Solid copper block between CPU and cooler for heat buffer

[Andres F Borrero]

So I am saving to make a new gaming PC... and while I want it to be top end and with excellent cooling and lots of fans, I also want it to be silent while I am not gaming or doing heavy workloads. I watched Linus' M1 Macbook Air video where they showed the cooling solution, a heat buffer piece for dissipating power spikes, and I thought if it could also apply for a gaming PC.

 

My idea (which might not be a new one) is the following: Place a copper block above the CPU and then place the cooler on top of the copper block. The purpose is to buffer power peaks under normal loads so that the fans don't ramp up (e.g. every time I compile my code). I'd like the forum's opinion on whether this could work, the impact in thermals under prolonged loads, possible effects on other components, and if it would achieve its initial purpose. I know water cooling has this buffer efect on thermals, but I am not a fan of it since I don't consider it reliable and the pump noise is annoying for me (personal opinion).

 

My main concern is having an extra contact in the heat exchange process which could affect prolonged workloads, so I'd be using liquid metal thermal paste. The expected CPU is a Ryzen 9 5900X (the 24 threads help me a lot in machine learning and development solutions).

 

I chose copper for its thermal conductivity and high thermal mass but I am open to options. I know I could work around the adaptations for making this solution without compromising the motherboard's integrity. Price is not a concern also.

 

Thank you all, I shall be reading all the comments eagerly.

[Levent]

Adding more surface area does not mean you can reduce sudden peaks. Smaller the core die, harder it will be to keep it cool when a sudden power burst occurs. Your idea will not work.

[AlwaysFSX]

2 minutes ago, Levent said:

Adding more surface area does not mean you can reduce sudden peaks. Smaller the core die, harder it will be to keep it cool when a sudden power burst occurs. Your idea will not work.

To add to this, it'll completely kill the efficiency of heat pipes too. Those need to be closer to the die to wick heat away.

[Entropy.]

Likely won't work well if at all. I'd just put the CPU under water, an aio like the arctic liquid freezer ii would be just fine. What's your budget for the build? 

[Guest]

It's not a bad idea, but in this fashion would increase your temp gradient. So the fans would be on for longer durations while the cooling process will take additional time to dissipate stored heat in your copper block. However, it would take a longer period of time to get to a high enough temp to spin the fans up. And Ultimately, your max temps will be higher.

 

Paste, no matter what kind you use is very poor of heat transfer. The more layers you add, the less thermal conductivity you get. 

 

The better approach would be to remove and replace the IHS plate with something a little larger. This will give you a better temp gradient (higher) but will increase the amount of time it takes to heat soak your cooler. 

 

Since you mentioned "all copper" block, I had originally thought all copper water block, which is probably the better way for you to go. Most custom loops can be built to be virtually silent. Just depends on how much radiator surface area you have. But the waterblock makes a load of difference removing heat.

 

Any how, do experiment. Find different sizes and thicknesses so you have some variety of testing to do. 

 

 

 

 

 

[AnonymousGuy]

The correct answer is to change the response interval of your fans to be longer.  That way it won't instantly try to spin them up just because the CPU temperature briefly went up a few degrees under load.

 

Or place a temperature probe elsewhere on the heatsink and base the fan speed off that, but not a lot of motherboards support external temperature probes in their fan configuration.

[Enderman]

This would only work if the thermal sensor was in the heatsink.

Since the thermal sensor is in the CPU it will still spike up and down depending on load.

If you want to stop the fans from ramping up and down you need to adjust the delay or ramping speed.

Also putting copper between the CPU and a cooler is bad because it acts as a bottleneck.

Heatpipes have more than 10x the conductivity of copper.

[metaleggman]

If you've got the budget for a 5900X, you probably have have the budget for a high end air cooler or triple rad AIO, along with a case to fit it all in.  Assuming you're using case fans and a cooler from someone like Noctua or Be Quiet!, along with a PSU with nice fan control, you should have a relatively quiet build, even at full tilt.  Using software like Argus Monitor will give you granular control over the fan curves if you go air cooled.  And with an AIO or custom loop triple rad, you pretty much just set fan speed to whatever the highest fan speeds are that are inaudible in your environment, since you'll have enough thermal mass unless you're talking hours of maxed out workloads (in which case, you'll want those fans to ramp up I'm guessing).

For my 3600 using a Noctua NH-U12S with 5 NH-P14s redux fans (3 in, 2 out), Seasonic Focus PX-850, an EVGA 2060 XC Black, and a modified fan curve using Argus, even at full tilt, my rig doesn't go above like 43 dB (though I'm using Smart Tools on my phone, so it's not super accurate or precise, but goo enough for government work), with ambient noise being about 25 dB.  The old fridge in the same room is significantly louder when it's heating up the coolant.  Plus Noctua fans aren't really annoying, they kind of pleasant, with a sound similar to white noise than a whining turbine like cheaper fans.

My point is, while your idea is cool and interesting, why fix something that isn't broken?  It could be a fun project on an older machine you found on the side of the road, but stick to something known to work for a nice piece of silicon like a 5900X.

[Andres F Borrero]

15 hours ago, Levent said:

Adding more surface area does not mean you can reduce sudden peaks. Smaller the core die, harder it will be to keep it cool when a sudden power burst occurs. Your idea will not work.

The surface area stays the same, the thermal mass is what increases.

[Levent]

1 minute ago, Andres F Borrero said:

The surface area stays the same, the thermal mass is what increases.

Your main issue is the size of the core die, not thermal mass.

[Andres F Borrero]

14 hours ago, Enderman said:

This would only work if the thermal sensor was in the heatsink.

Since the thermal sensor is in the CPU it will still spike up and down depending on load.

If you want to stop the fans from ramping up and down you need to adjust the delay or ramping speed.

Also putting copper between the CPU and a cooler is bad because it acts as a bottleneck.

Heatpipes have more than 10x the conductivity of copper.

True... I'd have to work around delaying the CPU ramp up... but the overall conductivity might be very bottlenecked

 

[Andres F Borrero]

14 hours ago, ShrimpBrime said:

It's not a bad idea, but in this fashion would increase your temp gradient. So the fans would be on for longer durations while the cooling process will take additional time to dissipate stored heat in your copper block. However, it would take a longer period of time to get to a high enough temp to spin the fans up. And Ultimately, your max temps will be higher.

 

Paste, no matter what kind you use is very poor of heat transfer. The more layers you add, the less thermal conductivity you get. 

 

The better approach would be to remove and replace the IHS plate with something a little larger. This will give you a better temp gradient (higher) but will increase the amount of time it takes to heat soak your cooler. 

 

Since you mentioned "all copper" block, I had originally thought all copper water block, which is probably the better way for you to go. Most custom loops can be built to be virtually silent. Just depends on how much radiator surface area you have. But the waterblock makes a load of difference removing heat.

 

Any how, do experiment. Find different sizes and thicknesses so you have some variety of testing to do. 

 

 

 

 

 

I don't mind fans on for longer durations, especially since high end fans are very silent at low-mid speeds. My main concern is the extra layers reducing thermal conductivity...

 

I might reconsider a waterblock, last time I saw one it was too noisy, but it might've been a cheap or old one. 

 

I could try experimenting on an older system. Still, I could try a copper block on top of the heatpipes that take heat to the radiator, this way there is a more direct and traditional heat path but a larger thermal mass for spikes... But I think this is way more complicated and requires a cooler with enough space below it...  

[Guest]

24 minutes ago, Andres F Borrero said:

I don't mind fans on for longer durations, especially since high end fans are very silent at low-mid speeds. My main concern is the extra layers reducing thermal conductivity...

 

I might reconsider a waterblock, last time I saw one it was too noisy, but it might've been a cheap or old one. 

 

I could try experimenting on an older system. Still, I could try a copper block on top of the heatpipes that take heat to the radiator, this way there is a more direct and traditional heat path but a larger thermal mass for spikes... But I think this is way more complicated and requires a cooler with enough space below it...  

The easiest way to get a tower cooler to do better is putting a fan right at the base. This will start dissipating heat right away opposed to waiting for it to travel up heat pipes.

[akio123008]

17 hours ago, Andres F Borrero said:

I watched Linus' M1 Macbook Air video where they showed the cooling solution, a heat buffer piece for dissipating power spikes

Ah I think I see where your theory might be wrong.

 

You see, the macbook cooler is tiny, and can't dissipate a lot of heat. Therefore the increased thermal mass allows it to "cheat" for a little bit, and absorb more heat than it can dissipate, absorbing power spikes.

 

However, a desktop cooler can dissipate plenty of heat, as it's much bigger. The bottleneck isn't heat dissipation here, the bottleneck is heat transfer. Hence why a buffer won't help, as it doesn't increase the heat transfer rate.