Direct IHS Liquid Cooling

[Moto7451]

I wouldn't bother modifying the IHS. The reason this could/would/should hopefully work better is because you're removing something thats in the way of thermal transfer. The problem with this method is that you may not get enough turbulence.

One idea I wanted to try, was to make a jet block by putting an insert with holes (think shower head) into the input barb and try hitting the IHS with high speed streams of water. Hopefully something like this would reduce the laminar flow of the water across the IHS.

The only way to know for sure is to try it... or get someone in here with access to scientific software .

 

[jaydee116]

I ran this tonight. Results were pretty much as expected.
Sempron 2500+ @ 1862mhz approx 82watts stock Vcore 1.45V
Direct IHS Cooler:
@3.6GPM water to cpu differential was 12C. (Water 25C, CPU 37C). I decided to stop there simply because there is no point in going further.

Stock AMD Sempron Air Cooler 50C CPU temp (32C air to CPU differential)

AquaXtreme MP-05 SP 432pin base:
@1.5GPM water to CPU temp differential 10C
@1GPM 11C
@.5GPM 12C

Danger Den Maze 4
@2GPM 14C
@1.5GPM 15C
@1GPM 15C
@.5GPM 16C

So obviously if the direct IHS cooler is performing worst at over twice the flow rate than the MP-05 SP then it shows this is not a viable option. I don't think modifying the IHS will make it enough better to even beat out the Maze 4.

So in my mind project successful and enough info to know this is not the path to follow. A water block will perform better and less chance of failure.

I wouldn't bother modifying the IHS. The reason this could/would/should hopefully work better is because you're removing something thats in the way of thermal transfer. The problem with this method is that you may not get enough turbulence.

One idea I wanted to try, was to make a jet block by putting an insert with holes (think shower head) into the input barb and try hitting the IHS with high speed streams of water. Hopefully something like this would reduce the laminar flow of the water across the IHS.

The only way to know for sure is to try it... or get someone in here with access to scientific software .

That is exactly what I did with my direct die cooler.
Ignore the link in the pics, was my old website several years back.

The results were not great. To much heat in to small an area.

 

[attack]

Any plans to de-lid the IHS and go direct die? Or perhaps a few rubs with 80 grit sand paper to allow for turbulated flow and more surface area?

 

[jaydee116]

Any plans to de-lid the IHS and go direct die? Or perhaps a few rubs with 80 grit sand paper to allow for turbulated flow and more surface area?

I don't see any point in it. Only way to make it cooler is a lot more surface area. Roughing up the IHS would get maybe a C difference. Taking the IHS off would just make it worst as you just removed even more surface area.

These parts are going right back into my Linux box with the stock air cooler attached.

 

[Moto7451]

Basically, unless they come out with finned IHS (basically a glued on heatsink) there's no point.

Thanks for the data. If nothing else... you saved me from cutting up my plastic collection .

 

[Mycobacteria]

I like the idea,

Its no different to any other block if it leaks your still stuffed.

You dont need to use AS5 anymore, so its direct contact.

The top can be made of acrylic its the water that is required to pick up and transfer the heat not the plastic.

 

[Rickpatbrown]

Water is not as thermally conductive as copper, right? (sorry, I don't feel like looking it up). Copper is better at conducting the heat from the CPU, but it can't move it very far since temperature gradient diminishes as you move from the source.

Water doesn't have this problem since you can move it with a pump. Increased surface area is the real problem. That probably why CPU blocks perform better. My guess is that you can't get any better than a well designed block on a naked core. Which is pretty much like you have, but the design on the block is very poor. Not yours, but the heatspreader isn't designed to create turbulence.

What we need for a revolution in liquid cooling is something that transferrs heat better than Water.

What do I know, though? I haven't taken physics yet. I did dream that I was failing physics this afternoon though. Does that count?

 

[MattODoom]

Pretty slick, wish it worked better, reminds me of my old Athlon XP with an exposed die.

 

[Moto7451]

Water has a much higher heat capacity and a few other factors that makes it better than water or a large copper heat sink. You are correct that copper conducts heat better than water.

 

[Jinu]

Not sure if this would work, or even if its a brilliant idea; however, have you thought about taking a dremel to the IHS and creating some groves to create turbulence?

 

[Joe Camel]

I have a CNC mill that I made this with so no need to saw. However I have no idea how thick the IHS is in the middle.

assuming AMD hasn't changed the IHS design in a few years... HERE is your answer.

(.094")



"LOTS" of copper to play with


.

 

[mepis]

I don't see how this is a failed experiment. You have just proven that you can easily produce and sell a very cheap "water block" or heat exchanger for a water cooling system for low income folks like myself. I would be excited to see something like this on the mass market. I could easily see prices sub $10 for something like this.

I mean even with the costs of a this with a total water cooling system your looking at more money or equivalent than a very good air cooler. But to sell the part stand alone it gives low income enthusiast a way to play with water cooling and scrounge the parts on my own. If you sold this for say $10, got a good old fashioned heater core for ~$20, pump used, scrounged, or from a junkyard, etc.. for roughly ~$20 and $10 for home depot parts you have a complete cooling system for $60. Again thats about as expensive as a good air cooler but would give me a viable option to go water cooling or others like my self to do it for a good performance.

 

[MattODoom]

So I guess that if you needed to remove this thing youd have to drain the entire loop and then kind of tilt the computer?

 

[Ben333]

JayDee, WTF? lol

 

[Anubis_386]

Looks awesome .. But wouldn't performance be pretty bad because of there not being much surface area?

 

[jaydee116]

I don't see how this is a failed experiment. You have just proven that you can easily produce and sell a very cheap "water block" or heat exchanger for a water cooling system for low income folks like myself. I would be excited to see something like this on the mass market. I could easily see prices sub $10 for something like this.

I mean even with the costs of a this with a total water cooling system your looking at more money or equivalent than a very good air cooler. But to sell the part stand alone it gives low income enthusiast a way to play with water cooling and scrounge the parts on my own. If you sold this for say $10, got a good old fashioned heater core for ~$20, pump used, scrounged, or from a junkyard, etc.. for roughly ~$20 and $10 for home depot parts you have a complete cooling system for $60. Again thats about as expensive as a good air cooler but would give me a viable option to go water cooling or others like my self to do it for a good performance.

The problem is liability. The seal is good but only in optimal conditions. If some average joe pulled the water hose a little to tight the seal would brake and water everywhere.

Another problem, as pointed out, is removing the thing. I had the luxury of having the motherboard not mounted to the case and flipped it upside down when I unmounted it. even after rocking the mobo before removing there was still a fair amount of water in there.

On top of that a $48 Thermalright SI 120 with a $10 Nexus silent fan will perform better than this does and be quieter.

Also be aware it is highly unlikely you are going to get 3.6GPM through this thing with a regular pump. I am using a Gen-X PCX40 1200GPH pump to get that much flow rate. I would not recommend at all that you go with much less. This thing must have high flow rates or it will suffer miserably. This is why I didn't bother testing at lower flow rates.

A rough guess is a regular DD D5 would get about 1.5GPM through this and temps would be about the same as a stock AMD air cooler. Completely pointless.