So you've thought of a water jacket concept. Have I got it right? An outside condensation jacket to assist with cooling the inner core which itself is a waterblock with its own in/out feed?
The water jacket itself I think would mainly deal with condensate.
The block would have to be designed with an ultra-slick coating. Teflon maybe.
When condensate formed on the block, it would immediately bead and drain down the water channel into runoff tubes which simply drain into a small reservoir which is free to let water evaporate. Or hell, it could have a small dehumidifer built into it which would accelerate the evaporation process.
The water condensation jacket would allow the block to condensate all it wanted without harming components.
Aye, which makes it even easier to drain. Because then all one must do is make the structure of the block such that it slopes off to the sides in spots where water could pool. Mind you that the shell would encompass the WHOLE block and drain to that valve. It would be sealed with silicone at the in/outlets, and it would be sealed with a temperature-resistant material where it came up against the core-contact. Basically, a hermetically-sealed shell around everything but the contact.
You have to make it so that the water wouldnt drip at the left corner. That sharp edge could make it deattach from the surface. IT would be better off to have something on the left and right to drop off of or gather.
So Thrax and I have very similar ideas. He proposes to have the waterblock within a container of sorts. The dead airspace prevents the condensation from getting to the components. This "jacket" also acts like a channel for that moisture to accumulate and run off.
I, however, proposed that the jacket be made of a material with a higher thermal insulating value. The jacket/container would a) prevent moisture from getting to the surrounding components. While the waterblock itself may get cold...the insulating barrier would prevent that cold from every reaching the visible surface.
Much like wrapping the waterblock in neoprene as an example. The fact that the container is an insulator would actually make the waterblock more efficient. (a very small % more efficient but more efficient nonetheless).
Actually...the dead airspace around Thrax's waterblock would aid in efficiency as well.
The two could be easily blended. My original design had the waterblock fitting into the jacket with no airspace. Adding a standoff design to the design would give the channels/dead airspace required by Thrax's design. The jacket would be made in two halves and the inner surface could be coated with Teflon.
The inner waterblock would have the raised processor core surface. That would fit into a hole at on the bottom half of the container. That hole on the inner surface would have a slight raised lip (the same "height" as the standoffs for the rest of the waterblock.)
The hole and waterblock raised core surface would be mated for size thus a tight fit. Simply apply a small bead of silicon around that hole on the inner top surface and when the waterblock slides in...squish and thus a seal.
Same again with the two halves...a gasket would be better. Four screws holds the two halves together and a threaded drain hole on 2 of the sides provides for draining at either orientation.
Now who here knows of a material that is a very good insulator, can be machined via a programmable router and is fairly cost effective? (No space shuttle tiles here)
The cost of the waterblock essentially does not change. The technology is proven. It wouldn't have to be as big or as thick either. Only enough to be durable against the pressure of the clip and stand up to the force of the hose nipple assembly over time.
To see if I have this right, you're basically thinking of taking a standard water block, then encasing it in a shell comprised of two halves?
As in, you set the block in the bottom half, seal it properly and affix it to the board. Then you bring the other half down on top of the block, seal it, and wrench it down to form the seal?
To see if I have this right, you're basically thinking of taking a standard water block, then encasing it in a shell comprised of two halves?
Bang on. The inner side to the two halves would have round standoffs (5mm. in diameter). At least four per half. 6 would be better. and two on each inner "side". This would provide for your airspace and a firm seat for the inner waterblock.
As in, you set the block in the bottom half, seal it properly and affix it to the board. Then you bring the other half down on top of the block, seal it, and wrench it down to form the seal?
Could do it that way but it woud be handier to assemble the unit out of the PC case. Four screws through the unit to hold it together. This may present a problem though four mounting. The most cost effective way would be to have the four screws holding the unit together double as the mounting screws.
That in itself is a problem. You want more torque on the screws to hold the two halves together but not as much torque through the PCB. It would probably be better to go with an eight screw pattern. The four corners would align to the mounting holes in the mobo. Those would be our pressure screws for adjusting waterblock to PCB/Processor core.
The other four screws could be Alan keyed screws inset (not on the corners) for holding the waterblock jacket together. It would probably be better to use hex head bolts with a locking washer.
EDIT: Okay...Thrax and I have done our job. Whose the material expert here?
EDIT EDIT: DAMN If I had access to an Autocad driven machining router I could have a prototype made out of Aluminum within a couple of hours. It'd be a simple job..really.
My neighbour is a machinist...maybe I could...hmmm.
/me thinks of ordering a waterblock...hell...custom design that too.
I think it would be better to torque the two halves together in your hand rather than mounting it on the motherboard then getting into an awkward area to do that.
Quick release hoses? Or good ol' slide on the tube and use a collar? I think the latter would be more watertight and cheaper.
Putting it all together outside of the case would make changing a processor easier and attaching hoses easier.
EDIT:
Okay...machining the waterblock is easy. Two 3" square chunks of copper 1 inch thick. Piece one: Side A: machine down the surface leaving a postage stamp sized "bump" equal to the thickness of the container jacket plus the height of the standoff. Hell...you could do it as a circle...makes for easier maching in later steps.
Piece one Side B: router a "S" channel for the water flow.
Piece two Side B: mirror of the "S" channel. Piece two side B: Drill in/out holes at either end of the "S" channel and tap them for the hose nipples to screw into. Apply machine sealant grease and torque in the nipples.
Weld the two halves together and buff whole unit and finish the heatsink contact area. Waterblock is done.
The two halves would be mirrors of each other. Side A would get a square (or round) hole equal to the waterblock postage stamp contact area. Side B would get two holes of sufficient diameter to accommodate the nipples.
You could get fancy and tap those holes to screw in a threaded collar. A metal Slide a metal sleeve over the hose followed by a gasket. Attach hose with crush collar for watertight-ed-ness. <---whoa...that's a new word.
Anyway slide gasket and metal sleeve over nipple assembly and screw hand tight.
Unit is now watertight. The only exposed surface of the waterblock is the postage stamp sized core area which get's cold as it should do.
I think you'd be better off using standard barbs and hoseclamps. It's cheaper AND more effective, which you can't beat.
But as for your waterblock's interior, you must remember that the more turbulent the flow is, the more effective it is at dissipating heat. You might recall Swiftech's diamond pin matrix on the interior of their blocks, which is one of the most effective cooling interiors in the industry.
I understand the ease of an S-channel, but something a little more advanced, lending itself to a turbulent flow (More surface area too) would be keen.
I looked at the Swiftech pin pattern and I don't think it's a matter of turbulence. It's a matter of maximum surface area. Heat comes up through the base and goes into all those pins. All those pins are surrounded by fluid so there is a heckuva lot more surface area for dissipating the heat.
Cheap Solution: Fine bit router drill to cut a checkerboard patter into one half of the waterblock. Like a tic-tac-toe box except a lot more of them. Surface area to fluid is thus increased.
Not so cheap solution.
The two halves are router'd out to form a box. One half has fins tacked onto the inner side. Heat transfers up the fins and this has even more surface area to fluid.
Imagine that heat is a water-soluble powder. Which gets the powder mixed into the water faster:
Sitting there, or being stirred (turbulence)?
We all know that surface area helps in shedding heat from the sink, however now you have to contend with the dynamics of a fluid arriving in a finite amount, upon which you must rely to carry heat away. Instead of an unlimited volume of ever-circulating air, you need to maximize the amount of heat taken away per cubic centimetre of water. That takes turbulence.
Swiftech's waterblock design seems to be a variation of their pin heatsink. It's like they encased the pin heatsink. Makes sense for production technology as the equipment/design exists already.
I think turbulence as churning river water. I keep forgetting to downscale it to low flow rates.
Hmmm...low flow but still able to "churn" up the water and have maximum surface area...and be able to be routered out.
"S" channel = Minimum effectiveness.
"Checkboard" pattern = better effectiveness
What about a waffle style? Fluid is force up then down then up and so on.
Cheaper to machine than mounting pins individually or tack welding fins? The waffle pattern could be at a 45 degree angle to the square of the waterblock. Fluid enters at a "corner" then broadens out to center and then narrows back to outlet. Double churning effect?
If ya got money to burn then sure. ANYTHING is possible. I'm thinking cost effective here as you know us modder types. We'll bitch about the cost of anything...but usually end up spending 3 x as much.
There are a bunch of solvents that are not too flamable that are possible to use. There are a bunch of thin oils also. The question is how do you want the system to run. Do you want phase change at the CPU or not?
I saw the ultimate phase change cooling system at HP once. The took a big RISC chip and removed the heat spreader. Then they took an old inkjet plotter print head, as I recall it had 120x240 jets. The print head and the chip were almost exactly the same size. They figured out how much heat was generated at each point in the CPU. Then they set the print head to shoot micro drops of water at each location so that the water just evaported when the next droplet hit. They were able to keep the actual die temp 102C under any load. I wanted to see them do it with alcohol and run cooler.
If you used a "water cooling" system with a fluid that had a low boiling point then you could use the phase change to absorb the heat, return the vapor, condense it and repeat. This would remove heat much more efficently that water cooling. These don't need to be refrigeration systems. You could do it without compresors and such if you selected a liquid that boiled at 80-100F.
What is wrong with Pelts? They can be water cooled, there is no cold surface exposed that isn't in contact with the CPU, you can use water to dump the excess heat and the are easy to control. Ok, so they are expensive and power hogs. Do you want everything at once?
I would love to bring home a couple of Vortec tubes and hook them to my compressed air system in the basement. Blow very cold air against the CPU, vent the hot side out of the case. Loud as hell but cool.
Ah, you're talking about direct-die evaporative cooling. A DIYer, whose name and project I've forgotten, did that a while back. I think he did it with a Barton, if anyone wants to combine "Barton" and "Direct die" in google.
The problem with pelts is condensation and frost. You actually have to contend with ice on your motherboard.
Hmmm... A very intriguing design. Simple interior, yet water would be sloshing around like crazy.
/me grins
Just remember that the barbs should be pointing in opposite directions from one another, to facilitate avoiding kinks in the tubing.
Bloody hell, if we actually made something of this, we could actually market it as "The first waterblock specifically designed for chilled-water applications!" or somesuch.
What would be great for cooling is hacing fins like thermal rights SLK HSs. Make the water flow across all those fins. That or something like swiftech has for their HSs. The coils. That has a lot of surface area and would create a lot of turbulence if the solution was forced thru the coils.
Neither of those solutions could be done in one step tho and this cost a lot more. I was just trying to apply the best of air cooling to the interior of a water block.
Fins do help heat transfer. The size of the fin is dictated by the thermal properties of the working fluid. Air=large fins, water=very small fins. If you tooled the inside of the flow passages in a pattern like a knurl it would be fine. In industrial power plants finned tubing has fins that are only ~0.025" tall.
Turbulance helps heat transfer, and eat power. You need to be careful. You also want the coldest water at the outside edges, and the warmest in the hotest area. This is called counterflow and it will increase the total heat transfer.
You could also look at making the contact face out of silver, no bigger than needed and only 3mm thick. Then make the block out of stainless steel. It has very poor heat transfer, it would help cut you losses. It would be easy to solder the two together.
Comments
So you've thought of a water jacket concept. Have I got it right? An outside condensation jacket to assist with cooling the inner core which itself is a waterblock with its own in/out feed?
The block would have to be designed with an ultra-slick coating. Teflon maybe.
When condensate formed on the block, it would immediately bead and drain down the water channel into runoff tubes which simply drain into a small reservoir which is free to let water evaporate. Or hell, it could have a small dehumidifer built into it which would accelerate the evaporation process.
The water condensation jacket would allow the block to condensate all it wanted without harming components.
All water would be downhill.
I, however, proposed that the jacket be made of a material with a higher thermal insulating value. The jacket/container would a) prevent moisture from getting to the surrounding components. While the waterblock itself may get cold...the insulating barrier would prevent that cold from every reaching the visible surface.
Much like wrapping the waterblock in neoprene as an example. The fact that the container is an insulator would actually make the waterblock more efficient. (a very small % more efficient but more efficient nonetheless).
Actually...the dead airspace around Thrax's waterblock would aid in efficiency as well.
The two could be easily blended. My original design had the waterblock fitting into the jacket with no airspace. Adding a standoff design to the design would give the channels/dead airspace required by Thrax's design. The jacket would be made in two halves and the inner surface could be coated with Teflon.
The inner waterblock would have the raised processor core surface. That would fit into a hole at on the bottom half of the container. That hole on the inner surface would have a slight raised lip (the same "height" as the standoffs for the rest of the waterblock.)
The hole and waterblock raised core surface would be mated for size thus a tight fit. Simply apply a small bead of silicon around that hole on the inner top surface and when the waterblock slides in...squish and thus a seal.
Same again with the two halves...a gasket would be better. Four screws holds the two halves together and a threaded drain hole on 2 of the sides provides for draining at either orientation.
Now who here knows of a material that is a very good insulator, can be machined via a programmable router and is fairly cost effective? (No space shuttle tiles here)
The cost of the waterblock essentially does not change. The technology is proven. It wouldn't have to be as big or as thick either. Only enough to be durable against the pressure of the clip and stand up to the force of the hose nipple assembly over time.
It's a pity we're a country away.
//EDIT:
To see if I have this right, you're basically thinking of taking a standard water block, then encasing it in a shell comprised of two halves?
As in, you set the block in the bottom half, seal it properly and affix it to the board. Then you bring the other half down on top of the block, seal it, and wrench it down to form the seal?
Bang on. The inner side to the two halves would have round standoffs (5mm. in diameter). At least four per half. 6 would be better. and two on each inner "side". This would provide for your airspace and a firm seat for the inner waterblock.
Could do it that way but it woud be handier to assemble the unit out of the PC case. Four screws through the unit to hold it together. This may present a problem though four mounting. The most cost effective way would be to have the four screws holding the unit together double as the mounting screws.
That in itself is a problem. You want more torque on the screws to hold the two halves together but not as much torque through the PCB. It would probably be better to go with an eight screw pattern. The four corners would align to the mounting holes in the mobo. Those would be our pressure screws for adjusting waterblock to PCB/Processor core.
The other four screws could be Alan keyed screws inset (not on the corners) for holding the waterblock jacket together. It would probably be better to use hex head bolts with a locking washer.
EDIT: Okay...Thrax and I have done our job. Whose the material expert here?
EDIT EDIT: DAMN If I had access to an Autocad driven machining router I could have a prototype made out of Aluminum within a couple of hours. It'd be a simple job..really.
My neighbour is a machinist...maybe I could...hmmm.
/me thinks of ordering a waterblock...hell...custom design that too.
Flush hex nuts would be most adequate to make a firm seal, I agree.
Quick release hoses? Or good ol' slide on the tube and use a collar? I think the latter would be more watertight and cheaper.
Putting it all together outside of the case would make changing a processor easier and attaching hoses easier.
EDIT:
Okay...machining the waterblock is easy. Two 3" square chunks of copper 1 inch thick. Piece one: Side A: machine down the surface leaving a postage stamp sized "bump" equal to the thickness of the container jacket plus the height of the standoff. Hell...you could do it as a circle...makes for easier maching in later steps.
Piece one Side B: router a "S" channel for the water flow.
Piece two Side B: mirror of the "S" channel. Piece two side B: Drill in/out holes at either end of the "S" channel and tap them for the hose nipples to screw into. Apply machine sealant grease and torque in the nipples.
Weld the two halves together and buff whole unit and finish the heatsink contact area. Waterblock is done.
The two halves would be mirrors of each other. Side A would get a square (or round) hole equal to the waterblock postage stamp contact area. Side B would get two holes of sufficient diameter to accommodate the nipples.
You could get fancy and tap those holes to screw in a threaded collar. A metal Slide a metal sleeve over the hose followed by a gasket. Attach hose with crush collar for watertight-ed-ness. <---whoa...that's a new word.
Anyway slide gasket and metal sleeve over nipple assembly and screw hand tight.
Unit is now watertight. The only exposed surface of the waterblock is the postage stamp sized core area which get's cold as it should do.
But as for your waterblock's interior, you must remember that the more turbulent the flow is, the more effective it is at dissipating heat. You might recall Swiftech's diamond pin matrix on the interior of their blocks, which is one of the most effective cooling interiors in the industry.
I understand the ease of an S-channel, but something a little more advanced, lending itself to a turbulent flow (More surface area too) would be keen.
I looked at the Swiftech pin pattern and I don't think it's a matter of turbulence. It's a matter of maximum surface area. Heat comes up through the base and goes into all those pins. All those pins are surrounded by fluid so there is a heckuva lot more surface area for dissipating the heat.
Cheap Solution: Fine bit router drill to cut a checkerboard patter into one half of the waterblock. Like a tic-tac-toe box except a lot more of them. Surface area to fluid is thus increased.
Not so cheap solution.
The two halves are router'd out to form a box. One half has fins tacked onto the inner side. Heat transfers up the fins and this has even more surface area to fluid.
Imagine that heat is a water-soluble powder. Which gets the powder mixed into the water faster:
Sitting there, or being stirred (turbulence)?
We all know that surface area helps in shedding heat from the sink, however now you have to contend with the dynamics of a fluid arriving in a finite amount, upon which you must rely to carry heat away. Instead of an unlimited volume of ever-circulating air, you need to maximize the amount of heat taken away per cubic centimetre of water. That takes turbulence.
Swiftech's waterblock design seems to be a variation of their pin heatsink. It's like they encased the pin heatsink. Makes sense for production technology as the equipment/design exists already.
I think turbulence as churning river water. I keep forgetting to downscale it to low flow rates.
Hmmm...low flow but still able to "churn" up the water and have maximum surface area...and be able to be routered out.
"S" channel = Minimum effectiveness.
"Checkboard" pattern = better effectiveness
What about a waffle style? Fluid is force up then down then up and so on.
Cheaper to machine than mounting pins individually or tack welding fins? The waffle pattern could be at a 45 degree angle to the square of the waterblock. Fluid enters at a "corner" then broadens out to center and then narrows back to outlet. Double churning effect?
If ya got money to burn then sure. ANYTHING is possible. I'm thinking cost effective here as you know us modder types. We'll bitch about the cost of anything...but usually end up spending 3 x as much.
We might as well admit it to ourselves.
I saw the ultimate phase change cooling system at HP once. The took a big RISC chip and removed the heat spreader. Then they took an old inkjet plotter print head, as I recall it had 120x240 jets. The print head and the chip were almost exactly the same size. They figured out how much heat was generated at each point in the CPU. Then they set the print head to shoot micro drops of water at each location so that the water just evaported when the next droplet hit. They were able to keep the actual die temp 102C under any load. I wanted to see them do it with alcohol and run cooler.
If you used a "water cooling" system with a fluid that had a low boiling point then you could use the phase change to absorb the heat, return the vapor, condense it and repeat. This would remove heat much more efficently that water cooling. These don't need to be refrigeration systems. You could do it without compresors and such if you selected a liquid that boiled at 80-100F.
What is wrong with Pelts? They can be water cooled, there is no cold surface exposed that isn't in contact with the CPU, you can use water to dump the excess heat and the are easy to control. Ok, so they are expensive and power hogs. Do you want everything at once?
I would love to bring home a couple of Vortec tubes and hook them to my compressed air system in the basement. Blow very cold air against the CPU, vent the hot side out of the case. Loud as hell but cool.
The problem with pelts is condensation and frost. You actually have to contend with ice on your motherboard.
Crude first rendering of what I was thinking for the inner waterblock.
/me grins
Just remember that the barbs should be pointing in opposite directions from one another, to facilitate avoiding kinks in the tubing.
Bloody hell, if we actually made something of this, we could actually market it as "The first waterblock specifically designed for chilled-water applications!" or somesuch.
Neither of those solutions could be done in one step tho and this cost a lot more. I was just trying to apply the best of air cooling to the interior of a water block.
Turbulance helps heat transfer, and eat power. You need to be careful. You also want the coldest water at the outside edges, and the warmest in the hotest area. This is called counterflow and it will increase the total heat transfer.
You could also look at making the contact face out of silver, no bigger than needed and only 3mm thick. Then make the block out of stainless steel. It has very poor heat transfer, it would help cut you losses. It would be easy to solder the two together.