News DIYer Passive Cools Core i9 with 8-Pound Copper Block

[Admin]

An Intel Core i9 CPU was successfully cooled by an 8lb copper block salvaged from medical machinery, but we aren’t sure how well it would have held up under prolonged workloads.

DIYer Passive Cools Core i9 with 8-Pound Copper Block : Read more

 

[PlaneInTheSky]

makes no sense since the surface area by size and weight is tiny

would be much better and smaller if he just stacked a few old copper zalman heatsinks

 

[USAFRet]

makes no sense since the surface area by size and weight is tiny

would be much better and smaller if he just stacked a few old copper zalman heatsinks

For a DIY "what if", it makes perfect sense.

This is a "I have this big block of copper. Let's try it, see what happens" type of thing.

 

[TerryLaze]

For a DIY "what if", it makes perfect sense.

This is a "I have this big block of copper. Let's try it, see what happens" type of thing.

I have this big cylindrical chunk of conductive material...let's try and balance it on a $600 CPU that is very susceptible to electric conductivity, without any protection!

(Yeah the CPU is protected by the heat spreader, but the rest of the system isn't, I'm just making a point )

 

[USAFRet]

I have this big cylindrical chunk of conductive material...let's try and balance it on a $600 CPU that is very susceptible to electric conductivity, without any protection!

(Yeah the CPU is protected by the heat spreader, but the rest of the system isn't, I'm just making a point )

There's that too.

But, for running for an hour or so, eh...lets try it.

Also, if it fell over, the CPU probably wouldn't be damaged. The motherboard, however...

 

[InvalidError]

The specific heat of copper is only 0.38J/gram vs 4.17/gram for water. He'd get much better results by hollowing out the middle (maybe leave a 1" cone at the bottom to increase copper-water surface area a bit, taper the wall thickness from 2mm at the open end to 1cm at the base, leave 5mm at the base plate for heat-spreading) and filling it with water. Or go one step further: fill the pile 2/3 of the way with water, seal the open end and vacuum it down to 4kPa (partial pressure of water at 20C) to make it into a huge heat pipe.

That way, he'd be able to run the CPU ~10X longer between bursts. Though it'll also take 10X as long for the water to cool back down.

 

[Early Microchip]

An Intel Core i9 CPU was successfully cooled by an 8lb copper block salvaged from medical machinery, but we aren’t sure how well it would have held up under prolonged workloads.

DIYer Passive Cools Core i9 with 8-Pound Copper Block : Read more

What is misunderstood here is that the copper block is acting only as a "heat sink". It will absorb [sink] heat from the device and at the same time it will lose some heat from its surface area until it reaches equilibrium, i.e., when the heat put in equals the heat radiated or convected to the ambient air from its surface. Then its temperature will rise and thus may not prevent the device's critical junction temperature [Tj] from being exceeded. What we normally refer to as a heat sink is better called a "cooler", its job is to collect heat from the device, conduct it to its fins and allow the heat to escape by a process of convection and radiation to the ambient air.
Heat flows from the hotter body to the colder, it does not flow the other way, and if the ambient air temperature is high the "cooler" becomes less efficient. Mistaking thermal conductivity for cooling ability is a common mistake, Aluminium is preferred for "cooler" manufacture due to low cost, workability and low weight. Copper is a better conductor, Silver is the, but both are costly. Fins are vital to increase surface area, but unless the air is forced air by a fan or blower, must be spaced more than 8 to 10mm apart to avoid radiation between fins. Staggered fins to increase turbulence are better than straight ones as they break up any boundary layers that might occur in the air flow. Way back in the 60's and 70's when devices were much less efficient, many useful "heat sinks" [although better known as coolers] were available. Mostly plain aluminium was used, black anodising was offered, but the improvement gained was not cost effective. Industry tried many alternatives [ebullition cooling, liquid cooling, phase change heat pipes and even thermally connecting the devices to the fuel flow in military aircraft, as an airplanes fuel tanks are a great source of "cool"!
But all methods must move the device's heat to somewhere else. In the extreme you could machine a PC Cabinet from a solid block of copper, machine grooves all over its surface, then crosscut them to stagger them, but if the device's dissipation is more than the cabinet can dissipate in free air, you might have to direct fans at it to help it. The thermal interface between the cooler and the device is CRITICAL. Unless the two surfaces are flat to a very high degree, some interface material, thermal grease, thermally conductive polymers, or metallic loaded elastomeric compounds are necessary.
Just as a reminder, a cooler for a semiconductor device will be rated in what I always called "Thermal Ohms" or more correctly, thermal resistance in Degrees Centigrade per Watt; if you knew the ambient temperature and the device's power dissipation you could easily calculate what size heatsink to use. Lastly it's important to know there are three methods of moving heat, Conduction, Convection and Radiation, most are misunderstood, the first example to come to mind is that here in the UK we heat our houses with water heated by gas or oil, and piped to Radiators, which in fact are not very good radiators at all [because we paint them] but are in fact convectors.

 

[TechieTwo]

Copper is a good absorber of heat but once it's saturated it's of little value unless the heat can be disappated. That's why heat sinks have fins and fans. Water is often an unnecessary liability.

 

[InvalidError]

Water is often an unnecessary liability.

If your goal is to buffer transient load, water is more effective than solid copper and much cheaper.

 

[jasonf2]

This thing is a thermal battery not a fully functional passive copper sink. The thermal mass of the metal is absorbing the the heat until the delta between the battery and cpu get too close and then throttles at load. Even with a fan big air coolers have quite a bit more surface area for exchange and they are often coupled with a vapor chamber to let the heat distribute evenly on the sinks.

 

[jasonf2]

The specific heat of copper is only 0.38J/gram vs 4.17/gram for water. He'd get much better results by hollowing out the middle (maybe leave a 1" cone at the bottom to increase copper-water surface area a bit, taper the wall thickness from 2mm at the open end to 1cm at the base, leave 5mm at the base plate for heat-spreading) and filling it with water. Or go one step further: fill the pile 2/3 of the way with water, seal the open end and vacuum it down to 4kPa (partial pressure of water at 20C) to make it into a huge heat pipe.

That way, he'd be able to run the CPU ~10X longer between bursts. Though it'll also take 10X as long for the water to cool back down.

It will take longer to trip, but only because you were able to distribute the heat in the battery more effectively. Due to the more even distribution of the heat to the external surface area cooling should be a bit more effective as well. Regardless though just like all thermal solutions the crux is always that final exchange to ambient. Without a fan to mix the ambient air you are completely at the mercy of the convective flow created by the air being heated on the surface of the sink. As such the cylinder gets hotter at the bottom, which then moves heated air up onto the cooler upper regions of the sink and creates an overall interface region with a really inefficient delta. I would imagine that a person would be much better off strapping a copper sculpture of a tulip plant on top. At least at that point you would have more surface area and the copper/ambient interface region would cause some turbulence and allow the upper regions of the sink to stay out of some of the air that the lower regions had already heated.

 

[InvalidError]

Regardless though just like all thermal solutions the crux is always that final exchange to ambient.

That depends on the delta between average and peak power. A large copper cylinder like that can probably handle 50W continuous at 60C surface temperature, all you need is enough specific heat capacity to average peaks down to 50W without exceeding Tcasemax.

 

[jasonf2]

That depends on the delta between average and peak power. A large copper cylinder like that can probably handle 50W continuous at 60C surface temperature, all you need is enough specific heat capacity to average peaks down to 50W without exceeding Tcasemax.

First off I have to start out and say I agree roughly with your transfer assumptions. But I qualify that, if I were doing this as a DIY project, success would only be achieved if the operating prototype was able to meet the requirements of a real cooler in my given application. So on the most basic level that means that the design must be able to keep the CPU from thermal throttling under normal operating conditions and design power envelopes. If there isn't enough surface area and the interface shape reduces conductive transfer efficacies to to a point where it cannot offload that heat to ambient the design is a failure by my metrics. Thermal mass will let you store and dissipate the energy over time but regardless the sum of heat in (from CPU plate) to heat out (to ambient) has to be net zero or the system eventually overheats. What this article really should have been titled is "Why don't CPU cooler manufactures just use copper bar stock?".

 

[InvalidError]

So on the most basic level that means that the design must be able to keep the CPU from thermal throttling under normal operating conditions and design power envelopes.

And that would depend on what the "normal conditions" or "design envelope" for a given application are. I'd imagine there is no shortage of practical applications where you need bursts of high performance processing with a sub-10% duty cycle such as just about all things camera-driven with external triggering telling algorithms when to do their things with the result desired ASAP followed by a relatively long pause until the next trigger or anything primarily limited by manual user input such as typing.

Not all applications require 100% sustained peak power 100% of the time.

 

[bit_user]

The specific heat of copper is only 0.38J/gram vs 4.17/gram for water. He'd get much better results by hollowing out the middle (maybe leave a 1" cone at the bottom to increase copper-water surface area a bit, taper the wall thickness from 2mm at the open end to 1cm at the base, leave 5mm at the base plate for heat-spreading) and filling it with water.

Are you saying an aluminum can full of water would make a better thermal mass than solid copper? I like it!

When I've left a can of soda in the freezer too long, what often happens is the bottom becomes convex, before the thing actually explodes. If you wait and get a bit lucky, you could have an almost perfect base for your passive water-based cooler. Then, perhaps some delicate work with a hammer could form a flat spot, in the center.

Of course, you'd need to add supports to ensure it didn't tip over onto the motherboard...

Or go one step further: fill the pile 2/3 of the way with water, seal the open end and vacuum it down to 4kPa (partial pressure of water at 20C) to make it into a huge heat pipe.

But this is the key point, isn't it? That heat pipes have a far higher thermal conductivity than even solid copper!

 

[bit_user]

In the extreme you could machine a PC Cabinet from a solid block of copper, machine grooves all over its surface, then crosscut them to stagger them, but if the device's dissipation is more than the cabinet can dissipate in free air, you might have to direct fans at it to help it.

That's often what you find amongst "industrial" PC cases that are sealed against dust, spray, or potentially corrosive gasses.

There have also been a smattering of high-end passively-cooled PCs for home & professional users, but the price can be very off-putting.

In the realm of lower-powered "NUC" type machines, it's not hard to find examples and they tend not to be so extreme.

https://www.tomshardware.com/news/bleujours-meta-u-nuc-intel-elements

 

[bit_user]

I would imagine that a person would be much better off strapping a copper sculpture of a tulip plant on top.

Relevant:

https://www.tomshardware.com/news/forget-water-cooling-check-out-this-7300-passive-cpu-cooler

​

BTW, that's obviously rendered, and the interface between the base & balls looks very poor, but it's getting towards the general idea. I'm thinking it would've been cool if you had copper trees springing out of the base, all the way to the point of fabricating tiny little leaves.

 

[InvalidError]

Are you saying an aluminum can full of water would make a better thermal mass than solid copper? I like it!

Even as a water pot, you'd still need enough base thickness to prevent flex under mounting pressure and provide a bit of heat-spreading. If you are going to run a vacuum to make it into a heat pipe, you'd also need enough thickness to prevent implosion and flex all around.

 

[wintercounter]

Let's make the house also from copper and attach this way to the sidepanel. (And you can also start cooking on the top, but at least keep your coffee warm.)

 

[PhilTroy]

Hi

FYI only, I took apart the case of an ASRock Industrial 4x4 with a AMD V1605B processor (having 4 cores and 8 threads). I rebuilt the case with a copper plate connected with thermal paste to the processor, another larger copper thermal plate connected to the first copper thermal plate, an aluminum plate ( around 4 inches by 4 inches) attached to the second copper plate, and 4 heat sinks attached to the aluminum plate. I painted the aluminum plate sides black with rustoleum (or similar) paint. I then added RyzenAdj to run automatically on my computer, and I usually run psensor to monitor the temperature.

About a week ago I was running the processor pretty much full out with ffmpeg on all 4 cores (and presumambly 8 threads). The power consumption monitored by my Belkin monitor was around 45 watts (much higher than usual but ffmpeg was running all out). I am pretty sure that the Tctl temperature was less than 80 degrees Celsius - usually it is much lower but the peak temperature typically runs up to the 80 celsius mark.

If I were to it over again, instead of using using 4 heat sinks connected to the aluminum plate, I would try to get a single larger and thicker heat sink as this would save a lot of the work.



Phil

 

[Hresna]

Der8auer uses small copper “dice” blocks to test delidded CPUs before mounting coolers on them. You won’t run them for long with any sort of heat mass that isn’t optimized for dissipating the heat, but they can be a buffer to hold enough for a practical purpose such as this

 

[who_farted]

Let's make the house also from copper and attach this way to the sidepanel. (And you can also start cooking on the top, but at least keep your coffee warm.)

good luck in a lightning storm.

 

[Amdlova]

A long time ago I have a passive 4770k with a thermalright ultra 120 rev c. And a 12cm akasa piranha Fan activates when the cpu reaches 90 degree celciuls. My friend still using that system. Even Xeon 5675 with another thermalrights coolers I have used passively. If have 800 gram can be done.

 

[Amdlova]

https://silentpcreview.com/thermalright-ultra-120-extreme-rev-c-ac-freezer-xtreme-rev-2/ miss that heatsink

 

[jmcgaw]

Not even a good test was run. Clearly the copper block will slowly reach some steady-state between the core temperature and the ambient and with the small surface area it will not be advantageous to the processor. More interesting would be to machine numerous vertical fins into the cylinder so that at at least a bit of efficient convective cooling will help out. But while we are on the subject, silver is a better thermal conductor than copper by a good margin. Even better, far better, is diamond. Somebody needs to do the test over again but with a bit of bling.