Silver vs. Copper

[Blain]

Water

Why do you say "water"?

 

[serpretetsky]

Why do you say "water"?

Choice of liquid depends on application of the heatpipe. I don't think common computer heatpipes use water, but water is one of the choices when designing heatpipes.

Acetone and methanol I think are the most common ones for 0-100C temperatures.

 

[LtGoonRush]

First, let me correct your statement that Ag shows a 5% edge over Cu in thermal conductivity. The table of metals and their thermal conductivities show it to be more like 20%. How that translates into actual cooling performance -- which might be a piddly 5% after all -- hangs on Aigo's analog.

Please check your sources, the measured conductivity difference should be about 5%. I think you might be looking at conductivity measurements performed at different temperatures or perhaps on different purities of metal, which would not be comparable. No disrespect intended as measuring and comparing conductivity can be complicated!

Here are some sources:
Engineering Toolbox: Cu 223 Ag 235, ~5% difference, both in BTU/hr/deg-F/ft, at 68F
Georgia State University Hyperphysics: Cu 385 Ag 406, ~5% difference, both in W/mK, temp not specified
Engineer's Edge: Cu 386 Ag 418, ~8% difference, both in W/mK, temp not specified
TIBTech: Cu 401 Ag 420, ~5% difference, both in W/mK, temp not specified

 

[BonzaiDuck]

Please check your sources, the measured conductivity difference should be about 5%. I think you might be looking at conductivity measurements performed at different temperatures or perhaps on different purities of metal, which would not be comparable. No disrespect intended as measuring and comparing conductivity can be complicated!

Here are some sources:
Engineering Toolbox: Cu 223 Ag 235, ~5% difference, both in BTU/hr/deg-F/ft, at 68F
Georgia State University Hyperphysics: Cu 385 Ag 406, ~5% difference, both in W/mK, temp not specified
Engineer's Edge: Cu 386 Ag 418, ~8% difference, both in W/mK, temp not specified
TIBTech: Cu 401 Ag 420, ~5% difference, both in W/mK, temp not specified

WOW! I STAND COR-RECK-TED!! As I said, it was a matter of "what I remembered" from a similar table. The over-the-hill brain had rounded down the figure for Cu, so I was imagining a ratio of 3 to 4 -- not 385 to ~406.

I feel so STOO-pid!! Sorry!

Also -- my assertion about diamond was excessive: it's only about 2.5 times that of the other two substances, or 250%. That's still a big difference, though.

 

[Zeldak]

no because your stuck at the limited element.

Lets assume this..

You have a 18Wheeler truck, and a Uhal truck.

The 18wheeler truck can deliver a full package of whatever, however when the Uhal truck can not move the same capacity in 1 trip unless it was to drive a lot faster, hence the potential of transfer will be limited at the Uhal with very little gain.

The same can be said for the other way around.
This however is then known as holding potential.
The uhal truck can bring packages to the 18wheeler, but will never saturate it or fill it up with 1 trip, the holding potential will be greater at the intial end, but when the system reaches equilibrium, you will again be limited by the lower potential as it exists the system.

I think this analogy makes sense. It reminds me of the rationale for RDRAM.

Please check your sources, the measured conductivity difference should be about 5%. I think you might be looking at conductivity measurements performed at different temperatures or perhaps on different purities of metal, which would not be comparable. No disrespect intended as measuring and comparing conductivity can be complicated!

Here are some sources:
Engineering Toolbox: Cu 223 Ag 235, ~5% difference, both in BTU/hr/deg-F/ft, at 68F
Georgia State University Hyperphysics: Cu 385 Ag 406, ~5% difference, both in W/mK, temp not specified
Engineer's Edge: Cu 386 Ag 418, ~8% difference, both in W/mK, temp not specified
TIBTech: Cu 401 Ag 420, ~5% difference, both in W/mK, temp not specified

Measuring thermal conductivity MUST be less than straightforward, else there wouldn't such differences in reported values. I've seen some reported of about 18%. Maybe not everything on the Internet is entirely accurate. Still, if the Lone Ranger were into overclocking, I'll bet he'd have a silver water block---or at least a silver heatsink.

WOW! I STAND COR-RECK-TED!! As I said, it was a matter of "what I remembered" from a similar table. The over-the-hill brain had rounded down the figure for Cu, so I was imagining a ratio of 3 to 4 -- not 385 to ~406.

I feel so STOO-pid!! Sorry!

Also -- my assertion about diamond was excessive: it's only about 2.5 times that of the other two substances, or 250%. That's still a big difference, though.

As OP, any STOO-pidity here is rightfully MINE! I am pleased, though, that as a result of this discussion I am less STOO-pid--or at least better informed--than I was at first.

 

[Subyman]

Would be fun to work up a diffusion equation and calculate the heat transfer for various interfaces and materials. Shouldn't be too hard, the boundary conditions are well defined. It would require a finite difference calculation once the equation was constructed, there wouldn't be an analytical solution. I may do it later. I haven't used Mathematica in a while.

 

[BonzaiDuck]

Measuring thermal conductivity MUST be less than straightforward, else there wouldn't such differences in reported values. I've seen some reported of about 18%. Maybe not everything on the Internet is entirely accurate. Still, if the Lone Ranger were into overclocking, I'll bet he'd have a silver water block---or at least a silver heatsink.

It might seem that way for me, but I was able to get my own accurate results comparing thermal pastes under controlled ambient and test-bed conditions. I became enthused about nano-diamond when I came across some white-papers and heatsink comparisons of Joe Citarella on "OverClockers.net" -- who was a founding principal in Innovation Cooling, if I'm not mistaken.

Citarella ran his tests comparing thermal paste formulations with a $20,000 device called a calorimeter. I don't know from this or that or any calorimeter, and I don't have $20K to spare. Howsoever that may be, somebody did.

If the Lone Ranger used silver bullets, he must have had a cheap source. Of course, there's that rather funny Johnny Depp ("Tonto") script-line: "Mmmm. Bad Train!" -- before the train filled with silver-ore goes over the cliff.

 

[blastingcap]

Without going into too much detail, you can't just look up numbers at engineering toolbox, which has wrong numbers for some of their materials. I mean WRONG. It's what you get for crowdsourcing scientific info, I guess. Don't believe everything you read on the internet.

Furthermore, nobody uses 100.00000% pure copper or aluminum in mass production. For heatsinks the copper is probably C11000 grade, which is very pure (at least 99.9%) but with some impurities. Not sure what purity silver they'd use for heatsinks but I'm guessing pretty high purity as well.

The thing is, even tiny amounts trace impurities can have devastating effects on electrical and thermal conductivity, depending on what the impurity is.

Anyway, long story short, if you compare C11000 copper to the equivalent purity silver, the thermal conductivity delta is probably less than 10% whereas the cost difference is way the hell more than 10%. Why bother jacking prices up so much for so little gain? Scientific tasks or some high-value military thing, maybe, but PC cooling? Not worth it. Unless it's for marketing purposes... witness the ridiculousness of companies trying to sell "military spec component" motherboards and such.

Also, silver is not butter soft, it's freaking hard compared to many things around the house made of wood or plastic or whatever. You would not want to be hit in the head with a silver hammer. But compared to other metals like copper, it is soft. Copper itself is kinda soft when compared to other metals, and especially to alloys like steel.

 

[BonzaiDuck]

If there were only a way to fabricate a solid synthetic (SYN- THETIC) diamond heatsink base to metal heatpipes with a minimum loss of thermal conductivity in the interface, water cooling would still prevail if you could make a water-block the same way.

But I might need two hands to count the people here who've already said this: if the effective difference between copper and silver is only <= 10%, it doesn't make sense to use silver unless money doesn't matter. What you'd get for the money is probably much less than what the forthcoming Noctua NH-D15 will provide in terms of cooling.

Let me put that another way. Water is of course a superior method of cooling because it transfers heat more efficiently than air. But using a silver water-block is only going to give an improvement to a water system that is proportional to that you get with a heatpipe cooler -- unless the fins and pipes of the latter are the final bottleneck in the air-cooling process that doesn't exist with water.

If the heatpipe coolers were bottlenecked primarily at the copper base, then the Noctua Cu-C(diamond) composite will remove some of that bottleneck to be replaced by the natural limitation of air, fins and pipes.

If they used the same composite for a water-block, you might see more significant gains in cooling performance. But the difference between Copper and Silver probably won't amount to a hill-a-beans. Certainly, the outlay for silver will amount to more -- than a hill-a-beans.

 

[Blain]

Even if a cooler transferred 100% of the heat produced, it would still never get the CPU below ambient air temp.

 

[Essence_of_War]

Water is of course a superior method of cooling because it transfers heat more efficiently than air. But using a silver water-block...

And don't forget, if you want to use pure water, not only do you have to have a silver waterblock, but you need a silver radiator too. On the plus side, does that mean you could do without a silver kill-coil?

 

[serpretetsky]

Even if a cooler transferred 100% of the heat produced, it would still never get the CPU below ambient air temp.

I agree with the point you're making, i just wanted to point out that at steady state, ALL coolers transfer 100% of the heat (assuming motherboards are perfect insulators).

I think you meant that even if you had a passive cooling system that had infinite thermal conductivity, it still would never get the temps below ambient (in fact, it would not allow temps to go below ambient!)