Question 7950X / NH-D15 (after 14 hours of render at 95C cooler is still cold to the touch)

[Aleximus96]

Good day, fellow PC enthusiasts.
Namely, whenever I had a problem in the past that was not Google-able, I came here, received multiple answers really fast, and had my problems resolved.

I call upon your might once again!

A friend asked me to build him a rendering PC for 3Ds Max and stuff like that, mainly to render the scenes for commercials as that is his job.

So I've put the list and he bought the parts and I assembled it and it was all fine until the temps hit 95C... I soon realized that's the "new normal" for the 7000 series.

The problem is as in the title, after many hours of full CPU load, the cooler is not even warm to the touch. Things I've tried so far:

Reseated the cooler and reapplied the paste (the paste in question is NT-H1 that came with the cooler).
Checked that there was nothing on the bottom of the cooler (plastic film or anything of that sort)
Clocks are not dropping below 5.0 Ghz, so the cooler must be doing some work.

I don't really know where to go from this point
any help is greatly appreciated!

Best wishes,
your fellow PC enthusiast.

 

[PEnns]

I am confused. Are you saying the not too hot cooler is really a problem?

It might be strange / unusual but if all is working fine, it's a blessing!! Maybe you got the most efficient cooler on the market that doesn't even break a sweat...

 

[Aleximus96]

I am confused. Are you saying the not too hot cooler is really a problem?

It might be strange / unusual but if all is working fine, it's a blessing!! Maybe you got the most efficient cooler on the market that doesn't even break a sweat...

I would be surprised that that's the case, like the bottom plate of the tower that sits on top of the CPU is cold to the touch after running at 95C for 14+ hours.

That doesn't sound right, it should be at least a little warm, no?

 

[Unolocogringo]

Not if you have good airflow over the towers.
The heatpipes carry the heat to the finstacks.
From there it radiates out the heatpipes into the fin stack to be blown away by the fans.
The fins should stay cool.
Put your finger on the heatpipe just above the CPU but below the fins and it should be quite warm. but not extreme HOT.
If the fin stack is hot your fan speed needs to be increased.

 

[Aleximus96]

Not if you have good airflow over the towers.
The heatpipes carry the heat to the finstacks.
From there it radiates out the heatpipes into the fin stack to be blown away by the fans.
The fins should stay cool.
Put your finger on the heatpipe just above the CPU but below the fins and it should be quite warm. but not extreme HOT.
If the fin stack is hot your fan speed needs to be increased.

Airflow is pretty solid, mesh front panel, with 2 intake 120mm, (middle one 120mm had to be removed to fit GPU inside).
I know coolers shouldn't be hot hot, of course, but it's just cold, not cold like the case, but not warm enough to feel it's radiating heat from a 95c CPU, will definitely check with a friend to touch the pipes near the socket to check that as well, thanks for suggestion!

 

[drea.drechsler]

Airflow is pretty solid, mesh front panel, with 2 intake 120mm, (middle one 120mm had to be removed to fit GPU inside).
I know coolers shouldn't be hot hot, of course, but it's just cold, not cold like the case, but not warm enough to feel it's radiating heat from a 95c CPU, will definitely check with a friend to touch the pipes near the socket to check that as well, thanks for suggestion!

I think this is pretty common to see with modern processors and Ryzen 5th gen in particular. The problem is that 7nm (and smaller) geometry offers insufficient surface area to adequately transfer heat out of the active devices of the CPU and into the heat spreader. Further, the processor has dozens of temperature sensors that are very close to the active components. The temperature reported (Tdie) is the hottest one of those at any given time. Even with very good cooling, it's very hard to keep temp down inside the die.

Also, never look at frequency of a Ryzen processor to infer performance; that's because the boost algorithm will find a way to boost to max clock whenever it can but also rapidly pull clock back as needed based on temperature to maintain safe operation. Average clock across time may be more helpful but better still is to look at performance with benchmarks. Something like Cinebench running for 10 minutes (take the last run for the results) should work well for an air cooler since they thermally saturate almost immediately if being asked to dissipate more power than they can handle.

 

[Unolocogringo]

Igor's lab has Flir images that explain it better than words.
Referred to as heat soak. If the outer part of the fins are hot, the fan is running too slow or your processor is putting out more heat than the cooler can handle.

 

[Aleximus96]

I think this is pretty common to see with modern processors and Ryzen 5th gen in particular. The problem is that 7nm (and smaller) geometry offers insufficient surface area to adequately transfer heat out of the active devices of the CPU and into the heat spreader. Further, the processor has dozens of temperature sensors that are very close to the active components. The temperature reported (Tdie) is the hottest one of those at any given time. Even with very good cooling, it's very hard to keep temp down inside the die.

Also, never look at frequency of a Ryzen processor to infer performance; that's because the boost algorithm will find a way to boost to max clock whenever it can but also rapidly pull clock back as needed based on temperature to maintain safe operation. Average clock across time may be more helpful but better still is to look at performance with benchmarks. Something like Cinebench running for 10 minutes (take the last run for the results) should work well for an air cooler since they thermally saturate almost immediately if being asked to dissipate more power than they can handle.

Great piece of info, would have never thought about that (smaller process> more denser heat)

I meant 5ghz all core boost has been going for 14 hours, so that's taken care of, it was similar when I ran the new cinebench for 10 minutes after assembly. Will get back to you guys when a friend touches the pipes near the socket after he start another render.

 

[CelicaGT]

Great piece of info, would have never thought about that (smaller process> more denser heat)

I meant 5ghz all core boost has been going for 14 hours, so that's taken care of, it was similar when I ran the new cinebench for 10 minutes after assembly. Will get back to you guys when a friend touches the pipes near the socket after he start another render.

I work in industrial cooling. Without getting into a pages long explanation..... everything you see is normal. Most of the initial heat exchange and movement is done by the medium changing phase from a liquid to a gas, storing the energy as latent heat energy (You won't feel this, as opposed to sensible heat energy). In opposition, as cooled, the latent heat is conducted into the ambient air by the pipe walls and fins, that latent heat is gone and the gas returns to a liquid and wicks back to the cooling plate area. If the fins were warm (as noted by others) you could indeed have an issue. As clock speed and temps are "as expected" don't go chasing zebras, just run the system.

 

[Unolocogringo]

I was typing a similar post when yours popped up!
Well not the technical part

I work in industrial cooling. Without getting into a pages long explanation..... everything you see is normal. Most of the initial heat exchange and movement is done by the medium changing phase from a liquid to a gas, storing the energy as latent heat energy (You won't feel this, as opposed to sensible heat energy). In opposition, as cooled, the latent heat is conducted into the ambient air by the pipe walls and fins, that latent heat is gone and the gas returns to a liquid and wicks back to the cooling plate area. If the fins were warm (as noted by others) you could indeed have an issue. As clock speed and temps are "as expected" don't go chasing zebras, just run the system.

 

[Aleximus96]

Thank you all very much for your input regarding this question, I will monitor the situation for a few more render runs just to be sure.

Once again TH forum proves to be THE place when in doubt, thanks a lot guys, again!

Best wishes and regards!

 

[thestryker]

In the case of AMD Zen 4 desktop CPUs this is basically caused by the IHS being too thick and not transferring heat as well as it could. A few tech youtubers have done videos showing the IHS temperature to be a fair bit lower than the CPU temperature.

 

[gamerk316]

A good test of the cooler would be to watch how fast temperatures go down once load is removed from the system. If it takes more then a couple of seconds for temps to decline, then the cooler is likely struggling (for whatever reason) to remove the heat.

That being said, this sounds like normal performance, as others have already stated.

 

[drea.drechsler]

A good test of the cooler would be to watch how fast temperatures go down once load is removed from the system. If it takes more then a couple of seconds for temps to decline, then the cooler is likely struggling (for whatever reason) to remove the heat.

That being said, this sounds like normal performance, as others have already stated.

When I was making my "new cooler or not" analysis I also look at that in an effort to decide. What I have found is a typical characteristic of my two Ryzen systems is how fast die temperatures drop as soon as processing load drops away. With the 3700X in particular it was very much the same under it's (barely adequate) stock cooler as it is under a 240mm AIO. That is even though actual performance, therefore thermal output as the CPU must be working harder, is much better under the AIO.

I think that, too, is a result of how small the actual heat output (not just temperature) is for any one of the very small core regions where individual temp sensors are reporting. Since it's really quite small once it stops working the heat dissipates rapidly not only into the heatsink but into the surrounding die and then into the heatsink.

 

[RodroX]

Another thing to keep an eye on is BIOS updates. Sometimes new microcode could provide imporvements to temps over load. This has happend in the past for Ryzen CPUs.

It also doesn't help that some motherboard manufactures (not all of them, and not for all the models) send the motherboard with some "tweaks" out of the box (factory OC) that make CPUs work harder than they should.

On the other hand if the system is working fine, theres no reason to updae the BIOS, so I would not do it for now.