Experimental Water Pump?

[bigben2k]

Well, I think you're partly right, in the miniaturization, but I think that it's going to be more of a wireless/central computing application.

Picture a slim notepad, with a tactile keyboard, that's less than a quarter inch thick. The processing is done at some central server, to which you connect by wireless.

But that server is still going to be an "old fashioned" box, and unless the die traces get really small, it's not going to be any smaller than what it is today. PCs today are pretty much the same size, if not bigger, than they were 25 years ago. When AT was replaced with ATX, we really didn't see any changes in size. uATX was nice though, maybe it'll pick up.

Right now, we've got 90 nm technology, and 60 and 40 is on the horizon, but it's getting real close to hitting some hard physics "wall".

If anything, watercooling could be integrated with the CPU. Kinda like my "CPU backside" cooler idea, but closer to the die. Then, that little pump might be just about ideal.

[Cova]

Lets not start assuming that that 3GPH is going to significantly change when you start putting some stuff in a loop with it. The osmosis effect through that glass may not be moving very fast, but I would bet it takes a LOT of force to slow it down any, whereas the mag-drive pumps we use don't even physically connect the impeller to the motor - you just have to overcome a weak magnetic field to stop flow completly. As someone else posted above, these pumps could probably power some serious micro-jet type blocks.

If they could get a high enough pressure difference and pump something that evaporates a little easier than water does (add a little methanol perhaps) it might even be possible to make easy/cheap phase-change coolers that have no moving parts

Though I agree with the other post farther up about running electricity through your coolant being a bad idea - especially since your WB will likely be electrically conductive and is attached to your core.

[jaydee]

Future CPU's I am refering to Ben:

http://www.sciam.com/article.cfm?art...CA809EC588EEDF

http://www.technologyreview.com/arti...istram1001.asp
http://www.theseus.com/
http://www.geek.com/procspec/newslet...oc10242001.htm

[bigben2k]

Well, eliminating the clock is one way to reduce the heatload, but it's a drop in the bucket, no? (I didn't read all of it).

[jaydee]

Originally posted by bigben2k
Well, eliminating the clock is one way to reduce the heatload, but it's a drop in the bucket, no? (I didn't read all of it).

It is a pretty big drop from what it makes out 15-30%. In fact Intel showed a CPU they said is scalable to 5gig that could be cooled with just a heat sink (in a different article that I cannot find). They will come up with a way to make them run cooler. After they have maxed out the current "style" of CPU's they will move onto something better that they are probably already working on. Also VIA (used to be Cyrix) has 1gig chips that can be used with a large HS and very slow quiet fan, in fact some use it passively. Sure performance isn't as good a AMD or Intel 1gig. But they do work good enough to get things done.

http://www.via.com.tw/en/viac3/c3E.jsp

Current CPU's are not nearly as effiecent as they can be. Once they become more efficent they will use less watts per die size. Just a matter of time. Just think how much tech has evolved in the last 8 years. Give it 18 and it will be totally different than today.

P.S.: I remember hearing 200mhz would never be achieved.

[joemac]

I don’t think they intend to drop the clock at all (I don’t see how they can). Instead they are probably going to use many different clocks to operate the many circuits inside the CPU. I think you can compare this to integrating a video/lan/sound card into the CPU each running at their own clock speed with its own clock, internal bus and calling it a breakthrough .

[Cathar]

My post copied from #Rotor's thread:

Well a few questions naturally.

1) What's the power draw to move the 200ml/min
2) Can it be scaled up? Say using a CD sized disk and pushing 1l/min
3) What about clogging issues? 1 micron tubes make for a pretty effective solids filter

3gpm is 3.3ml/sec, or about 14W/C for the water.

i.e. every 14W pushed into the water will cause it to heat up by 1C. So even a 140W CPU will only cause the water to rise by 10C as the water flows through the block, so 3gpm is still "enough" flow.

The waterblock design to take advantage of this would have a fairly thick base-plate. I would imagine around 5mm would do it. Using a 1mm wide channelled maze design would make for a pretty effective water block. Just have 1mm wide channels zig-zagging back and forth across the block. 1mm wide fins by 4-5mm high. This would keep coolant velocity in the realm of "decent", and due to the super-low flow rates, would be far less restrictive than one would think.

Sure, it'd get warm, but if the block is designed properly I think the performance would probably surprise quite a few people. It won't be a high-performance monster, but it could cool a very hot CPU far more effectively and with far less noise and bulk than a HSF arrangement.

For laptops you could probably use the back of the display as a large passive heatsink.

[TerraMex]

> For laptops you could probably use the back of the display as a
> large passive heatsink.

Makes sense.

Just as reference, if you check the Hitachi link i posted, they used , for the prototype , the lower metal plate for heat dissipation. Stainless steel. Probably because corrosion issues.

But then again, taking BigBen's idea, couldnt you also use the backplate connected to a series of heatpipes attached to a heatsink on the processor? Something similar to what Zalman is doing in the "all heatpipe case" .

And does anybody remember this ? If i recall correctly, it has a series of microchannels inside.

Edit: just found this

http://www-me.stanford.edu/globals/goodson2.html

Crap pictures thou. I guess Cathar hit the spot again .

PS:
http://ssd-rd.web.cern.ch/ssd-rd/sem...eniers-pt2.pdf
http://cjmems.seas.ucla.edu/papers/Joo_MEMS95.PDF

[bigben2k]

The heat needs to be dissipated by the screen, becvause most people who sit their laptops on their laps (hence the name!) don't want the heat on their laps...

The problem is that the display also generates a fair amount of heat, so a "radiator" of some kind needs to be somewhat isolated from it, which in turn will make the whole contraption somewhat thicker, which is not appealing.

So combining the small pump, an ultra low flow block, and some (almost) capilary tubing for a rad, a laptop could be watercooled. The only problem to resolve then, is the reliability of the tubing run, across the screen's hinge.

(Man, I ought to just shut up and market this stuff!)

[Althornin]

Originally posted by Cathar
My post copied from #Rotor's thread:

Well a few questions naturally.

1) What's the power draw to move the 200ml/min
2) Can it be scaled up? Say using a CD sized disk and pushing 1l/min
3) What about clogging issues? 1 micron tubes make for a pretty effective solids filter

3gpm is 3.3ml/sec, or about 14W/C for the water.

i.e. every 14W pushed into the water will cause it to heat up by 1C. So even a 140W CPU will only cause the water to rise by 10C as the water flows through the block, so 3gpm is still "enough" flow.

The waterblock design to take advantage of this would have a fairly thick base-plate. I would imagine around 5mm would do it. Using a 1mm wide channelled maze design would make for a pretty effective water block. Just have 1mm wide channels zig-zagging back and forth across the block. 1mm wide fins by 4-5mm high. This would keep coolant velocity in the realm of "decent", and due to the super-low flow rates, would be far less restrictive than one would think.

Sure, it'd get warm, but if the block is designed properly I think the performance would probably surprise quite a few people. It won't be a high-performance monster, but it could cool a very hot CPU far more effectively and with far less noise and bulk than a HSF arrangement.

For laptops you could probably use the back of the display as a large passive heatsink.

problem: Its not 3 gallons per minute, its three gallons per HOUR.

[Zhentar]

thats just a typo, the calculations are for an hour.

I'd be interested in seeing the math for that; my calculations using two or thee webpages came up considerably different; your probably right but I'd like to know why