Watercooling in the new scientist
 

http://www.newscientist.com/news/news.jsp?id=ns99994086

Thought you guys might be interested.

It's ice to see fellow NS raeder :)

I loved the last bit by chap from QuietPC... embarassing to say the least...

Electro-osmosis - my first though was 'Wow, sounds like a battery effect going on there' or current passing through this glass disc is isolated from coolant somehow (doesn't seem to be since migration of ions is what makes batteries going).
Nice find dude :)

yeah I don't really understand how the pump thing works. Thought it maybe to do with the forces (I think it is called the weak force or something, obviously too long since I did any physics) from the ions attracting the molecules in the water bit like gravity from the sun and moon for the tides.

They clain that 200 milliliters/m (0.2 L/m) would be enough to cool a processor, emitting 120W plus, with 500W hotspots. That's rather bold! That flow would have to be right on top of the die, in fact, that flow would have to go through the processor core, to be that effective! We've all seen direct die experiments, and they require large flow rates: even with inpingement, 1.0 gpm (60 gph, ~4 L/m) doesn't even cut it.

I just don't see it. :shrug:

They may not be expecting that 0.2LPM to cool that kind of heat output. I think they meant that the technology can be ramped up to a much higher flow rate. The thing that bugs me: my pump is already pretty much slient. The advantage here is the lack of moving parts, and that would DEFINITLY be a good thing for us. No breakdowns resulting in meltdowns.

Even with no moving parts, a pump (or any other devices) can still fail... The advantage of having no moving parts really is noise reduction/elimination.

No moving parts does not mean no breakdowns.

They've been talking this up for a while now; I'll believe it when I see independent tests.

I'll give my point of view
the statement on the heat output that can be transferred is more of a marketting point than practical
I mean, it may be able to do it with the properly engineered waterblock, and radiator setup, but to cool that amount of heat (effectively) that particular flow rate is not going to do
it certainly doesn't mean laptop cpu's are going to radiate that amount of heat any time soon though- we still have the problem of inefficient battery technology
the statement made at the end of the article sums it up though- unless the industry 'needs' to, its not going to put water into any of its laptops for purposes other than enthusiastic etc, ones
if the time does come that laptops require watercooling, flourinet or a similar dielectric fluid will be used, probably in very small amounts per unit

200mL/min is more than enough to "handle" a processor. That is, of course, to keep it in operating temperatures... They would need a VERY good waterblock design, and in the end, it will cost WAY too much, for WAY too long, unless someone like Intel embrasses it. Can someone explain about these 500W hotspots, if the total energy input is only 100?

They are missing some units when mentioning 500W hotspots. 500W/sq-cm? 500W/hectare? Damages their credibility.

One thing I haven't seen mentioned, is the tendency of the electro-kinetic pumps to electrolyze water. Some designs I've seen include a platinum catalyst to reassemble the broken molecules, but this won't be 100%. What happens over the course of a year or two with highly reactive oxygen and hydrogen running around your cooling loop? How much water will need to be replaced? Will the end user be able to do this without introducing contaminents that clog the pores of the EK pump and the micro-channels of the waterblock?

Interesting stuff intellectually, not yet interested as a product.

If you're interested, I wrote an article on this exact topic at my home site, PCSynapse, a couple months ago. If you're interested in the subject, we're in the process of writing a sequel to my article, with some exclusive information you won't want to miss, so I think it's worth a look if you are interested in this system.

Read my article on Cooligy here.

Remember, Part II of that article is coming up soon, so If you're interested, come back in a few days and we should have it done.

The reason it doesn't need that much flow is because of it's use of microchannels to dissipate concentrated heat. The pump is just a small, silent, non moving addition, which hasn't reached it's full potential yet probably...

Think about how pumps have evolved from simple things like the archimedes screw to the super high flow racing pumps of today...

Think of the version there as the screw, and what it will become as the high flow racing pumps. It's still under development

We can't really wait thousands of years, I must say. That electroylzing thing sounds very not-good. The price of these things will be astronomical for years to come, and that is the reason that water blocks don't use this type of microchannels.

Au Contrare Angry Alpaca...

If you've read cooligy's press briefs, they clearly show that microchannels can be accomplished using regular techniques. You're thinking very linear, you have to explore your thoughts. Think of waterblocks today. They're all made with CNC machines basically. I'm sure with some effort, molten copper could be poured into molds, and blocks would be made by the thousands.

They'll probably appear in OEM systems first anyway, so it will have it's first test there.
We can't really say anything until it's released.

Quite true, take Danger Den as an example
the RBX is being sold quite cheap in comparison to the competition for a reason, and that has alot to do with economies of scale

Groth, more to do with the waterblock than anything else I suppose

Must say I'm sceptical of someone who states:
While the number of transistors doubles, the heat does not - as processors get thinner they also get cooler.

Also...
Is quite dubious. That's about what current chips generate, how many of you would run a 0.2l/m pump? :p

And finally...
That's a good finish because you're looking at a maya render of the setup not a real product. :p

It's been done. However, they supposedly aren't the quality we're looking for, and that's why we use machining. It's a lot better in quality, but costs more. These guys NEED quality, because even a minor mistake will make it not work altogether, due to being on the edge of acceptable cooling, rather than leaving a large margin, like we try to.

Interesting all this mucro channel thing.
I do not quite catch on this micro channels thou... do the y mean die's silicone to be microchanneled?
I think they should stay with this electroosmosis technology and leave heatsinks design to somebody else (look for foamed graphite heatsink technology and radioators based on the same material).
Why use very poor heat conducting material tofor water block and grove intricate microchannels in it (size of water blocks makes price difference between copper and silicone immaterial, not to mention massive embodied energy of any silicone products) if you can do it (microchannels) in one of the best known/cheapest heat conductors (copper)?
Microosmosis would work if used married with tunneling acceleration technique (magnetic particles accelerator, or however it is called-can never remember).
As AngryAlpace rightly observed price of this technology compared to puny performance imakes id still very much R&D project, but with lots of prospects.

I wrote about foamed graphite material, has anyone ever tried to produce foamed copper (open cell structure)? I seem to remeber sth about cigarette filter making technology using high voltage electron guns to produce micro channels, anyone?
It reminded me of recent experiment with slowing down of light, extremely inspiring:)

200ml/min is plenty to cool the CPU. Do the math, specific heat of water is 4.186j/ml, many times what it would need to be at 200ml/min. You would need specialized heatsinks of course, but no one expects to fit a WW in a laptop anyway.

You know, what would I give for a single solid piece of performance information that is not stacked to the gills with marketing hyperbolae?

What is the C/W of the block? I can roughly estimate it given what we've been told so far, and at 200ml/min, made of silicon (~150W/m-K), ~1.5mm thick base-plate, and even being generous and assuming a value for convection (h) of 250000W/m^2K, I arrive at an estimated C/W of around 0.25 for a 100mm^2 die, and that's not including the thermal goop/interface layer. Raising h to ∞ only improves that C/W by ~0.01. The use of silicon and the low flow rates seem to be the dominant sources of thermal resistance to me. If the silicon bp is really 0.75mm thick, then this knocks about ~0.06C/W off that figure.

I'd like to see how close to the mark my estimates are. I think it's quite telling that the marketing literature is full of hyperbolaic predictions, but quite short on actual hard information, which researchers of the calibre that the company is based on would obviously actually have and know, but they're not sharing it. Why?

You know, if I had something that was truly ahead of the game and wanted to push a hard sale, I'd have hard performance data right on the tails of the marketing material to back it up. If the block had a performance leading (BillA-style) T/W of 0.05, then why wouldn't they rush to make that public?

Excuse me for being cynical. Have read about twenty too many "Best....Performance....Ever" marketing material in my time.

You're under the assumption that the system is running water...

Chances are, it will be using more of an alchohol.
Secondly, the reason why huge amounts of flow aren't neccessary is because of the microchannels. The whole point of the block is to remove the densely packed heat on the core like modern heatsinks cant. As opposed to modern heatsinks, the heat travels only about a few millimeters to get dissipated. The physics of microchannel cooling are very different.

Secondly, the system has yet to be finalized, but you can't really draw a conclusion from the included information just yet. Think about it Cathar. You probably went through a few designs for your Cascade block before you made a final release version, they're probably doing the same. It's not going to make it to the enthusiast level until it's been tested on OEM high-performance machines first. I honestly think the next logical step would be to release a case with the cooligy system pre-installed, much like a Koolance watercooling case. This is uncertain though.

Alcohol is worse than water in almost every conceivable way for liquid cooling.

Most of us are quite familiar with micro-channels. True, the system, for its size and it's ability to locate the liquid-air heat exchanger in different locations/shapes makes it more convenient than any HSF and higher-performance than most any of them to boot, but the physical properties of the materials used don't indicate a dramatic improvement, and certainly nothing better than even moderate performing waterblocks of today.

Speaking of heat only needing to travel a few millimeters, this is no different to a number of waterblocks today. The bulk of the heat, in the Cascade design, travels less than 1mm before reaching a convective surface.

I personally believe it's aimed at the OEM level because that's where its price/performance will force it to live.

What about the heat in the water? What gets done with that? What kind of radiator will they use? Won't THAT be bulky?

why bulky? I mean for a prospective laptop system it would have to be small
I'm thinking a flattened heatercore design (meaning low thickness)
or something like the senfu, a large heatsink with channels flowing through the base- probably mounted on the back
all speculation

Well, if we've got a crap radiator, and no flow, and apparently a crap block, then we're getting into problems. Major problems.

"Goodson's experiments have produced a flow rate of 200 millilitres per minute. Keane says this would be enough to cool chips that radiate 120 watts of heat per square centimetre, with hotspots of up to 500 watts."

That really doesn't sound right...200 mL a min? As in, 0.200 L a minute? As in, 12 L an HOUR? Thats...NOTHING!

I know you're all talking about it, but that really doesn't seem like it'd cool much...

And cather, he said AN alcohol. There's more than one kind of alcohol.

Yes, there's a bunch of different alcohols, but they all suck.

I'm with stupid!

If they do it perfectly, 200mL/minute can handle a LOT of heat. However, the block isn't perfect, as Cathar points out, and I doubt they can do much with the space they have radiator-wise.

The point your missing is that the radiator won't neccessarily look like the ones we have today. Looking at the maya renderings, it looks more like an Aerocool heatsink that will be mounted in a remote location. It's not going to have the same cooling properties that regular heatercores/automotive radiators have. It'll be designed differently.

Cathar, as you have much more experience with cooling, I'll agree with you.

It should do fine as an OEM heatsink, and I doubt it's been perfected yet. I'm sure they'll figure out more efficient ways of conducting heat before it's final release to the OEM market.

Speaking of microchannels Cathar, how's that microchannel block coming along (Hydra I believe)?

I've pretty much given up on micro-channels. The smallest channel block I had made up was 0.5mm. It performed well enough but really only roughly matched the 1mm channelled White Water as the impingement nozzle becomes largely ineffectual as the channels are reduced, and it's better/easier to just do the standard one-side-in to one-side-out style of channels once you get below about 0.8mm channel width.

The issue with smaller channels (truly micro-channels) that I could see was that as channel size is reduced further, meaning much below 0.5mm, the pressure drop required to push the water through those channels starts to climb fairly steeply, meaning that the pumps that hobbyists use today start to struggle to push even moderate amounts of flow.

My predictions were that somewhere around 0.2mm was the practical channel width limit for the pumps hobbyists typically use. It's possible to make 0.2mm channelled blocks in copper using photochemical resist etching. Given that we need to cool about a 20x20mm area to cover all types and orientations of CPUs, a 0.2mm wide set of channels with appropriately sized channel walls would present a total block restriction that would make the block barely able to achieve 1LPM flow rates even for higher-end pressure hobbyist pumps (MCP600, Mag3, Laing D4).

Now 1LPM, with water, is what I personally estimate the balance point is in terms of the point where the thermal capacity of the liquid flow through a block starts to become an increasingly dominant source of total thermal resistance within the block's operation regardless of how high h, being the rate of thermal convection, is raised.

The other issue with making the channels smaller and then raising the pressure is pump heat. This is something that the EK pumps that Cooligy are looking at are an absolute necessity to solve. I do see a problem here though, Cooligy keep quoting 1W pump power consumption, but at some of the flow rates and estimated pressure drops that are being quoted, I see that 1W must be a lower end figure, and that to achieve some of what they're predicting to happen, pump power draw (according to the physics) must be approaching 10W or more.

Cooligy's graph:

http://www.cooligy.com/images/microchannel2.gif

Is true, but only for a fixed flow rate. Once we start factoring in pressure drops, and perhaps more correctly plotting a 3-D surface of pressure-drop vs channel-width vs h, we would see a surface that was hyperbolaic for h as well as for pressure-drop. Plot a flat realistically achievable pressure-drop plane through the surface and the picture doesn't look so attractive any more.

Look, I'm not out to discredit Cooligy at all, but instead to critically analyse what they have presented so far. It does look like a decent form of waterblock technology, but I absolutely will not swallow the marketing material hook, line and sinker without first applying what I understand, have experimented, and can see, to the material as it is presented.

The only way I can see this technology really working well is by 1) raising flow rates, 2) setting it up as a heat tube that incorparates the micropump, which means the entire system must be sealed (read oem), 3) incorparating the channels into the design of the CPU (read expensive, oem, and years down the road). In any case whatever is used for materials would have to insure that absolutlely no contaminants are able to form in the system to avoid blocking the pump and channels. In short and if it can be made to work, I would not expect to see this sort of cooling for quite some time. I do find it interesting but it is definetly not anything that will be DIY.