Waterblocks with heat pipes pins possible?

[Jag]

I've been thinking, would it be possible to manufacture a waterblock with heat pipes pins(inside the block or protruding from it) something in the style of Swiftech?
I mean, is it a viable possibility?
I know that heat pipes pins exist, and they are very similar in shape to those utlilized by Swiftech in their air cooled heatsinks.

[BillA]

think on it some more, and the obvious will become apparent
do you know what a heat pipe is ?

[8-Ball]

There appears to have been a lot of interest in heat pipes recently for what ever reason.

A while ago, I found a company by the name of thermacore who had developed a heatsink base using heat pipe technology.

I emailed them about the possibilities of using such a device for waterblocks, and was sent a wealth of information.

After getting out the scrap of paper and my calculator I came to a fairly quick conclusion.

There is a significant bottleneck.

The thermal resistance of the vaporisation stage alone was more than the TOTAL thermal resistance of most waterblocks.

Heat pipes are not a performance cooling technology unless you are wanting to move thermal energy over a relatively large distance.

8-ball

[Jag]

Thank you.
Well sometimes ignorance and stupidity come hand in hand.
I just asked, because (it's just my assertion, don't killl me for it!) it seems to me that even when you try hard, redefining and improving the waterblocks and so on you just can't get much higher, I mean is watercooling reaching a plateau, that from there on the benefits will be marginal, just like the F1 cars that compete for seconds? - Maybe that's the trill of it.

[killernoodle]

Well, you already have water running through a waterblock, so what would be the point of adding another medium that is essentially acting in the same fashion-- allowing a liquid to flow through a tube for cooling. Water will do it better than whatever they put in those things.

Also, water cooling is reaching a plateau as far as performance goes. I like to call that plateau "room temperature". If my dad would only let me turn the AC down to 10C I could break the room temp barrier and usher in a new era of simple water cooling

[8-Ball]

Originally Posted by Jag

I mean is watercooling reaching a plateau, that from there on the benefits will be marginal, just like the F1 cars that compete for seconds? - Maybe that's the trill of it.

That's exactly it.

At least with waterblocks anyway.

Areas that still have room for improvement.

radiator - smaller and more efficient
TIM joint - still accounts for a large portion of the total thermal resistance between the cpu and the ambient air.

8-ball

[Jag]

So, what can we expect in terms of cooling as the cpu's get warmer?

[BillA]

increased cooling cost

[8-Ball]

I think the question should be, what happens when the cpus start dissipating more thermal energy.

The answer is, they are gonna get warmer. Full stop, unless you move to active cooling.

8-ball

[Jag]

For example, if the TIM joint is very influenced by the its application and compression, why did some manufacturers moved to the OEM heatsink/waterblock mounting scheme (socket 478), instead of the mounting holes they used in their previous products?
Does it makes sense?

[Jabo]

Yes, it is cheaper

I cannot agree with a point of view that there is not much to improve on water blocks front.
What is available currently is based on two types of design. First and the most popular one is pin/fin type. The other one is impingement type design (Cathar's design and rip-offs).

With warmer CPUs blocks design will have to change to accomodate this and keep silicone at acceptable temps. Sub ambient cooling is a different matter and probably will never go mainstream.

I second 8-ball's point of view towards radiators getting smaller and mre efficient. Mind you, it is dead easy to and cheap to build a setup keeping your coolant at ambient temps or just above it.
TIM interface and generally mounting issues have some potential for improvement since even the best testers do it like 10 mounts and then extract av. values to get close to meaningful performnce values.

Coolants are dead end imo due to massive R&D costs involved.

My two pennies

[Boli]

Intel pulled a chip (I think the dual core thingy unsure) because it was too hot and they are trying to make it stay under 100W of heat, i.e. at the best a quiet cooler can manage and still keep it within acceptable long lasting temps.

Its not the CPU I'm worried about heatwise, manufacturers know they can't raise it much higher without problems its all the other parts that have my attention.

Remember when a PSU only used to have one fan and some didn't even have that, a southbridge now comes with a heatsink and will only need more and more active cooling as the PCI-X and SATA speeds get ever faster. That's another thing hard drives they can get REALLY hot.

I have a feeling watercooling will concentrate less and less on the CPU block as it is starting to become increasingly obvious that only insanely priced Silver or even Diamond can hope to get you a measurable temperature difference away from the rest of the pack.

When it becomes obvious that the bare silicon cannot adequetly conduct the heat away in a satisfactory manner, first the heatsink will tred - then the imortal fan before the almighty waterblock.

My vision of the future unless a new cooler technology is developed is a motherboard full of mircro waterblocks with capilary "fluid pipes" snaking across both sides - before entering into a self contained radiator/pump/res stored in a couple of the 5 3/4" bays.

[jlrii]

Originally Posted by Jag

I've been thinking, would it be possible to manufacture a waterblock with heat pipes pins(inside the block or protruding from it) something in the style of Swiftech?
I mean, is it a viable possibility?
I know that heat pipes pins exist, and they are very similar in shape to those utlilized by Swiftech in their air cooled heatsinks.

A typical water cooling has several interfaces to think about: air to the exterior rad surface, conduction through the material the rad is made of, water to the interior surface of the rad, water to the block, conduction through the material of that block, and lastly the TIM joint from the block to the CPU. Of all those junctions the TIM joint is the weakeast.

In the setup you propose poorer performance would probably result unless the heat tube were directly immersed in the water on the cold side and directly on the die on the hot side. In that case you might get close to water cooling with a block alone but not exceed it. The reason for this is, while the tube has excellent conduction, but many more interfaces are created.

The proposed system would include all of the above plus: a TIM joint where the heat tube is fastened to the water block, the conduction of the heat tube, possibly another TIM joint where the TIM joint is fastened to the CPU block, and the conduction through the material of that block.

Optimally the cold end of the heat tube could extend directly into the water block allowing it to have direct contact with the water, and the hot side would contact the die directly. But there are a couple of problems to overcome with that scenerio.

If the water block is to be of the impingement type- How is the heat pipe to accept the jet(s)? Keep in mind that if the pipe does not have the cold side above the hot side it must have some sort of wicking material inside to get the liquid back to the cold side. So the interior of a machined baseplate for the waterblock would probably have to be coated with a sintered metal to allow liquid return. Also the tubes are usually made of copper and are quite thin. We have seen how thin baseplates on waterblocks can distort under the clamping pressure required to create a good TIM joint. So attaching the tube directly to the core becomes a problem as it will distort when pressure is applied. Also note that most tubes use water under a vaccum which may further induce distortion under pressure.

Finally consider that the tube will probably be made of copper. Just because it is a heat tube does not alter the conductivity of the shell material. As in the Cascade, the baseplate is already close to, if not at, the minimum thickness to avoid distortion. If the tube must be that thick what have you gained? The conductivity of copper remains the same at the base plate and at the block on the CPU. Between those you have the added thermal resistence, however good, of the heat tube. Lots of work...lots of cost...the one advantage would be being able to relocate the water block in cases if severe hieght restrictions within the case.

Yup, I did think about this a fair amount when I first found out about heat tubes. I had thought about incorparating one inside a water block. But the same applies, water cooling is close to it's pinnacle with impingment blocks like the cascade. A large conduction area within the block is not required and adding more components beyond the the impingment surface only adds more thermal resistance.

[BillA]

Originally Posted by Jabo

. . . . .
The other one is impingement type design (Cathar's design and rip-offs).
. . . . . .

so Cathar 'invented' impingement ?
spoken like a fanboy noob

[pHaestus]

Bill:

Who did produce the first commercial wb with central inlet? First ones I remember were Danger Den Maze2 (maze1C was later actually) and the old Swiftech (name escapes me). Or was there another wb out before then even? Impingement as a technology is very old though for sure.

[BillA]

no idea really, the first wb I bought was a MCW462 (and I modded the bp);
possibly the Maze by DD

anyone ?

[pHaestus]

Was that one blue anodized? Swiftech had a central inlet wb before that one that was bare aluminum (or possibly clear anodized). I think that block and the original Maze2 were pretty close coming out.

I assume that you first used the term "impingement" to describe HOW these wbs worked though.

[BillA]

the first were bare aluminum

I coined the phrase "die area impingment"
Gabe used the term "jet impingment"
- both to describe the inlet 'action'

[Les]

Originally Posted by pHaestus

....
I assume that you first used the term "impingement" to describe HOW these wbs worked though.

My earliest recollections are Bill's post here :-

" jeeez, now that guy's started a new thread

offering free samples

and asking where there is a forum that is un-critically accepting

not here, babe

Ice Czar, wrt wbs:
We are probably looking at the same parameters, but I would describe them differently

3 "generic' wb designs
- labyrinth (MazeI)
- finned (OCWC home-made)
- die area impingement (MCW462)

despite the many pronouncements, it (to me) is still very much unresolved which type might be (able to be) called "superior"

but some few facts are known:
5 or 6 years ago Intel published a paper (which is on Google, lost the URL) comparing labyrinth and finned wbs for (large) server applications, and they concluded that finned were better due to the reduced pumping requirements. As I recall the cooling, per se, was not a determining factor.

This would at least suggest that the design parameters available for manipulation for those 2 will yield similar results. (And looking at your listed variables it can be reasonably supposed.) Note that in a systemic watercooling system evaluation the pumping requirements are hugely important, yet in looking at only the wb the pump head losses attributable to the wb are often not quantified.

Which leaves the dai (die area impingement) type, of which the POS waterbox that is the subject of this thread is another variant. (It should be noted that there are many hybrid designs also, MazeII etc.) Is there any reason to suspect that this type might be superior ?
Yes indeed.

Much difficulty is caused by sloppy terminology, and the use of the engineering term "turbulent" is a case in point. Turbulent flow is a mathematically described condition, yet is used by many as a synonym for a "mixing" type of flow. The enhanced heat transfer benefits that are applicable to the truly "turbulent" flow are substantially higher than the benefit from better mixing of the fluid.

Turbulent flow is a consequence of velocity, and is affected by the surface roughness. While turbulence may be induced by a change in direction, such will not (normally) change the flow regime. It is a fact that turbulent flow cannot be achieved in a labyrinth or finned wb (with any reasonably sized pump), but it can be in a dai type wb.

I've just started wb testing, but the results are startling.

be cool

IP: Logged

11-24-2001 07:09 AM "

[BillA]

2001 eh ?
learned a bit since then, but a tolerable post despite the errors

there should be procooling threads even before that one,
went to H after JoeK banned me here

[kronchev]

Originally Posted by Les

My earliest recollections are Bill's post here :-

" jeeez, now that guy's started a new thread

offering free samples

and asking where there is a forum that is un-critically accepting

not here, babe

Ice Czar, wrt wbs:
We are probably looking at the same parameters, but I would describe them differently

3 "generic' wb designs
- labyrinth (MazeI)
- finned (OCWC home-made)
- die area impingement (MCW462)

despite the many pronouncements, it (to me) is still very much unresolved which type might be (able to be) called "superior"

but some few facts are known:
5 or 6 years ago Intel published a paper (which is on Google, lost the URL) comparing labyrinth and finned wbs for (large) server applications, and they concluded that finned were better due to the reduced pumping requirements. As I recall the cooling, per se, was not a determining factor.

This would at least suggest that the design parameters available for manipulation for those 2 will yield similar results. (And looking at your listed variables it can be reasonably supposed.) Note that in a systemic watercooling system evaluation the pumping requirements are hugely important, yet in looking at only the wb the pump head losses attributable to the wb are often not quantified.

Which leaves the dai (die area impingement) type, of which the POS waterbox that is the subject of this thread is another variant. (It should be noted that there are many hybrid designs also, MazeII etc.) Is there any reason to suspect that this type might be superior ?
Yes indeed.

Much difficulty is caused by sloppy terminology, and the use of the engineering term "turbulent" is a case in point. Turbulent flow is a mathematically described condition, yet is used by many as a synonym for a "mixing" type of flow. The enhanced heat transfer benefits that are applicable to the truly "turbulent" flow are substantially higher than the benefit from better mixing of the fluid.

Turbulent flow is a consequence of velocity, and is affected by the surface roughness. While turbulence may be induced by a change in direction, such will not (normally) change the flow regime. It is a fact that turbulent flow cannot be achieved in a labyrinth or finned wb (with any reasonably sized pump), but it can be in a dai type wb.

I've just started wb testing, but the results are startling.

be cool

IP: Logged

11-24-2001 07:09 AM "

um how did you get a link from hardforum when theyve been down for 2 weeks?

[Les]

Originally Posted by kronchev

um how did you get a link from hardforum when theyve been down for 2 weeks?

Have a copy of the post - considered it significant at the time.
Only pasted copy because hardforum is down.

[Cathar]

There was also the Leufken Technologies DFI that was introduced mid-late 2001, which was a direct-die cooling block that pointed a jet directly at the CPU die, albeit from a 45° angle.

There were direct-die cooling efforts from before that time as well, although I am unaware of any direct links.

While not "Enclosed DAI" it does still classify as core impingement.

The Silverprop Cyclone 1 was also out and about in July 2001, and featured DAI.

Edit: Maze 2 was released prior to June 2001 though...

[BillA]

just to clarify, no claim is being made as to 'first'
only that, by whatever name, jet impingment has been around a while
(as in every single lathe)
as adapted by various individuals to CPU cooling

[SysCrusher]

Originally Posted by unregistered

2001 eh ?
learned a bit since then, but a tolerable post despite the errors

there should be procooling threads even before that one,
went to H after JoeK banned me here

Did you ever find that white paper again Bill? I seen that one at one time but it's no where to be found on Intel anymore. They also talked about impingement in a "cascade design". I think they took it off after the presshott situation.