Waterblocks: Why a block?

I'm planning to water-cool a CPU, and, in searching for the best design I've noticed what to me appears a very limiting assumption in one area.

What do you call the part of a water-cooling setup that carries heat off of a CPU?

If you say 'waterblock', you then must start design with a block of solid material which is nothing more than the worst heatsink form possible, then try to correct the problem by pumping water through (usually) an interior channel. Some of these are just lovely with Celtic knots and rollercoaster switchbacks, wormpaths wending poetically, or a girlfriend's initials. I get lost in idiot spirals trying to design The Better Waterblock.

Most block patterns are layed out in 2D! This is obviously going nowhere.

So, let's forget blocks and start again. Let's say the object is to hit the CPU hard and fast with alot of water and get out quick to make room for the next "parcel" of water. By hitting hard I mean getting the water very close to the heat source. I think everyone agrees that circulating around there is no good. An inlet next to and outlet with a pea-sized channel comes to mind, but it has little surface coverage and poor volume.

All we really want is an inlet, an outlet, and a splitsecond moment of heat transfer before the carrier makes room for the next carrier. So, align the carrier movement at 90* to the CPU surface. Of course you'll never get water to flow back through itself, molecule by molecule (at pump speeds, anyway), but we can very nearly accomplish this by bundling many inlets and outlets together. For a rough image, picture a bundle of soda straws.

There are several ways to smoothly seperate the inflow and outflow internally, so you only have two connections on such a unit.

Have you forgotten about waterblocks yet? I really hope other people can contribute to this idea, so I will not post my own design yet.

I'm working on (yes still, I haven't given up on it yet), a block that is triple layered. So the coolant will stay in the block a little longer but make it's way out the top, after having passed through the third layer's maze. I don't know if it's going to work, but my friends sure do give me weird, loks when I try to explain it to them.

huh?

Sean Lindstrom, perhaps you could make a quick paint sketch of the design you had in mind so we could understand it better?

CrystalMethod said:

I'm working on (yes still, I haven't given up on it yet), a block that is triple layered. So the coolant will stay in the block a little longer but make it's way out the top, after having passed through the third layer's maze. I don't know if it's going to work, but my friends sure do give me weird, loks when I try to explain it to them.

Click to expand...

That seems like its a very good idea, keep us informed on the progress

i think I get it...

Basically a block is boring, since contact time isnt as important with high flow pumps, why not get creative with the water block. Ergo, dont make it a block at all..

i have seen articles from people who use something other than blocks... didnt cool very well at all. However, I will let you guys in on a secret. I could care less how cool my cpu gets. Im going with a h20 and Pelt system just to overclock a little. If my temps run in the 60's... thats fine by me.
SO.. i may get creative with the waterblock. Problem is, i have no way to machine metal So i guess I wont get that creative.

Here's the first draft. What you're looking at is the side view of a cyrindrical unit, constucted mainly of standard PVC pipe fittings.

It occured to be that moving water removes heat better than a lump of metal, so I want to maximize the water volume.

A: This is the only heat conductor, a thin sheet of copper. It could be a 2X2 square or disk. I want it just thick enough to lap flat after assembly. Just above this is the water exchange layer. The goal is to have the cool water contact the copper and evacuate quickly. You'll notice that the whole bottom surface does this.

B: The weirdest part of this unit. Tubes within tubes. The inside (blue) tubes are probably soda straws (thin-walled and nearly frictionless). The outside tubes are just the voids between the the inside tubes and the oversize holes drilled through the (grey) fitting at B. There're other ways to accomplish this water movement, and, since I've never seen tubes within tubes in nature, it's probably a bad idea. Then again, I've never seen anything like a conventional waterblock in nature.

C: The inflow (blue) tubes pass through this chamber. The less heat gain from the warmer chamber water, the better. The warm (red) water makes it's way out through gates at...

D: Here the inside PVC pipe has a few legs projecting downward to seat it against the inside walls of the outside PVC pipe.

E: Here's another plate, with holes to tightly fit the sodastraws. A honeycomb pattern gets the most straws. A very hard rubber would be a good material, although nearly anything that can snugly hold those straws (well, they don't have to be straws) so they're perfectly aligned is OK.

F: Here's the 1" outlet and inlet (both are PVC because it welds great to other PVC). The inlet needn't be welded to the inner chamber. I'm thinking of heat-forming the outlet side of the inner chamber into a crude hourglass, to improve flow at the mouth of the outlet. 1" is for volume. Ideally, though, the pump system would use tubes having a sectional area equal to one-half the surface area of the heat source.

G: End caps. These could be clear. Not shown are the 3 or 4 setscrews used to position the inner chamber.


Any questions, just ask.

I'll be using a chilled reservior, but I think this CPU cooler would be effective with ambient-water systems too.

Ok, I think I follow what your saying. I added (quickly) arrows into your pic. Now that is the direction of the water correct? The water will go down, then up all the sides and back out the opposite of the in. Correct?

The only problem Im seeing here would be that when the water hits the bottom of the "cooler" where the cpu is, it looks like the area where the water comes out is very very thin. So the 'straws' end almost directly on top of the metal plate that comes into contact with the die. Here is where I see a possible problem.

You would have to make sure that the water that is pushed in from the inlet will not be slowed down here. If it does, then the water would end up being stuck there longer than wanted and would heat up a little more than wanted as well.

Like the idea. Some possiblities I think. Keep us posted if you get something like this working and how it handles it.

IFMU

ack... dang pic....

ack... dang pic....

i think their are to main reasons why a block was picked first ONE most heatsinks and squares because u can cram the most in it within a fair size and i think thats another reasons why its a block u can get the most tubelence in the smallest amount of room which if im right means better cooling.. now if u can find a better design then by all means..

Only drawback that I can see is the warm return water is being taken passed the incoming cold water, hence warming it up a little.
This is probably a minute amount of heat transference, but bears taking into account.

Like the design tho, and talk about turbulence.

there are many flaws i see in your design...

is has been proven in waterblock designs that surface area plays a critical role in cooling capacity, which is why the spiral/maze design with the walls often perform best. (Spir@l/Maze-2/Maze-3)

turbulance is another factor, the flat smooth area you have that touches the CPU Die will not be very affective because like i said above, it lacks surface area, and it lacks turbulance. I would suggest reading this thread before confirming your design: http://overclockers.com/articles511/

another reason "blocks" have been chosen as the best medium for thermal transfer is because they can absorb heat, and if you know anything about thermodynamics, you know that when something is cold, it still contains heat. So, when at idle, the block ussualy absorbs the coldness fromt he water, when the CPU suddenly goes to load it will take a while for the CPU die to fully heat up, which is why temps ussually rise slowly. In your design, i see a lack of medium to absorb the "coldness" of the water, therefor when the CPU gets load it will heat up VERY quickly, like in a mater of seconds, where as normal WB designs take a matter of minutes. the sudden flash from COLD to HOT will cause CPU instability and it can and will cause damage to the CPU itself because when things get hotter they expand, and yours will expand very rapidly (much like the cracks in your driveway are formed from theremal expansion)

Not to be offensive, but i do think a conventional waterblock design is much better than yours. Sure yours might cool better than air, but not as good as a good waterblock that can keep the load temps less than 6 degrees C above ambient.

that's allright. nice idea. looks like a nightmare to build though. let us know how you make out.

looks like a direct-die cooling setup, with a thin copper base added to it.

You are going to have massive amounts of head loss with that design. While it's a nice idea, you should consider ways to have a less restrictive flow path.

uhmmm id hate to tell you this but

he said it looks like a direct die setup but i think that it looks like a peice of metal with some metal surface area inside so to me it is a "water block". but keep in mind that when i say water block i mean something that water pases through to transfer heat away from another object. but all im trying to say is that in some and most of our areas of intrest something as indepth as that just isnt needed when another much simpler idea can do the same thing maybe even better due to the restrictions that your block calls on.

you may also like this

im not sure if you like this idea howeveer it goes one step farther. if you wher to shorten the straws you could place a heat sink on the bottom this would cause better heat transfer, but ps dont get to crazy you may forget that a block sitis in the computer sideways on standard towers.

Very interesting. I think for the most part, it should work pretty well, but there are a few considerations to take into account.

1. Base thickness- It is a fact that a thicker base spreads heat over a wider area than a thin base. with such a thin base, it would be frivolous, and perhaps counter-productive to have more than 1-3 tubes spraying down onto the base.
2. Flow rate- Although the design will increase turbulance, and minimize the boundary layer, it looks as if it could be restrictive to flow. This might not be a problem, however, because the velocity of a fluid increases as the restriction increases (the venturi effect), and the resulting increase in turbulance, and decrease in boundary layer may make up for the losses in flow rate.
3. Mounting System- Come up with a way to mount it securely. It shouldn't be too hard, but it should be adressed.
   [/list=1]
   
   The only thing I have to say, is that I can't wait for someone to make one.

Syk0tiK, your arguments against this block design seem to be without merit. Sorry, but your post was kind of condescending, and not based on any sound engineering principles. I am certainly not an expert on the subject either, but I have done a lot of reading on the subject.

1. There are blocks that perform better than the spiral/maze designs.

2. This design could create fully developed turbulence from constriction of flow to many, smaller-diameter conduits. Most mortals can't predict block performance without empirical data, unless it's *really* obviously flawed.

3. Very massive blocks, with huge heat capacities can perform very poorly. I won't meantion any names, but I have proof. Really, heat transfer is confined to a very small portion of the water block. The only advantage of a massive block is a safety factor, in case the pump fails.

I think it might be a kick-*** design, with a little refinement! The major problem with the block is that it would be expensive to produce in numbers.