Infrabasse

Waterblocks in general seem to be hitting a wall with the i7, not being able to cope very well with the massive heatload they produce when overclocked/overvolted.
I imagine this is due to the innability we have in getting rid of that heat fast enough.
We see impossibly complex block design which hit the limits of our manufacturing abilities (tiny fragile channels, expensive and dificult to produce, damaged on delivery, cloged up in no time etc, etc ...)

Now we've been trying to lap them, all the way to the point of direct die/block contact, even direct die watercooling (although that's a fail).

Now I was wondering wether that would work better:
Heatpipes used to move the heat really quickly out of a hotplate (in contact with the CPU)
Cool the pipes with a larger area waterblock seperated from the actual CPU and which wouldn't need to be as small area optimised with jets etc like we see on some blocks ... simpler design, lower manufacturing costs ....
We could even imagine a dual stage block with water cooling both ends of the heatpipes (although block efficiency would need to be calibrated so that the heatpipes don't start working in the wrong direction)

-------------

What do you guys think of that? Anyone ever attempted such a thing?

__________________
Work in progress

Slider46

I doubt heatpipes would be more efficient than a solid copper block at transferring heat. A more efficient waterblock design would probably show better performance.

Or better yet, some SAFE way of having the cooling liquid in direct contact with the heat spreader on the chip itself - plausible, but probably not viable for affordable mass-production.

__________________

Infrabasse

hmmm ok
I thought heatpipes were so good at transferring heat they might be better than solid copper ...
My bad then, and there I was hoping to revolutionize the industry hehe

Why dont they use solid copper rods on HSF then if heatpipes aren't as good as copper?

__________________
Work in progress

Slider46

Heatpipes are better for transferring heat over a distance (like up through the fins of a heatsink) however, a solid copper block (I would think) would be much better at transferring heat from the CPU directly to the water flowing through the block. The chip is in direct contact with the block and the block is in direct contact with the cooling liquid.

__________________

Infrabasse

Yeah but with pipes you'd be able to move a large quantity of this heat away to a bigger simpler block, in the meantime you could also carry on watercooling the base in the same way we do now, except there wouldn't be as much heat in that spot anymore ... (and you'd have to do that working around the base of the pipes )

You get the idea?

Let's wait to hear other opinions on the subject ...

__________________
Work in progress

fstfrddy

As a block maker I agree that the wall has been hit. I have experimented with different configurations, pins, channel designs and have been unable to see any real improvement. I return to the two favorite designs, which is my G or S channel. No matter which block you buy today they will all be within a few degrees of each other. The variance comes from mounting techniques, TIM used and even the construction of the CPU (IHS). Sad but true, ambient water can only go so far. If you want more than that the only option is chilling of some method. Unfortunately IMO the heatpipe/waterblock combo wouldn't help and would certainly translate to higher production costs.

__________________
Heatware Rating
4 GHz Overclock Club

Slider46

You're assuming that heatpipes are going to be more thermally conductive than a copper-to-water interface and you're also not factoring in efficency when it comes to how many times the heat has to be transferred. (i.e. the fewer number of times heat has to be transferred, the simpler and more efficient a system becomes)

Your idea would go something like this:

Chip > 'Hotplate' > Heatpipes > Copper Waterblock > Cooling Liquid

Rather than:

Chip > Copper Waterblock > Cooling Liquid

The heat would have to travel further, through more interfaces (Chip to hotplate, hotplate to heatpipes, heatpipes to waterblock, etc) and to top it off, through heatpipes which i'm pretty sure would be less thermally conductive than a simple copper-to-water interface (when moving heat comparible distances)

__________________

Infrabasse

I still don't understand why it wouldn't work better ...
The wall we're hitting is the innability to displace the heat from the surface of the CPU fast enough, right?
Current blocks efficiency I believe are dictated by the thickness of the base (which is required to expand the water/copper area) and the efficiency of the channel design.

now if we used direct IHS contact heatpipes we'd basically only be limited by the heatpipes contact area and wall thickness. And we could carry away more heat than we currently do using a larger block on the other end of the pipes.

edit: I posted this before reading the above post. I'm still taking that one in

__________________
Work in progress

Clockadile Dundee

What would happen if you just put a pump with a tremendous amount of head pressure on a HK3.0? Like an rd20 or rd30? Would the added turbulence and pressure cause a drop in temps?

Would that just asplode my car radiator?

OP : I have been wondering if soldering heatsink elements to the IHS would increase the transfer. Of course, this would be highly impractical. But, if you could somehow have the unit be modular, you could have it so that part of the assembly could be removed.

On the other hand, I think heatpipes are moving in the wrong direction.

(rimshot)

Infrabasse

What kind of temps are people hitting when using madly ventilated direct contact heatpipe HSFs?
What I'm picturing would basically be that but replace the fins&fans with a waterblock ...
And then maybe mildly watercool the opposite side of the pipes at the CPU end.

If that could get us 5c lower I'm sure their'll be people to fork out the extra cash.
Oh well if that's not gonna work, you can't blame me for trying

__________________
Work in progress