Question High CPU temp but no throttle

Kamiryu

Reputable
Mar 2, 2019
11
1
4,515
Hello,

I've got a 6700K paired with a 240mm AIO and no OC. I get abnormally high temperatures under load considering the cooler and the fact that the CPU isn't OC.
So, I'm not sure whether the AIO has an issue or whether it is the sensors. It's worth mentioning that it is my third 6700K, the previous 2 died.

When I run stress tests, the CPU doesn't throttle despite the high temperatures (99°C). Using HWinfo and setting the CPU to performance, the minimum frequency is the CPU's max frequency (on all cores).
 
the CPU doesn't throttle
If your CPU is hitting 99°C, it sounds like it is throttling.
Watch the HWiNFO - Performance Limit Reasons data.

0zVUnDo.png


it is my third 6700K, the previous 2 died
Have you ever checked how much voltage is being sent to your CPU? Look for VCore voltage in HWiNFO.

A 240mm AIO should be able to easily handle a 6700K that is not overclocked. What thermal paste are you using? Have you tried replacing the thermal paste recently? Sounds like the AIO is probably dead. This is a very common problem. Do you have an old air cooler that you can test with? They are far more reliable than AIO coolers.
 

Kamiryu

Reputable
Mar 2, 2019
11
1
4,515
If your CPU is hitting 99°C, it sounds like it is throttling.
Watch the HWiNFO - Performance Limit Reasons data.

0zVUnDo.png



Have you ever checked how much voltage is being sent to your CPU? Look for VCore voltage in HWiNFO.

A 240mm AIO should be able to easily handle a 6700K that is not overclocked. What thermal paste are you using? Have you tried replacing the thermal paste recently? Sounds like the AIO is probably dead. This is a very common problem. Do you have an old air cooler that you can test with? They are far more reliable than AIO coolers.




Thank you for your reply !

I've checked the "CPU Performance Limit Reasons data", and it's full of No's. As I mentioned, the frequency hasn't dropped at all, even at 99°C.

Max voltage: 1.376 under load. I'm using the Thermal Grizzly Kryonaut. Even when I had freshly changed the thermal paste, temperatures were still too high. Unfortunately, I don't have any extra cooler to test with...
 
temperatures were still too high
If the AIO is properly secured and the thermal paste was applied correctly then you have to conclude that your AIO is not doing a very good job. How old is it?

I have a slightly bigger Corsair H115i RGB PRO XT 280 mm AIO. It can dissipate 180W with ease and has no problem keeping my 10 core 10850K cool even when it is overclocked 100 MHz. Here is an example with the CPU fully loaded during a Cinebench R23 run.

zJuSAPA.png


If you do not want to spend any money on a new cooler, I would try reducing the CPU voltage. Using a negative offset of -50 mV to -75 mV is usually stable. This can reduce Intel CPU temperatures significantly. At default settings, motherboards always use more voltage than the CPU actually needs to be stable.
 
  • Like
Reactions: KyaraM

Kamiryu

Reputable
Mar 2, 2019
11
1
4,515
If the AIO is properly secured and the thermal paste was applied correctly then you have to conclude that your AIO is not doing a very good job. How old is it?

I have a slightly bigger Corsair H115i RGB PRO XT 280 mm AIO. It can dissipate 180W with ease and has no problem keeping my 10 core 10850K cool even when it is overclocked 100 MHz. Here is an example with the CPU fully loaded during a Cinebench R23 run.

zJuSAPA.png


If you do not want to spend any money on a new cooler, I would try reducing the CPU voltage. Using a negative offset of -50 mV to -75 mV is usually stable. This can reduce Intel CPU temperatures significantly. At default settings, motherboards always use more voltage than the CPU actually needs to be stable.


Well my bad, I have a 280mm as well actually: H110iGT. It's around 6 years old. I think that the temps were never great (way higher than they should). I've applied the thermal paste as per the manufacturer's instructions, so it shouldn't be an issue.
What surprises me is the lack of throttling: could it be that the CPU is actually cooler than the sensors indicate? Assuming the throttling is based on the sensors only, it should still throttle though. Is it a motherboard issue?
 
Intel CPU throttling begins at 100°C. They can run at full speed all day at 99°C. The 4 individual core temperatures should be accurate and similar to each other when the 4 cores are equally loaded running a stress test like Cinebench. Do you see any significant temperature variation from core to core when the CPU is loaded?

Some Intel CPUs have a problem where the heat spreader on top of the CPU cores does not make even and full contact with each of the cores. This can be a problem when you first get a CPU or it can get worse over time. I would reduce the voltage or buy a new cooler before I would buy a delid kit.

https://rockitcool.myshopify.com/products/rockit-88
 
  • Like
Reactions: KyaraM and Kamiryu

Kamiryu

Reputable
Mar 2, 2019
11
1
4,515
Intel CPU throttling begins at 100°C. They can run at full speed all day at 99°C. The 4 individual core temperatures should be accurate and similar to each other when the 4 cores are equally loaded running a stress test like Cinebench. Do you see any significant temperature variation from core to core when the CPU is loaded?

Some Intel CPUs have a problem where the heat spreader on top of the CPU cores does not make even and full contact with each of the cores. This can be a problem when you first get a CPU or it can get worse over time. I would reduce the voltage or buy a new cooler before I would buy a delid kit.

https://rockitcool.myshopify.com/products/rockit-88

Oh, I thought that 99°C would be enough to throttle. The cores have similar temps, so it seems to be a cooler issue only. I plan to upgrade with a new build using the upcoming CPU's, but I wanted to resolve the issue in case I'm selling my current system as a whole.
Thanks for your help!
 
  • Like
Reactions: uWebb429

Karadjgne

Titan
Ambassador
Well my bad, I have a 280mm as well actually: H110iGT. It's around 6 years old.
Most likely not the pump at 6 yrs old, most likely is coolant levels insufficient. The pump will still shove the coolant around, what's left of it, but there's too much air (for lack of a better word) in the rad piping, so there's actually very little transfer of heat. This is easily checked with a vertical rad, the exhaust will be hot at the bottom, but noticibly cooler behind the top fan.

If the coolant has deteriorated enough to start allowing algae buildup, one tube will be noticibly warmer than the other as the trapped inlet side doesn't get much chance at absorbing the heat from the cold plate, and lack of flow prevents the coolant at the cold plate from leaving, so cooks like a pan on a burner.

6 years old is ± the time that coolant starts breaking down into its constituent molecular makeup. Oxygen molecules are small, so pass through the tubing, nitrogen is much larger, so gets trapped. Same applies to several other gases. Sooner or later there's simply not enough coolant left to do the job.

Time to replace the cooler. Either bigish air or 240mm/280mm aio.
 
Aug 17, 2022
8
1
15
Download "Throttlestop" go through the walkthrough to Uneervolt and make sure you select SpeedShift and unselect SpeedStep : https://www.ultrabookreview.com/31385-the-throttlestop-guide/

Older motherboards don't have BIOS setup to select Speedshift or Speedstep, so you have to rely on XTU or Throttlestop (XTU doesn't support 6700K)

Speedshift was introduced by Intel in Skylakes, but older motherboards did not get the memo...so they are default on Speedstep.

I had stock 6700K @ 33C running idle at 2-5% load and a massive jump to 70C at 25% load and unbearable 93C at 100% load. After the Speedshift toggle, I was able to OC @ 44 ratio with a 20mV overvolt, @28-30C idle at 2-10%, and 81C at max load. A beauty-All by using a small chassis with aircooler at 21C ambient.
 
  • Like
Reactions: uWebb429

Karadjgne

Titan
Ambassador
I've never seen a pc go to idle with SpeedStep or SpeedShift disabled. And yes, older boards did have it, even legacy bios had it, it just wasn't a visible/user accessible toggle on boards not designed for OC.

SpeedStep or its successors entire purpose was to cut power/frequency from the cpu after a determined time period of user inactivity. Disabling that negated the power/frequency drop which along with C-states/C1E/EIST was necessary for a Static OC as power fluctuations created instability.

2 kinds of OC were Static and Turbo, Static held the same frequency at all times, regardless of load, Turbo kept the same default behavior but raised the maximum default turbo levels.

Static could be tuned for lower temps as single thread vcore/vid didn't change when switching to multi core. Turbo required a higher vid, higher vcore and subsequently ran slightly higher temps because LLC had to be more inline with vdroop seen during the switch from idle to turbo power use.

SpeedStep and SpeedShift are the same thing. The term 'Intel SpeedStep' was stopped in 2012 when the trademark was invalidated, Intel changed it to Speed Shift instead.