Overclock.net banner

1 - 20 of 42 Posts

·
Registered
Joined
·
1,022 Posts
Discussion Starter #1
If any of you have already removed the ihs, why not do direct die water cooling.

All you need to do is drill a rectangle about 20% larger than the die though a water block.

Than place a large o ring over the die and clamp the water block down 100% evenly on each screw using a torque pattern.

Lots of people did this with the old P4s but i haven't seen it done in a while

Iam sure it would pay off for ivy bridge
 

·
Registered
Joined
·
2,860 Posts
This interests me a lot. Wouldn't the die get damaged from direct water contact?
 

·
Registered
Joined
·
1,064 Posts
Quote:
Originally Posted by PepeLapiu View Post

This interests me a lot. Wouldn't the die get damaged from direct water contact?
It might crack if you ran it dry to build heat and then got an influx of cold water, but I wouldn't see what could be damaged by this. It's just a slab of silicon.
 

·
Registered
Joined
·
1,745 Posts
That's a small amount of surface area to cool, would there even be that much benefit? The whole point of grooves, fins, etc. in the copper plate is to increase surface area as much as possible, much greater than the die itself. I can't see this being very practical, but then again - what do I know? ^.^
 

·
Registered
Joined
·
1,064 Posts
Yeah, I just though about it. You need the metals to spread the heat because water is not a good of a conductor of heat. No matter the speed you run the water through the thing, it just won't take away the heat. On a side note there is also no such thing as water moving too fast (a general gripe I have when people say that water needs to spend a certain amount of time in a radiator).
 

·
Registered
Joined
·
483 Posts
Quote:
Originally Posted by ramicio View Post

Yeah, I just though about it. You need the metals to spread the heat because water is not a good of a conductor of heat. No matter the speed you run the water through the thing, it just won't take away the heat. On a side note there is also no such thing as water moving too fast (a general gripe I have when people say that water needs to spend a certain amount of time in a radiator).
so is it better to run the water speed faster?
 

·
Registered
Joined
·
1,286 Posts
It's counterproductive to direct cool the die. You want to use copper or another metal that will immediately pull the heat away faster than water will, and then buy more surface area for the water to swoop in. When Intel builds a heatsink, it's a giant copper block with 2-3 loops going through it.
 

·
Registered
Joined
·
1,745 Posts
Yeah, I think the direct-die water cooling was from the early days of water cooling. Now, our "direct die" solutions would involve sitting the block directly on the die (rather than the heat-spreader), as opposed to having the water itself contact the die. That way you greatly increase surface area for heat dissipation - the only thing this would remove from the equation is the IHS and one layer of TIM.

However, in that situation, you are still reducing surface area , and most blocks are designed for a larger heat spreader instead of a die so would likely be less efficient in that case anyway.

Long story short, it doesn't make much sense to do.
 

·
Registered
Joined
·
1,064 Posts
Generally not. You just need to make sure that the volume of water in the water block is moving fast enough to remove the heat and that it spends enough time in the radiator to shed the heat. There is a point where that happens. Above that, more flow does literally nothing but pump heat into the coolant via the pump. For super efficiency a person would measure the block temperature, the coolant temperature, the radiator temperature, and the ambient temperature and do measurements at what pump speed keeps things the coolest over all of these various temperature ranges and pump speeds and make a controller to keep everything operating at the minimum speed.
 

·
Registered
Joined
·
2,860 Posts
I LOL every time I hear people saying the water needs to spend more/less time in the rad.

Let's say you have a loop with 3 gallons of water. And let's say that your radiator can contain 1 gallon of water.

Your water is going to spend 33% of the time in the radiator, no matter what your fllow is. If you pump 6 gpm, the water will spend 10 seconds in your rad, but it will go into your rad twice every minute. So it will spend 20 seconds of every minute inside your rad.

If you only have a flow of 3 gpm, your water will only go into your rad once every minute, but it will stay there 20 seconds every time.

Flow will not increase/decrease the amount of time your water spends in the rad. If you want your water to spend more time in your rad, get a bigger rad that contains more water.

And I am sorry, but this direct die/water cooling can not work as well as with a block. Imagine it you will blowing air directly on the die. Air can not absord heat as well as water. And water can not absorb heat as well as well as copper. This is why we need the copper to spread the heat for the water. And this is also why we need rads to spread the heat for the air.

I will however try to delid my CPU and put the block directly on the die. That should increase cooling efficiency.
 

·
Registered
Joined
·
1,745 Posts
Quote:
Originally Posted by PepeLapiu View Post

I will however try to delid my CPU and put the block directly on the die. That should increase cooling efficiency.
Keep in mind that blocks are designed for an IHS-sized contact area, not a die-sized one... it may or may not yield better temps.
 

·
Registered
Joined
·
2,860 Posts
Quote:
Originally Posted by Powermonkey500 View Post

Keep in mind that blocks are designed for an IHS-sized contact area, not a die-sized one... it may or may not yield better temps.
The die makes a 'die size' contact with the IHS, and the IHS spreads it for the copper block. I see no benefit of a copper IHS spreading the heat for a block that is also made of copper.

Theorically, I see no reasons why it wouldn't work. ......nutin' like finding out for myself I guess.
 

·
Registered
Joined
·
1,286 Posts
Quote:
Originally Posted by PepeLapiu View Post

Quote:
Originally Posted by Powermonkey500 View Post

Keep in mind that blocks are designed for an IHS-sized contact area, not a die-sized one... it may or may not yield better temps.
The die makes a 'die size' contact with the IHS, and the IHS spreads it for the copper block. I see no benefit of a copper IHS spreading the heat for a block that is also made of copper.

Theorically, I see no reasons why it wouldn't work. ......nutin' like finding out for myself I guess.
Heat transfer is a function of surface area, thickness, and thermal conductivity of the material used. By direct cooling the die by soaking it in water, you decrease the thickness, but you also lower the surface area and thermal conductivity (water is a relatively poor thermal conductor compared to metals). Improvement in 1 category but worsening in 2 others means a loss in cooling performance.
 

·
Registered
Joined
·
1,745 Posts
Quote:
Originally Posted by dr/owned View Post

Heat transfer is a function of surface area, thickness, and thermal conductivity of the material used. By direct cooling the die by soaking it in water, you decrease the thickness, but you also lower the surface area and thermal conductivity (water is a relatively poor thermal conductor compared to metals). Improvement in 1 category but worsening in 2 others means a loss in cooling performance.
Yes but we are now talking about regular setup vs block directly on die, instead of direct-die cooling (water on die).

I would think it wouldn't make much difference because blocks are designed for IHS, not die - but I am not sure.
 

·
Registered
Joined
·
1,286 Posts
^^ Ah yeah missed that we changed scenarios. Cooling without the IHS and putting a waterblock directly on the die would be an improvement, if for no other reason than you eliminate 1 layer of thermal paste (between the die and the IHS).

However practically this won't always be better, because the IHS is applied consistently to make contact with the die whereas you trying to screw down a waterblock by hand is likely to be more imperfect.
 

·
Premium Member
Joined
·
3,639 Posts
As we take the water flow rate to infinity, we can hold the water temp at ambient and run the heat transfer from the surface area of the die into an infinite volume of water. Ignoring the delta T of the water and going entirely on SFC across die area gets us the theoretical maximum performance. Now, compare this to observed block performance to see if it's worth pursuing.

I think a direct vapor chamber would be better as long as voids can be avoided. This system would have a void coefficient not unlike Chernobyl. Anyone willing to drain half their loop and pull an almost complete vacuum on it?
 

·
Registered
Joined
·
1,745 Posts
1 - 20 of 42 Posts
Top