Monitoring, loop tuning and expansion

[Radeon28]

Aside from better contact between the core and waterblock, we could also consider the plated material between the block and the processor.

Thermal conductivity (measured in W/(m·K)) is the property that tells us how well a material can pull heat off of something. Copper has a value of 401, which is only topped by silver (429), diamond (900-2300) and Graphene (5000). Now considering no one here has enough money to make a WB out of any of those substances (and if someone does, please adopt me!).

Side note: contrary to popular believe, copper is a better material for WB's and rad's than gold (even pure) because it has a higher thermal conductivity and lower thermal resistance.

Getting back on track, of all the materials mentioned, silver is a POSSIBLE option, since its thermal conductivity is higher than the copper, and has a lower resistance than the copper, it could theoretically increase heat transfer if a THIN plate of silver were put between the processor and block. This is also why TIM is made with silver, also, if you were to make the 2 surfaces so flush that TIM wasn't needed... that would increase HT quite a bit.

This is all dependent upon the thickness of the material, I'd have to do a little bit of math to see how thin the sheet would need to be to break even, and then anything thinner than that would be an improvement.

I know its not exactly cost effective, but when your trying to redesign something... its rarely cheap.

 

[QuietIce]

Once you've got the lid off I don't see how putting anything (reasonable) between the block and the core could help unless you can find something to replace TIM ...

 

[Radeon28]

Once you've got the lid off I don't see how putting anything (reasonable) between the block and the core could help unless you can find something to replace TIM ...

Agreed, but I'm just trying to think of ideas other than IHS popping. As far as replacing the TIM goes, I'd make a thin sheet of silver (thickness of about .5mm) that has the dimensions of the core, and set that between the core and the WC. This way we don't have to worry about TIM "spilling" onto the conductive components around the core and its TC and TR would (theoretically) increase the HT since its properties are better than that of copper.

I'm really enjoying this discussion, let's keep it rollin'!

 

[evolvedpc]

Once you've got the lid off I don't see how putting anything (reasonable) between the block and the core could help unless you can find something to replace TIM ...

One thing that you could put between the block and the core is a heat spreader with a wider surface area. The size the the area that is in contact with the block on a typical CPU IHS is pitiful.

Heat will travel fairly readily through the solder to the IHS, but then it has to squeeze through a tiny little square of TIM. A great big square of TIM would work MUCH better.

 

[evolvedpc]

A great big square of TIM would work MUCH better.

Now that I think about it, and IIRC, an i7 has a fairly big area, but only a little square in the middle that is raised up a millimetre or so, for contact with the block. I wonder if it would make any difference to just fill the gap in with TIM, so as to increase the thermal interface area?

In fact, why not just pile up a bunch of solder on TOP of the IHS, around the edges of the raised up bit, (NOT on top of the raised up bit,) and then sand it down flush with the raised up bit. This would form a large surface area that would contact more of the block, allowing a wider thermal path.

 

[m0r7if3r]

snip

Side note: contrary to popular believe, copper is a better material for WB's and rad's than gold (even pure) because it has a higher thermal conductivity and lower thermal resistance.

I have to make a note to this, gold could be considered superior in the same way copper is considered superior to sliver. Gold doesn't tarnish or corrode ever, so, thermally speaking, copper is better, but for things exposed to air, I'm all for gold plating it (inside of it should be copper/silver)

@evolved i think you'd still have thermal inefficiency (and maybe a bottleneck) at the solder. I can only assume they're using lead solder (if they're using silver, correct me) and lead's heat transfer is less than that of silver. This said, radeon's idea of using a thin silver plate is

Now, I have to say, I think that the silver idea would have to be implemented perfectly for it to work right. First off, processor safety precautions, silver is as electrically conductive as it is thermally conductive (iirc, its the best conductor electrically), so you'd have to prevent it from spilling over with this method. I also don't know where silver melts, so i don't know if this is practical, but I think the proper way to apply this idea would be in an indigo xtreme style mount where you melt it in so that it forms the perfect shape and perfectly contacts the cpu. To do this, you would have to burn it in with the computer on its side (indigo xtreme again) and your temps would also not be able to reach a point where the silver melted again, so you couldn't be underradded. Also, with something this small, every nanometer of contact area is vital, so if you could shape a cavity to the cores exactly (getting this with good tolerances is out of my reach, but I've seen the automills they use that are used to manufacture these blocks, and this would be a joke for them. This is dependent upon consistency on intel's part. Now, with this in place, you also want silver on the sides of the core. I also can't help but wonder how much heat you could pull out of the non-core part of the cpu...why don't we all have jobs doing this yet?

 

[QuietIce]

Agreed, but I'm just trying to think of ideas other than IHS popping. As far as replacing the TIM goes, I'd make a thin sheet of silver (thickness of about .5mm) that has the dimensions of the core, and set that between the core and the WC. This way we don't have to worry about TIM "spilling" onto the conductive components around the core and its TC and TR would (theoretically) increase the HT since its properties are better than that of copper.

I'm really enjoying this discussion, let's keep it rollin'!

May as well electroplate it onto the water block if you're going to go that route - but I'm still not sure it's better then direct contact with the water block. Any material you add just makes things worse, doesn't it, unless it's like TIM and dramatically increases contact area? I suppose if the electroplate ended up with an extremely smooth surface it would work but I don't know enough about that process to even guess if it's feasible ...

 

[m0r7if3r]

May as well electroplate it onto the water block if you're going to go that route - but I'm still not sure it's better then direct contact with the water block. Any material you add just makes things worse, doesn't it, unless it's like TIM and dramatically increases contact area? I suppose if the electroplate ended up with an extremely smooth surface it would work but I don't know enough about that process to even guess if it's feasible ...

I think electroplating would leave the same rough surface like you see on an ihs...but I'm not familiar enough with the process to know. You bring up an interesting point, lets assume we can make 2 perfectly smooth surfaces (and I mean perfectly, down to the atomic level or however low you can go), one was your processor core, the other your cpu block, would you need to use tim for this? I think not...so a tim with a weaker structure that can seep into the pores in the metal can bring it one step closer to this ideal situation...what if you were to ridge the top of the cpu core and ridge the bottom of the heatsink so that they meshed, it'd either have to be a not-so-fine ridging (or maybe pins if you wanted it to be less directional) or incredibly precisely manufactured...someone call intel!

 

[QuietIce]

The only trouble I see with the ridges is the TIM application (unless you're still talking about the "perfect surface"). However, making the CPU concave with a convex sink would also increase contact area and the TIM application wouldn't be much different than it is now ...

 

[m0r7if3r]

That would work, but a more effective way might be to reverse the current setup, and have the cpu be convex so that the total distance to the core is less. The counterpoint to this is that, unless you're using heatpipe direct touch you increase the distance that the heat has to travel before it starts being dissipated...

now, if you compare this to ridges/diamonds at a 45 deg angle (which is to say, 90deg bends at each turn, 45 deg above parallel angles), you're effectively doubling the surface area with only a nominal increase in how far the heat has to travel...now, what's to stop you from doing both?

 

[evolvedpc]

On a related note, I found this diamond-based thermal paste on the net. Given that diamond is many times more thermally conductive than copper or silver, it should (theoretically) be better.

Anyone used it?

 

[Elfie]

I've used it. It appears to take about a week to cure fully and gives me the same results as MX-2. In addition when you clean it off it abrades the markings off the CPU so don't use it if there's the possibility of an RMA.

It's a pain to apply and you need to stick the tube in hot water for a few minutes to soften it enough to get it out.

Other people have had better results but I don't think there's enough diamond in it.

 

[evolvedpc]

I've used it. It appears to take about a week to cure fully and gives me the same results as MX-2. In addition when you clean it off it abrades the markings off the CPU so don't use it if there's the possibility of an RMA.

It's a pain to apply and you need to stick the tube in hot water for a few minutes to soften it enough to get it out.

Other people have had better results but I don't think there's enough diamond in it.

Thanks for the feedback - saves me buying it. I was a little suspicious of the "94% diamond content by weight" claim. Seems to me that volume is that only thing that matters here - not weight. I'll stick with my MX-3.

 

[Elfie]

Cool Lab Liquid Pro was better for me http://www.specialtech.co.uk/spshop/customer/product.php?productid=8440&cat=1273&page=1 This gave me a couple of degrees improvement compared to MX-2 but it can cause your heatsink to stick to the block and it's a real pain to clean off. They say you can use metal polish but I've had to lap the CPU(s) and heatsinks/blocks to get it completely off.

I think I still may have some Diamond paste left if you want to try it out, pm me with your details if you want me to send it. I got a 20g tube and used most of it experimenting with different mounts. Contrary to the instructions I got best results with spreading it with a razor blade.

EDIT: found the Diamond TIM, there's enough left for 2-3 applications.

 

[evolvedpc]

I think I still may have some Diamond paste left if you want to try it out,

Many thanks for the offer, but after your less-than-sterling recommendation, I think I'll take your advice and pass.

 

[Owenator]

A few points about the mount/thermal interface and heat spreaders.

TIM:
If we could get the surfaces to be perfectly smooth you wouldn't need the TIM. The TIM is there to fill in the gaps caused by surface's roughness. It looks smooth the the naked eye but on the microscopic level it is not.
from:http://www.bit-tech.net/hardware/2009/02/16/all-about-tim/1

If you had two very smooth surfaces put together they will not need much of a TIM. Older HP UNIX worksations actually used a microfine finish on the CPU and cooler and a simple piece of foil as the TIM.

You can get precision machined shims that are so smooth they practically bond when put together.

We need both flat and smooth to get the best mount possible.

Heat Spreaders:
The idea of a heat spreader is to use something that conducts heat in a "horizontal" direction so the surface area heated can be cooled by a larger contact area.

I used to bond copper plates to the bottom of Aluminium heatsinks to improve their performance. I've even seen silver coins attached to some heatsinks for the same reason. Cold plates are used in peltier setups in a similar fashion.

Basically you get diminishing returns as you add materials. A heat spreader can help but you have more interfaces to deal with. All of these sandwiched materials create more layers which inturn increases the amount of interface resistance to heat transfer.

The top/cover on todays CPU's is not thick enough to act as a true heat spreader. It works more to protect the core and give a bigger surface to mount to.

 

[QuietIce]

I know this post looks long but most of it is quotes.

A few points about the mount/thermal interface and heat spreaders.

TIM:
If we could get the surfaces to be perfectly smooth you wouldn't need the TIM. The TIM is there to fill in the gaps caused by surface's roughness. It looks smooth the the naked eye but on the microscopic level it is not.

If you had two very smooth surfaces put together they will not need much of a TIM. Older HP UNIX worksations actually used a microfine finish on the CPU and cooler and a simple piece of foil as the TIM.

You can get precision machined shims that are so smooth they practically bond when put together.

We need both flat and smooth to get the best mount possible.

Heat Spreaders:
The idea of a heat spreader is to use something that conducts heat in a "horizontal" direction so the surface area heated can be cooled by a larger contact area.

I used to bond copper plates to the bottom of Aluminium heatsinks to improve their performance. I've even seen silver coins attached to some heatsinks for the same reason. Cold plates are used in peltier setups in a similar fashion.

Basically you get diminishing returns as you add materials. A heat spreader can help but you have more interfaces to deal with. All of these sandwiched materials create more layers which inturn increases the amount of interface resistance to heat transfer.

The top/cover on todays CPU's is not thick enough to act as a true heat spreader. It works more to protect the core and give a bigger surface to mount to.

Obviously not impossible to do then! I wonder if something like silver filigree would work?

On a related note, I found this diamond-based thermal paste on the net. Given that diamond is many times more thermally conductive than copper or silver, it should (theoretically) be better.

Anyone used it?

LOL - that's a local product here at OCF!
Here are a few threads started by JoeC, one of their founders (and an old admin of ours ):

http://www.overclockers.com/forums/showthread.php?t=630909
http://www.overclockers.com/forums/showthread.php?t=632321
http://www.overclockers.com/forums/showthread.php?t=645888

They've even tried formulas up to 96% diamond but they didn't work out as well.

That would work, but a more effective way might be to reverse the current setup, and have the cpu be convex so that the total distance to the core is less. The counterpoint to this is that, unless you're using heatpipe direct touch you increase the distance that the heat has to travel before it starts being dissipated...
That would be concave (curved inward) - I got the terms right.
(from the same root as cave, cavern, and cavity)

now, if you compare this to ridges/diamonds at a 45 deg angle (which is to say, 90deg bends at each turn, 45 deg above parallel angles), you're effectively doubling the surface area with only a nominal increase in how far the heat has to travel...now, what's to stop you from doing both?

Again, you'll have TIM issues trying to use grooves (or pins). If you double the contact area but double the amount of trapped air as well that's no good ...

 

[m0r7if3r]

Gah, this is what I get for posting after a LOOONG day or 2, thanks for catching that. I think that if you could go back to the idea of letting the processor heat melt silver into the air pockets you could pull it off, maybe even combine that with the micropolishing. As for the heatspreader concept, its good if you've got a tremendously large cooler surface, but the surface of a heatsink or waterblock is pretty close to that of the core, so I don't think its quite necessary for how the high performance pc community uses it.

 

[Owenator]

Gah, this is what I get for posting after a LOOONG day or 2, thanks for catching that. I think that if you could go back to the idea of letting the processor heat melt silver into the air pockets you could pull it off, maybe even combine that with the micropolishing. As for the heatspreader concept, its good if you've got a tremendously large cooler surface, but the surface of a heatsink or waterblock is pretty close to that of the core, so I don't think its quite necessary for how the high performance pc community uses it.

I wouldn't discount the heat spreader that quickly. The actual core is not that big compared to the heatsink. Even an i7 core has a surface area of 263square mm under it's cap.

A typical water block is on the order of 50 mm x 50 mm or 2500 square mm. That's a 10:1 ratio. This is not a trivial difference. Especially since the "top" on the i7 is far to thin to act as an efficient heat spreader. For a waterblock the copper base acts as a heat spreader. Most good heatsinks have some for of copper integrated to help spread the heat as well. They essentially have incorporated the heat spreader in this way.

I agree that heat spreaders - in their current form - are not needed as an add-on if you have a good cooler. But once we can get a better flat heatpipe/heatspreader I would be willing to bet that cooling would increase significantly. Well, I can dream right?

 

[m0r7if3r]

Wonder what the plausibility of making a larger core on the same architecture is...basically increase the surface area so that you can cool more efficiently...