016
165
0
18,690
You might consider installing Intel's XTU, and applying a negative offset?

(Also, per HWmonitor, what clock speeds are being sustained on all-core tests? If all cores are 4.7 GHz, then it sounds as if MCE is enabled in BIOS, which will contribute substantially to heat, as 'stock' clock speeds for all core operation would be only 4.4 GHz or so, at best...)

MCE is disabled in Bios. I have just manually set an underclock for 4.3ghz on all cores, and set manual voltage for 1.15v. Ran Prime95 small FFT test and max package temperature is not reaching above 76C after 5-10minutes of running it. Not sure why max vcore is going up to 1.29-1.3v if I set it to manually 1.15v still.
 
I have three 120mm fans on the bottom for intake, one 120mm exhaust fan in the back, and three 120mm exhaust fans on top. the 240mm AIO is mounted in the front as an intake. Case is a Lian Li Dynamic XL.
You should probably put radiator in top position (as exhaust).
I don't think currently it has very much access to fresh air. And it's not front intake. More like obstructed side intake.

Can you show any photos of your system?
(upload to imgur.com and post link)

It's this case? Correct?
81fbjKwKgsL._AC_SL1500_.jpg
 
Solution

thewalleprod

Distinguished
Feb 28, 2016
165
0
18,690
You should probably put radiator in top position (as exhaust).
I don't think currently it has very much access to fresh air. And it's not front intake. More like obstructed side intake.

Can you show any photos of your system?
(upload to imgur.com and post link)

It's this case? Correct?
81fbjKwKgsL._AC_SL1500_.jpg
Yes exactly that case. Radiator is on the side there.. I thought that putting radiator on top is bad for an AIO pump.
 
Just ran all fans that are connected to the hub at full speed, and ran the AIO fans/pump at Extreme from iCue software, ran Aida and still got 88-90c max CPU package temps. I noticed that my max VID volts hits 1.394 which seems way too high for a non-oc 8700k..
Is there any particular reason why you can't use a 360mm radiator? I guess its too late to exchange the 240mm one.
 

thewalleprod

Distinguished
Feb 28, 2016
165
0
18,690
Is there any particular reason why you can't use a 360mm radiator? I guess its too late to exchange the 240mm one.
Yeah exactly. Maybe in the future. What is weird is that Prime95 with AVX disabled i'm hitting max temps of 77C during testing, and in Aida64 i'm hitting 87-90C. I thought that Prime95 was way more demanding.
 

thewalleprod

Distinguished
Feb 28, 2016
165
0
18,690
I may have figured it out. I just OCed the 8700k to 4.7ghz on all cores, with a manual voltage of 1.225 and LLC at level 5. Temps are not going above 80C while stress testing on Prime95 and Aida64. Core clocks are running at 3600-4000mhz when running the stress tests though. Does this mean this is an unstable OC?
 
Totally up to you but you should think about deliding the cpu, I used the Rockit Cool with their copper IHS liquid metal and Thermal Grizzly Kryonaut. I did mine when i first bought it and havent touched it since ( feb 2018). Running 5.1Ghz all core @ 1.4v 0 AVX offset with a Cryorig R1 ultimate on Prime95 small test im topping out at 78*c.


Depending on your AVX offset if you're set at 4.7ghz all core you should only be dropping down to 4.4ghz, which would be a AVX offset of 3 and you should be able to set it in the bios. I would probably guess you dont have the bios setup right and when it goes all core its only hitting the 4.3Ghz that intel has set for it, with an AVX offset would drop it down to 4.0Ghz. If you dont have the all core set up correctly then it could also be thermal throttling as well, that why you could be seeing the dip to 3.6ghz.

I watched this video when i overclocked my 8700k. LINK
 

thewalleprod

Distinguished
Feb 28, 2016
165
0
18,690
Totally up to you but you should think about deliding the cpu, I used the Rockit Cool with their copper IHS liquid metal and Thermal Grizzly Kryonaut. I did mine when i first bought it and havent touched it since ( feb 2018). Running 5.1Ghz all core @ 1.4v 0 AVX offset with a Cryorig R1 ultimate on Prime95 small test im topping out at 78*c.


Depending on your AVX offset if you're set at 4.7ghz all core you should only be dropping down to 4.4ghz, which would be a AVX offset of 3 and you should be able to set it in the bios. I would probably guess you dont have the bios setup right and when it goes all core its only hitting the 4.3Ghz that intel has set for it, with an AVX offset would drop it down to 4.0Ghz. If you dont have the all core set up correctly then it could also be thermal throttling as well, that why you could be seeing the dip to 3.6ghz.

I watched this video when i overclocked my 8700k. LINK

Sync all cores is set to 47. AVX offset is 0. How would it be thermal throttling if max temps are below 75C? Maybe I set voltage too low?
 
If cores are set to 47 and AVX is set to 0 then it should stay at 4.7Ghz, it should not dip down to 3.6-4.0Ghz during the test. If it can boot at 4.7ghz at 1.225v then during the test if there is an issue it would crash or freeze not drop the core speed down to match voltage needs.
 

thewalleprod

Distinguished
Feb 28, 2016
165
0
18,690
That is correct.

@thewalleprod
If the turbo power limits are set to default values, that could cause the CPU speed to decrease during a full load stress test. Rather than guessing what the problem might be, run HWiNFO and monitor the Performance Limit Reasons data. This will tell you if it is a power limit throttling problem or something else.

DTPcQYV.png

Fixed it in the BIOS, now with I put max limit with long duration/short duration package power and CPU cache ratio to 42. Running Prime95 with AVX disabled, max CPU package temp is 78-79C after 10 minutes no crashes. I had to adjust voltage to 1.3v with LLC set to Level 2. AVX offset is 3. VCore max is showing as 1.186v though. Core clock speeds are at 4700mhz constant. Sometimes dipping to 4400-4500.
 
AVX offset is 3
CPU-Z will show your CPU flip flopping back and forth between 4400 MHz and 4700 MHz when using an AVX offset. If you are happy with that then everything is good. If you rarely ever run AVX heavy software then consider setting AVX offset back to 0.

ThrottleStop gives you access to this setting so you can make changes to the offset value on the fly without having to reboot.

bFCQZ32.png


HWiNFO reports AVX throttling as,

IA: Max Turbo Limit