There isn't much difference really. An o-ring (for example) seperates two areas with different pressures, regardless of the actual pressures. It's the difference in pressure that determines how good a seal has to be.

There are no specifications listed in the article, but a complete vacuum is 14.7 psi. A water cooled loop has to deal with, worst case, 5 psi.

The big difference is that one can afford to loose a bit of pressure in water cooling (i.e. a negative pressure leak that injects air in the loop, or coolant evaporation through the tubing), but here one cannot afford to loose any of the vacuum: if the vacuum fails, what you have left is a really bad cooling solution. The cooling unit is then (essentially) unserviceable, aka junk.

AC systems use solid tubing, because they have to hold upwards of 300 psi, but they still break down and leak, on occasion. They're generally considered "high-maintenance" solutions, requiring yearly pressure checks.

All in all, it's the connection failures that's preventing water cooling from going mainstream (that, and cost). In the electronic cooling industry, Stainless Steel tubing is often perceived as being needed. :shrug:

Heat pipes are a bit better, as Cathar already pointed out; their construction is simple, and they remain as a fixed part of a larger assembly (i.e. no mechanical stress to bend and break it).

IBM recently annouces a rackmount cooling solution, but all it is, is a large radiator running chilled water, to cool the exhaust from the rack (to keep the data center temperature cool). IBM's ultimate plan is to incorporate the same setup within each unit placed within a rack, but it'll still be a "radiator after the exhaust" kind of solution; there are no plans to run water within a computer. (it sounds suspiciously like a plumbing exercise to me!)

BTW, did ya'll know that water cooling has been around since the sixties? (yes, 60's!).

1. Whilst I like innovation and experimentation, the thing that kills this idea for me is the fans. I run a PC, and I have it set to beep when it gets too hot, and shutdown when the temperature "runs away". So, given that, whats the diference between having a pump and having fans? Its not a passive setup, its not a silent setup, its just pumpless. If it was pumpless AND fanless, now that would be something. But its not.

2. I also don't like the radiator/condensor. To me, as per that pumpless-peltier rig, the radiator should be single pass, and have the "hot end" raised slightly to the "cold end".

3. Similarly, the waterblock is curious in its design. The inlet and outlet barbs are on the corners. I would have thought the optimum design would be for a vertical-motherboard orientation, with the outlet barb exiting parallel to the motherboard, or at least at 45° ala Swiftech MC6000's. The inlet would then be below the "contact" area.

I'd then have the "air chamber", where the vacuum is created, in a short tube above the height of the radiator inlet, so that the radiator is totally flooded.
But then the bubbles aren't going through the radiator to be cooled.

Hmmm....

Long Haired Girl: do you think it's possible to use a large passive radiator instead? I think it's possible the way I understand it. The ones that are marketed claims to be able to dissipate up to 125W, though i'm not sure at all what that really means. My question is, does the boiling water make any noise?

I do not think vacuums are such a big deal, the technology has been refined over the last hundred years - light bulbs, thermos bottles to name a couple. They have high reliability, Just need the right process

Metal hose, metal condenser poly carbonate cap does not look to be to tough of a challenge. Add a few ml of fluid and you are only talking a couple of pounds vac at rest and a couple pounds pressure at load, considering the small system volume.

A condenser is not a radiator and the boiler is not a block. This is a completely different animal. I would not be to quick to apply a water cooling template, I believe design considerations are much removed from water cooling. Plenty of thermosyphone literature out there

Re radiators, "You can't make butter with a toothpick". Ie, size matters.
However, the graph is awfully steep getting near zero air flow - every little bit of CFM helps, especially the first few. A truely passive rig would be an acheivement, and probably not portable.

I've not done any research on this, and so I agree that my design ideas above are probably newbie and illconsidered - hence my "Hmmmm".

Joe didn't just chose to remove the pump for noise, but also the high cost and failure rate of pumps.

Single/double pass doesn't strike me as too important for this. In fact, single pass sounds like it'd be worse because of the tubing required to get to the other end. I don't think the idea is that the water will go in one end and exit the other, but rather some water will enter and some will exit from barb A, and some water will enter and some will exit from barb B.

I'd think that both the inlet and the exit should be facing up at 45 degrees from the MB, in the same direction (up).


What I want to know is if having two hoses is necessary. It seems to me that having one, larger hose would do the job better, take up less space, cost less, and reduce the failure rate.

Sounds like a Vapochill Micro.

http://www.burnoutpc.com/modules/wfs...?articleid=213

I couldn't find any detailed reviews where it is put up against other high-end air coolers.

So... how is this setup not a heatpipe?
There's state change at the hot end and state change at the cold end (vacuum simply being a way to lower coolant state change temp - one could just as easily use something like butane and a slight positive pressure to, again, get coolant state change temp into a useful range).
Anyway, I'd always thought that this was the way heatpipes worked - many of them now have a wick to move liquid coolant against gravity, but the original ones I'd run into (late '60s) did not and were indeed position sensitive.
Is it that there is an "outbound" and "return" line? AFAICT the only thing that makes the return side a return is that is exits from the lower side of the radiator (which maybe we should be calling a condenser?). And, again, I have this foggy memory that early heatpippes, at least, also had outbound and return lines.

Feel free to point out what I've missed, the error of my ways, etc. - so long as you've got a decent reason why this thing is not kind of a heatpipe.

Thanks!

It works like a heatpipe but the means is different - way more surface area at both ends, and no solder joint between the heatpipe and the heatsink, or between the pipe and the metal chunk on the CPu.

Maybe we'll get a clarification.

I don't see any difference between a heatpipe and how AngryAlpaca thinks this works. The amount of SA at the condensor, and number of thermal interfaces at the condensor, are merely implementation compromises of commercial heatpipe waterblocks. The "home made" heatpipe (already linked within this thread) did away with the latter. Anyway, if the sole means of heat transfer is the evap/condense (bubbles), then its a heat pipe. Given two parallel paths, and returning of liquid via the same path, I can't see how any real coolant "flow" would be created, and the cooling of any flow would be immaterial compared to the cooling by the bubbles. So its a heat pipe.
Given this, then one large tube would be probably better, and yes AngryAlpaca's comments on the waterblock apply.

But I don't think its a heat pipe.

I think its a "syphon" as described. I think , as per a posting above, the bubbles contribute to fluid flow in that they go up one tube only, and hence there is a real "flow" eventually established. Not exactly a high flow rate, but flow none-the-less. This flow allows "normal watercooling" to contribute significantly to the overall efficiency of the rig.

However, UNDERBYTE's comments on condensor/evaporator design still apply, so I'll shut up now.

Well, in the article, there is a link to an article on thermosyphons, where the following is said:

It also says this:
So, what it is describing is a heatpipe - there is no "liquid flow", but a cycle.

Home Depot sells those same gas line tubes with flared and (IIRC) brazed endings, so theres no reason you couldn't make this thing tight enough for decades of operation. People use them in ghetto refrigeration systems already, and those often have 200+ PSI in them. Ditch the plastic for copper and you're good to go.

Edit: And put a $0.75 shrader valve on there. That way you can pump the pressure as high or low as you like.

shrader valve - interesting

hard plastics may be OK, Bic lighters do pretty well with butane.

is that a shrader fill valve on the bottom of a bic?

I think the correct spelling is "schrader valve". I believe they are used on car tires and inner-tubes and what not to fill air (and other stuff such as AC units). The part at the tip of the valve stem.

I don't think a schrader will hold vacuum - at least not to any greater (negative) pressure than the spring holding the valve closed - when holding positive pressure the internal pressure also holds the valve closed, so seal clamping force goes up with the pressure it's holding.
If you were thinking of something that required positive pressure to move state-change temp (like butane, maybe) then they should work fine.
Also - at least some schrader valves stand up to butane and propane just fine - they're used as part of the gas regulation systems in a number of outdoor gas grills (usually with a diaphragm pushing the schrader open - but I digress).
Presta valves might work for vacuum as they screw shut rather than being pushed shut - but I don't think presta valves with threaded ends are very common (I've never seen one).

If those are Schrader Valves, they're not the kind I'm thinking of:

http://www.rparts.com/Catalog/Major_...s/schrader.jpg

I got a 5 pack for a project for about $4 locally.

Interesting point. I've not tested the negative pressure limit, however I believe its fairly low. At least I've never seen or heard of a problem on systems running at below ambient pressure. Most likely you would not need to go that low for common coolants.

The worst case is probably water, where you'd be around 40 mmHg. Typically refrigeration systems are vacuumed at much lower pressure then this, often for hours and I've not heard of valves leaking. But I suppose its possible they leak slow enough that the vacuum pump overcomes the leak. At any rate, if I were to design something like this I'd probably use a coolant that worked at a slightly positive pressure anyway. R134a might be a good choice because its so cheap, nonflammable and comes in pure form. Pressure would be a bit high though, so you'd have to braze or at least use flare fittings on everything.

What kind of valves are used?

I was just looking at boiling points - too bad pentane's flammable (and a suspected neurotoxicant :() a boiling point of ~36C at 1ATM looks pretty good...

What about SP34E boiling point of -15 vs R134a at -26 - which should mean lower pressures needed. I know they make a big deal about R12 computability (mostly ester oil vs mineral oil, right?) but for a heatpipe-like-thing with a target temp around, say 22 to 26 then a boiling point closer to that target temp is a good thing as there'd be need for less pressure (plus or minus). Or have I missed something important? Other than playing with heat-driven refrigeration in the late '70's (arab oil embargo, energy panic) I have no experience with this stuff.

They're simply called "Schrader Valves" which in the context of HVAC means the ones I linked above. You can buy them for less then a buck a piece at any supplier or from Rparts.com

That site is highly confusing and contridicts itself repeatidly. They say its not a blend like R134a and then they say its mostly R134a. Not to mention the boiling point they claim isn't the same as the one in the data sheet :)

I'm going to assume the data sheet is correct, in which case the boiling point is virtually identical to R134a (within 1 PSI). So you'd be looking at ~100 PSI under operating conditions, and maybe 200 PSI worst case (dead fan). Sounds like a lot, but its actually quite manageable with common materials. Realistically you could use a propane torch to put the thing together (though MAPP or oxyace would be better).

Mainly the reason I looked at R134A is you can get it in cans for about 10 bucks a pop, and its relatively easy to hook these into standard manifold fittings. No need to fit a square peg in a round hole. I'm not sure if its actually a good choice from a performance perspetive, and the pressure is needlessly high. If you find something cheap with a boiling point closer to 20C, then I think its worth considering.

Also, while searching, I found this:

http://www.benchtest.com/heat_pipe1.html

Seems like we're a few years late to the game :) Though he made the condesner himself which I think really hurts his performance. Should have picked up a real one.

Twas reading that last night too.
What's interesting is that in the heat_pipe_2 article, the pipe works best when about 1/3rd full of liquid. That suprised me, and opened up a world of thinking.....

As in "what would happen to the thermosyphon device efficiency if about 2/3 of the coolant were removed?" Something less obvious that I missed...?

Read the papers on thermosyphons I would be surprised if there was more than 50 -100ml, what ever the coolant is

Yeah these things probably run on almost empty. You just need enough so that theres liquid in contact with the CPU block. Maybe a little more to be safe.

I was actually thinking more in line with Cathar's lament you can't easily add a GPU block....

Havent read the whole thread but Im interested in the results of this.

To me this seems like the logical way heatpipes should be made.

I assume the coolant is water, at vaccume. While there are good resons not to use them, it may be better to consider other coolants lower boiling points at normal pressure.

I have always thought the block for a heatpipe cooler should acctually be part of the heatpipe - and not just a bit of copper with a hole and the heatpipe inserted into it - which must hurt performance a lot.

There was a thread about this on bit-tech, but it didnt go anywhere as no suitable coolant exists at normal pressure for DIY research.


Block deigns for evaporative cooling would also be very different to normal watercooling, so maybe there is a lot of space there for improvement - as you wouldnt judge watercooling based on a maze1 cpu block performance would you?


Unfortunatly this would not be useable on GPUs? where as normal heatpipes and watercooling can.

Wouldn't adding a copper wick improve performance of thermosyphons? Anyone know?

The performance seems very good even as it is. Would work great with a peltier to freeze things up a bit, for extreme overclocking.

How about a peltier on the condenser instead? increase the rate of condensation ?

Idle speculation here as I do not know just some thoughts


they report the large die C/W as being .12 C/W overall

assume the the syphon condenser is comparable to an equivalent water radiator, maybe .08C/W?

Ok, that leaves ..04 C/W Boiler and TIM.

Take away .02 C/W on the TIM That leaves .02 for the boiler? Not bad

double the boiler performance to .01 C/W takes you to an overall of .11

double the condenser performance .04 C/W and you are at .08C/W overall

flip the ratios if you want, as with water cooling you are on a pretty rareified edge.

Big changes sometime make little differences, little changes can make big differences

There have been updates to the site regarding this, including an independent test from an unnamed source.
http://www.overclockers.com/articles1249/

I don't suppose BillA's fingerprints are on this?

Hmmm.

Paint me black and white if you wish, but is anyone else perturbed by the transition from OC.com being an industry independent test source to now being a site that is conducting competitive performance analysis against a product that they are now going to have a vested commercial interest in seeing succeed?

I am not saying that results are biased or inaccurate or anything of that nature, but in declaring a now very public conflict of interest with respect to what they're working on doesn't that then now remove the very independent nature upon which OC.com reviews were once thought to exist under?

What I mean is that OC.com has now crossed the line from an industry independent review source, to effectively being a nascent manufacturer hosting their own internally conducted competitive analysis reviews. This now means that the very independence with which OC.com used to operate under is now conflicted. This doesn't extend to water-cooling, but includes all cooling reviews of all products that they have conducted. This thermosyphon product has by their admission been in development for over two years - where does that place the independence of any review conducted by OC.com in the last two years?

I guess I'm just saddened to see the loss of independence of OC.com, especially since it existed for so long.

Good Point, Have to see how they handle it in the long run when they actually have a product. If he sends it out for independent review and it is validated then drops it from the site and no further product pumping occurs OC will have done a service to market by developing an alternative. If not, and the product is focused to making sales on the site you are right. Right now he seems just to be introducing the technology;

Well it gets a little seedier than that IMO, the more I think about it.

Under the guise of an independent review site, over the last two years JoeC has been receiving free competing manufacturer equipment with which to conduct competitive analysis for a product that has been in development by himself. In fact, his testbed is largely the result of freely donated time, equipment and efforts by various people.

What nascent (upstart) manufacturer wouldn't love to freely receive competing products with which to assess one's new products against in order to find a market for it, and position it appropriately?

Some people may call it clever. I call it something else.