Cooling problem NOCTUA NH-D15 + 13700k

[Alex Scr]

Hello,

 

Recenlty i upgraded my MOBO + RAM+ CPU and kept the old noctua cooler NH-D15, thinking it will be good enough.

After installation i noticed that, the CPU is thermal throttling when i'm doing the stress test in cpu-z. The odd part is that, the temperatures don't change, even when i'm keeping the side panel open.

When i tried to check if the cooler is working (i touched the metal part) and i noticed that the cooler is not even warm, so my guess is it doesn't make right contact?

Also, after first time when i mounted the new parts, beside de cpu cooler i had only the stock fans from the case (1 intake- 1 pulling out) and it took 2 minutes for the cpu-z stress test to get the CPU at 100 degrees C. Sawing this, i bought 2x NF-A14 PWM for the intake but the difference is really 2-3 degrees, and it takes la 5 minutes to thermal throttling.

 

My specs:

CPU: I7 13700k

MOBO: z790- Gaming X -Gigabyte

RAM: G.Skill Trident Z5 K2 - 6000MHZ, CL30

GPU: ASUS TUF 3080Ti

PSU: ROG THOR 850W Platinum+

SSD: 970EVO Plus 1TB

CASE: 4000D- Airflow

Fans: 2x NF-A14 PWM as intake, 2x Stock Corsair fans (the ones that came with the case) as pulling out.

Cooler: Noctua NH-D15, 2x140mm Fans

 

Currently i'm waiting to get the thermalright contact frame for lga1700, hopefully it will fix this problem.

 

Do you guys have any suggestions? Did anyone had the same issues like me?

 

[will0hlep]

That behaviour sounds kinda normal for a stress test. What is the cpu temps when idel?

[knch]

https://ncc.noctua.at/cpus/model/Core-i7-13700K-1640

https://noctua.at/en/intel-lga1700-mounting-kit-order-form

 

 

When did you buy that NH-D15?

[duckargent]

I have the same D15 cooler on a 10900k, the temps are perfectly fine on stock / no OC. I just replaced the front Noctua A15 fan with an A12 because my G.Skill TridentZ RAM was too tall for the cooler. The contact frame may help, of course, but I don't think this is the problem: Intel is most probably testing these CPUs without any frame.

 

Take out the cooler and replace the paste; don't put just a droplet in the middle and press, instead use a small improvised shovel to cover all the area, including the corners.

[Alex Scr]

1 hour ago, will0hlep said:

That behaviour sounds kinda normal for a stress test. What is the cpu temps when idel?

@38-45 celcius for idle, in gaming after 25 mins of walking around in Saint Denis in RDR2 the CPU had 87 maximum in HWMONITOR

[Alex Scr]

1 hour ago, knch said:

https://ncc.noctua.at/cpus/model/Core-i7-13700K-1640

https://noctua.at/en/intel-lga1700-mounting-kit-order-form

 

 

When did you buy that NH-D15?

I have it from 2017-2018, i do have the mounting kit for LGA1700, i bought it from amazon 

[Alex Scr]

24 minutes ago, duckargent said:

I have the same D15 cooler on a 10900k, the temps are perfectly fine on stock / no OC. I just replaced the front Noctua A15 fan with an A12 because my G.Skill TridentZ RAM was too tall for the cooler. The contact frame may help, of course, but I don't think this is the problem: Intel is most probably testing these CPUs without any frame.

 

Take out the cooler and replace the paste; don't put just a droplet in the middle and press, instead use a small improvised shovel to cover all the area, including the corners.

The paste is good, covering all the CPU, but i will try one more time.

I did read from a post on reddit that this ICS for LGA1700 is bending the CPU-s on the middle (on the hotspot) because of the mounting pressure and I'm assuming this is my case too because it doesn't make any sense, why my cooler is cold while my CPU is very hot...

[will0hlep]

Then my opinion would be that your PC is behaving as normal.

 

PCs do get very hot during stress tests (the point is to push the machine too it's limits) and CPUs are designed to thermal throttle when they get too hot to protect themselves. (13700K is designed for 100C at T-Junction)

 

Also CPU coolers don't typically feel that hot at the points furthest from the CPU cause they are constantly losing heat to the air as fast as they can (and the further you are from the CPU the more the heat has spread out or been transfered to the air).

(I'm not an expert and I might be wrong but I think your fine)

[Alex Scr]

3 minutes ago, will0hlep said:

Then my opinion would be that your PC is behaving as normal.

 

PCs do get very hot during stress tests (the point is to push the machine too it's limits) and CPUs are designed to thermal throttle when they get too hot to protect themselves. (13700K is designed for 100C at T-Junction)

 

Also CPU coolers don't typically feel that hot at the points furthest from the CPU cause they are constantly losing heat to the air as fast as they can (and the further you are from the CPU the more the heat has spread out or been transfered to the air).

(I'm not an expert and I might be wrong but I think your fine)

Thank you for your opinion.

And about the cooler, i did touch it close to the CPU and still didn't feel heat regarding the CPU being at @100 degrees.

Also, the CPU's temperature is fluctuating on different task (like gaming can be from 65-87, but it's not constant, it's like jumping around), is this also ok? 

[will0hlep]

4 minutes ago, Alex Scr said:

Thank you for your opinion.

And about the cooler, i did touch it close to the CPU and still didn't feel heat regarding the CPU being at @100 degrees.

Also, the CPU's temperature is fluctuating on different task (like gaming can be from 65-87, but it's not constant, it's like jumping around), is this also ok? 

CPU temp does fluctuate alot, I wouldn't be concerned.

[duckargent]

On 3/23/2023 at 2:53 PM, will0hlep said:

Then my opinion would be that your PC is behaving as normal.

No, it's not behaving as normal, that CPU should NOT thermal throttle with a Noctua D15 cooler. My 10900k which is also a power hungry CPU is not touching 90C when I overclocked it at 5Ghz all cores with manual voltages on Prime95 SmallFFTs (so maximum CPU stress); on overclock with auto voltages is was reaching 90-91C (again on Prime95 SFFTs). I don't have all the details for I have been using my CPU on stock since I installed it. Just look at that double-tower D15 (with 2x14cm fans), it's almost half the case area, it is perfectly able to soak the heat. I tested even the Noctua U14s (single tower), it didn't thermal throttle on Prime95 SmallFFTs (stock, no OC).

 

So, a desktop CPU should not thermal throttle in any circumstances, this is not a laptop: the OP has a high quality cooler and a good case airflow. The problem can be with the thermal paste (either reapply it or buy a new one), a defective cooler or the IHS is so bent/convex that he needs a contact frame. In the worst case he may need a CPU delidding.

[Alex Scr]

On 3/28/2023 at 10:50 AM, duckargent said:

No, it's not behaving as normal, that CPU should NOT thermal throttle with a Noctua D15 cooler. My 10900k which is also a power hungry CPU is not touching 90C when I overclocked it at 5Ghz all cores with manual voltages on Prime95 SmallFFTs (so maximum CPU stress); on overclock with auto voltages is was reaching 90-91C (again on Prime95 SFFTs). I don't have all the details for I have been using my CPU on stock since I installed it. Just look at that double-tower D15 (with 2x14cm fans), it's almost half the case area, it is perfectly able to soak the heat. I tested even the Noctua U14s (single tower), it didn't thermal throttle on Prime95 SmallFFTs (stock, no OC).

 

So, a desktop CPU should not thermal throttle in any circumstances, this is not a laptop: the OP has a high quality cooler and a good case airflow. The problem can be with the thermal paste (either reapply it or buy a new one), a defective cooler or the IHS is so bent/convex that he needs a contact frame. In the worst case he may need a CPU delidding.

Hello, thank you for your answer.

 

I did apply a contact frame. What i noticed is this : after i mounted the contact frame and tested the cpu in cinerbench r23 it started to draw more power (240w). Before, with the motherboard's mounting frame it was drawing maximum 230W. However, it did not help with temps much, is still going to 100 but it seems like is drawing more power.

As thermal paste im using noctuas nh-1, do you have a better recommendation that i can try?

And about the cooler, what i can try to see if is defective? The fans are going all the way up to 1500 RPM and don't have visual damage on the sink.

Also, i noticed that the front 2x NF-A14 PWM are doing only @1000 RPM even on stress test, which is weird, i set them to go up to 100% once the CPU reaches @70C

[duckargent]

On 3/29/2023 at 1:49 PM, Alex Scr said:

What i noticed is this : after i mounted the contact frame and tested the cpu in cinerbench r23 it started to draw more power (240w). Before, with the motherboard's mounting frame it was drawing maximum 230W. However, it did not help with temps much, is still going to 100 but it seems like is drawing more power.

As thermal paste im using noctuas nh-1, do you have a better recommendation that i can try?

And about the cooler, what i can try to see if is defective? The fans are going all the way up to 1500 RPM and don't have visual damage on the sink.

Also, i noticed that the front 2x NF-A14 PWM are doing only @1000 RPM even on stress test, which is weird, i set them to go up to 100% once the CPU reaches @70C

I think you improved the temps by a bit and thus the CPU is reaching new heights, hence it draws even more power, hence the temperatures rise again due to the additional consumption. Basically now the CPU is thermal throttling less; it's maximum turbo TDP is 253W, so you are getting close to that. I cannot say whether the throttling stopped since it is reaching the 100C turning point, so I am just guessing. In my opinion it is still throttling since it is not reaching 250W yet.

 

I think the thermal paste is fine, I too used the Noctua paste which came with my cooler. Regarding the cooler, you cannot test it unless you change it. But given that it is Noctua, I don't think this is the culprit, as I have not ever seen a complaint regarding Noctua products (I also have this cooler and 4 x A14 case fans).

 

It's either that you need to delid the CPU or the D15 cannot handle this TDP. Regarding the latter, my CPU also has a 250W TDP and I am using the same cooler and my temps are fine. I think you should search for a service in your area and ask for a delid offer, and explain them the issue. If you delid the CPU you will decrease the temps by as much as 10C, which is the difference between keeping your D15 or buying an AIO. Deliding is a bit risky, but it's either this or buying and expensive AIO (or buying, maybe for nothing, a new D15 just to rule out the possibility of a defective cooler).

 

How to Delid Your Processor and Why it's Worth It | Tom's Hardware (tomshardware.com)

 

Regarding the A14 fans, be sure not to connect them through the low noise adapter (LNA), which limits the speed to 1200 rpm.

NF-A14 PWM || Specifications (noctua.at)

In my opinion, you don't need an aggressive curve for the intake fans, these are big 140mm fans which push a lot of air (max 140 cm/h at max speed), even at low rpm. I am keeping my 2x A14 intake and 2x A14 exhaust fans on the silent profile (700 rpm), it's enough and the temps are very good, and they are silent. If they were 120mm or smaller, then yes, you would need to increase their speed.

 

[Alex Scr]

19 hours ago, duckargent said:

I think you improved the temps by a bit and thus the CPU is reaching new heights, hence it draws even more power, hence the temperatures rise again due to the additional consumption. Basically now the CPU is thermal throttling less; it's maximum turbo TDP is 253W, so you are getting close to that. I cannot say whether the throttling stopped since it is reaching the 100C turning point, so I am just guessing. In my opinion it is still throttling since it is not reaching 250W yet.

 

I think the thermal paste is fine, I too used the Noctua paste which came with my cooler. Regarding the cooler, you cannot test it unless you change it. But given that it is Noctua, I don't think this is the culprit, as I have not ever seen a complaint regarding Noctua products (I also have this cooler and 4 x A14 case fans).

 

It's either that you need to delid the CPU or the D15 cannot handle this TDP. Regarding the latter, my CPU also has a 250W TDP and I am using the same cooler and my temps are fine. I think you should search for a service in your area and ask for a delid offer, and explain them the issue. If you delid the CPU you will decrease the temps by as much as 10C, which is the difference between keeping your D15 or buying an AIO. Deliding is a bit risky, but it's either this or buying and expensive AIO (or buying, maybe for nothing, a new D15 just to rule out the possibility of a defective cooler).

 

How to Delid Your Processor and Why it's Worth It | Tom's Hardware (tomshardware.com)

 

Regarding the A14 fans, be sure not to connect them through the low noise adapter (LNA), which limits the speed to 1200 rpm.

NF-A14 PWM || Specifications (noctua.at)

In my opinion, you don't need an aggressive curve for the intake fans, these are big 140mm fans which push a lot of air (max 140 cm/h at max speed), even at low rpm. I am keeping my 2x A14 intake and 2x A14 exhaust fans on the silent profile (700 rpm), it's enough and the temps are very good, and they are silent. If they were 120mm or smaller, then yes, you would need to increase their speed.

 

I just finished another session of demounting-mounting + new paste for the cooler. I'm starting to lose my patience as i see this cooler work on other's systems. Another thing, when i take off the cooler, the thermal paste patern seems to go on the sides, leaving the middle almost empty. This is something i dont understand at all. I don't know what is the issue, even with the contact frame (which was supposed to fix this) and i'm verry sure this is the problem causing higher temps (poor contact on the middle where was supposed to be the most) and don't know how to fix it.

 

This are my temps in core temp after a 10 min cinebench r23 test. Also i noticed that, HWMonitor is not showing real stats for some reason.

Anyway, in this test i get @30293 points with fans at 100% and i see the CPU drawing up to 241 W.

 

As for gaming, the most i've seen was a spike to 87 degrees in fifa 23, in rest 50-75C depends on game or scene.

I'm starting to get verry desperate because the CPU, MOBO and rams are bought for 2 weeks and i don't see other option than buying an X73 kraken or h150i (wich i have doubts too that will fix the problem because of the poor contact still being there).

 

 

PS: I'm trying to avoid the delid. Also i was thinking about my heat pipes being not efficient anymore since i have this cooler for a long time.

 

[duckargent]

2 hours ago, Alex Scr said:

Another thing, when i take off the cooler, the thermal paste patern seems to go on the sides, leaving the middle almost empty. I don't know what is the issue, even with the contact frame (which was supposed to fix this) and i'm verry sure this is the problem causing higher temps (poor contact on the middle where was supposed to be the most) and don't know how to fix it.

 

This are my temps in core temp after a 10 min cinebench r23 test. Anyway, in this test i get @30293 points with fans at 100% and i see the CPU drawing up to 241 W.

 

As for gaming, the most i've seen was a spike to 87 degrees in fifa 23, in rest 50-75C depends on game or scene.

I ran the Cinebench R23 benchmark myself, it seems that I exceed the 90C threshold, but the maximum which I record is 95C. I took this picture at the 9:50 mark (after I finished a previous 10min loop). Also, I record a maximum of 244W, so the consumption is the same. What is not the same is the score: after 10 min I get 15872, which is nowhere near your double score. I'll give this to Intel: they created some powerful CPUs with these latest models. My CPU didn't throttle, since I maxed out at 95C on the 4th and 6th core core.

 

 

 

This is the difference in temperatures since we are using the same cooler: the bent/convex IHS on your CPU. This is why there is no paste in the middle if you take off the cooler, because the pressure is not even on the surface: when you screw in the cooler, the pressure is the biggest on the center where the bump resides (thus spilling the paste on the sides), and it gets weaker on the sides, where you see more paste. I reckon that the frame is not very efficient at mitigating this flaw. If something, don't screw in too hard the cooler in order not to apply too much pressure on the center.

 

You know what? Don't do anything. If you are willing to give another try, apply again the paste and let the screws a little bit looser (don't screw in too hard). Since you get this max temp only while running the stress tests, and since your CPU throttles only in this scenario, you will be fine in any other scenario which doesn't fully utilize all the cores (100%). For gaming, if this is what you seek, you will be perfectly fine. I reckon that the maximum of 87C in gaming is on the core which gets to 100C in Cinebench, and 87 might be the maximum, which is perfectly fine in the long run (even if it maxes out at 90). And if it doesn't throttle at 100C, you won't incur any performance loss.

 

Trust me, I have been using this CPU since 2020 only on gaming and I've never had any problem with it: either performance loss/drops or too high temperatures. Enjoy the CPU and forget about the temps.