block designs

[Volenti]

I'd file down/cut off some of that copper in the base of the fan part. Preferably so that the bottom two fins have about as much clearance over the WB base as there is space between the fans as you go around the fan.

What I intend to do is cut out a rectangular hole in the base of my block and insert the fan into that, then solder seal it, lap ect. the die of my (naked) p4 will directly contact the bottom of the fan, the rest of my block will be there just to direct the water flow, in theory anyway...

I'll make the hole a very tight fit (hopefully) so the plates shouldn't seperate with the heat, (well we'll see )

[MeltMan]

That is kinda what I was thinking, but on that same note, I had another idea on how to put the fan into the block... If you were to make the fan a part of the base by cutting edges on the fan...
_ __ _
| | |__| | |
| | |fn | | |
|_| |__| |_|

Solid copper on the ends

Fins in between, with "flanges" to make up the north and south ends of the block... This way you could clamp the crap out of it to keep the fins tight.

You would have to make a new fan tho

[jaydee]

You can find "some" of mine on my website at www.custom-cooling.com and a few articles from Fixittt and one from Rotor.

If anyone has any idea's on hold down device that uses the socket clips instead of the mounting holes and can fit around a block with a center inlet let me know. The board made a block(s) for does not have mounting holes.

[Cova]

Just had an idea for a hold-down, may or may not work depending on what the spec says for clearance around the outside of the socket. I know this isn't quite perfect, and I think it might have problems with the ends curving up away from the MB when pressure is applied - but maybe someone else can think of a good solution to that part.

Anyways - what if you had a plastic U-shaped piece (thin but very stiff, might have to be made of metal for strength) where the hollow inside the U shape fit exactly around a socket so you have to slide it on from the side and it can't be lifted straight up (perhaps have little teeth on the ends so it clips around the edge of the socket, though probably not needed) On the 2 vertical sides of the U there would be a hole drilled at the top and at the bottom - forming our favorite array of 4 holes in a square around the socket. The holes would be drilled not quite all the way through (to prevent screwing down into the MB) and would be threaded. Or - drill all the way through and screw stand-offs into it where the threads of the standoff aren't long enough to damage the board, and screw the block into those. This would pull the U-clip up against the socket tabs. Preferably, the open end of the U would be on the side with the socket lever so the CPU can still be removed. Also - the holes could be arranged to be in the same place as older Socket-A boards to be backwards compatible with older blocks.

I just worry about the whole U twisting - but if it was made out of aluminum instead of plastic and then coated in something non-conductive so it can't short the MB it's sitting on, it should be ok.

Hmm..., going away for the weekend, leaving in about an hour. I expect you to have solved all problems relating to this hold-down device and to the fan-style WB by the time I return

[Miss_Man]

Wow. Lotsa unique looking blocks here. Nice to see designs moving out from the chanel approach.

[airspirit]

Okay. I got to thinking when I saw a pic of this gnarly block/artwork this guy did in another post that involved swirling the water to prevent collision. Then I got to thinking about how hot spots form due to flow rate differentials in the block. Then I thought of the simple direct design of MeltMan's (good idea!), and put this together.

One inlet drops down into two exits that have a sectional volume sum equalling the inlet sectional volume. The water would be dumped directly onto the area directly above the die and there would be a small cone that would be centered on this flow that would, at its deepest point (the point) be 8mm from the die (at the thinnest point it would be 5mm thick). The two outlets would be offset to cause a whirlpool effect. This would cause there to be NO stagnant flow areas in the block and if the surfaces were roughed up for turbulence it would be a low resistance/high heat transfer design with NO low flow areas.

The inlet and outlet would be threaded for hose fittings, and te base plate would extend to the four corners for the mounting holes. This would be unsuitable for peltier use (obviously) but would be very VERY good for straight WC apps. Also note: a high pressure pump would be required to maintain a hard centripetal force on the water causing maximal flow. Also, the reason the top was "beefed up" with excess copper was to allow for more heat to be stored in the block for sudden heat spikes.

What do you guys think? It would be a bastard to make, but would work like a champ, methinks.

[bigben2k]

That WOULD be a bastard to make...

I think that you could optimise the flow a better way (but don't ask me which!). What I see is that the water is going to rush straight out to the exits, so you won't have your swirl at all, but hey, maybe you don't need a swirl...

[airspirit]

Oh, I guarantee it will swirl. The forces that would cause it to be sucked out of the outlets (suction of the pump) would come in a straight line into the outlets, not at an angle from the intake (like is supposed in your statement). The force from the inlet would be straight down over the cone. Slowly (in this case a couple of seconds) momentum in the fluid would build until there is a tornado effect inside of the block. This is why a strong pump is required. If the pump barely has the guts to power the system, it may not be able to trigger this effect that well.

The red is F and the blue is water flow.

[airspirit]

See above.

edit: The entire purpose of this would be the swirling effect. Without it this would run like crap. The whole idea is to maximise interior surface area and get maximum flow with no slow/hot spots while allowing the coldest flow to hit dead center on the die. If someone could mill this for a decent amount of cash, I'd probably buy one ... the only caveat is the need for a Y adaptor to relink the outlet streams and get them back up to size. The nice part is that you'd be able to scale the input up to 3/4" or so and keep the idea intact. You'd be able to use the entire amount of flow for cooling. Combine this with the 1" superflow unit somebody posted in the last couple of days (scale the interior up but leave depth dimensions on the bottom the same) and the cooling effect would be SICK. Imagine that whirlpool of his on top of your die ... *booge*.

[bigben2k]

Well, here's another pic.

In short, the water will take the path of least resistance. It doesn't matter what sucks and what doesn't.

red=waterflow

blue splatter=dead spot

[airspirit]

It will take the path of least resistance. You are correct. In this case, though, the path of least resistance will line up, after a couple of seconds, with the path of the force exerted by the pump. In any fluid system you'll see this happen. You are right in your diagram, though. When the system starts up, you will have exactly what you are describing happen. As the force from the outlets exerts itself on the system, though, the water will slowly build up momentum in a spiralling pattern, and eventually you'll get a full-on whirlpool effect. It is the same reason that in that picture on that post with the 1" guy the water in the funnel started a whirlpool: the inlet was offset (the hose) in relation to the funnel and it started forcing it to whirl like mad. This same thing will happen when it is the outlets that are offset. This is a well documented phenomenon ... I don't know how else to describe it except that: it will happen (if you build it they will come ... hehe).

edit: this diagram I made is definitely not to scale. The outlets would be narrower (they are half the volume of the inlet). Part of the confusion might be coming from my bad diagram. In any event, to prevent this possibility (it's not, but it'll make things clearer) you could do it like I'll diagram in my next post.

[bigben2k]

Well, the best way to observe it would be to run it outside the case, and add sparkles to the water: you should be able to visually observe the phenomenon (window notwithstanding).

I still think that it won't swirl...

[airspirit]

The only problem with this is that it will introduce stagnant areas behind the outlet baffles. But it illustrates how the flow is manipulated by the suction from the outlets.

Oh, and another way to illustrate this: those lawn sprinklers that spin around use this same principle to cause them to spin. Since the BLOCK isn't going to spin, the water will. You can't just neutralize forces in any equation.

Edit: nevermind the illustration. It wouldn't work because of the baffles. They would impede the circular flow. You're sure to have someone with a spinning lawn sprinkler living around you ... just look at the part that makes it spin. It looks remarkably like this block design (not the below illustration, but the above ones).

[airspirit]

Do we have any volunteers to mold this bastard out of something clear? Then we could sit around it and watch the swirlage.

Edit: This runs on centripetal force. We could also spin someone on a tire swing and see how far his/her barf flies. That would illustrate it. Anybody else have any yea/nay contributions to this? This site normally crawls with civil engineer types ... Skulemate? You out there?

[bigben2k]

Originally posted by airspirit
The only problem with this is that it will introduce stagnant areas behind the outlet baffles. But it illustrates how the flow is manipulated by the suction from the outlets.

Oh, and another way to illustrate this: those lawn sprinklers that spin around use this same principle to cause them to spin. Since the BLOCK isn't going to spin, the water will. You can't just neutralize forces in any equation.

Edit: nevermind the illustration. It wouldn't work because of the baffles. They would impede the circular flow. You're sure to have someone with a spinning lawn sprinkler living around you ... just look at the part that makes it spin. It looks remarkably like this block design (not the below illustration, but the above ones).

I hear you about the lawn... but that's different: the water is being ejected at an angle. In this case, there isn't anything to force the angle, so I (still) maintain that the coolant will flow in pretty much a straight line from the inlet to the outlets.

But hey, maybe it would work, just try it. I'm sure that you could jerry rig something out of pop cans or something, just for the purpose of observing the swirl.

[airspirit]

Ah, so you see: in the case of the sprinkler it is being ejected at an angle and in the case of the block it is being SUCKED at an angle (re: the force arrows above). Pushing and pulling are the same thing in the end: a force in the direction of travel.

[airspirit]

I'll make a rig out of some garden hose set in at an angle on opposing sides of a 5gal bucket. I should be able to do that at lunch. I'll fill it with water and then release the ends of the hose and see what happens. They'll be inset as shown below. I hope duct tape will hold it long enough for my experiment ... I'll report back later.

[myv65]

You can make water swirl with the right design, no question. Though it's off-track, I'm going to comment about your lawn-sprinkler note. This essentially comes down to a conservation of momentum. You bring water up into a chamber that has radial arms and then direct the arms such that the "jet" exits opposite the direction you wish the arms to spin. The reaction force imposed on the arms by the water as the arms redirect the water causes them to spin. This is purely a function of where the water comes in and where the water leaves (vector qualities).

Per your diagram, your block would try to spin. This is not necessarily because (or would not necessarily cause) a swirl exists. It's simply a question of where water enters and exits. It won't spin, of course, because it's held in place.

You say the diagram isn't to scale. Fair enough, I'll not try to guess whether or not it'll do what you wish. I will say that simply making the exits 1/2 the area of the entrance is probably not wise. Yeah, that'll force the average velocity to be the same in each, but it'll also reduce your flow. In hydraulics, orifice loss is not linear with area. Two ports of equal size that total the same area as a larger port will have a higher net flow loss. So long as the geometry and piping are symmetrical, flow will be relatively balanced between the exits. I would recommend they each be at least 2/3 the area of the entrance.

[bigben2k]

Originally posted by airspirit
Ah, so you see: in the case of the sprinkler it is being ejected at an angle and in the case of the block it is being SUCKED at an angle (re: the force arrows above). Pushing and pulling are the same thing in the end: a force in the direction of travel.

Looking forward to your experiment!

Still (I maintain) the water will tend to flow straight out:

green: cross-section of suction
purple splatter: turbulent area resulting from suction
blue: dead spot

[airspirit]

myv65: Exactly. The block will want to spin. That force will be retained even though the block will be held stationary by the bolt tackle. Since the block can't spin and the force is being exerted by the water flow, the water will then spin.

For every force there is an equal and opposite force.

My prediction is that because of the volume/mass of the water in the bucket, the swirling will be unnoticable until it is at least halfway empty (I'm praying for a little dimple like what you see when you drain the sink). In the end, though, if the water was to even spin 1/2-1 revolution in the block, the block would be a success. The whole idea is to eliminate slow spots. But, as you said we will see (I'll feel like a boob if this doesn't work ... plus, in any event, the detail guys are going to be pissed that I'm going to be chopping one of their new buckets ... hehe).

[bigben2k]

Originally posted by airspirit
myv65: Exactly. The block will want to spin. That force will be retained even though the block will be held stationary by the bolt tackle. Since the block can't spin and the force is being exerted by the water flow, the water will then spin.

OK, I can visualize what you mean, but all that will do, is change a circular motion (btw which also induces centrifugal force in a sprinkler) into one with the water, where no part moves, so you're basically using this force to "bend" the straight flow into what I call a "near straight" line.

The faster the flow, the more "bend" there will be, from the straight line, but there will be no swirl: i predict less than 1/4 turn, probably 1/8.

[Skulemate]

Originally posted by airspirit
Skulemate? You out there?

Yeah, I am, but for problems like these I'm afraid I am not much use. I'm afraid that most of the fluids I use is pretty simple stuff... I try to stick to figuring out simple stuff like flow and head loss using Bernoulli, or Hazen-Williams as my equations of choice. Unfortunately, until I take some further courses in fluids, I'm useless for stuff like this whirlpool block of yours. You're better off talking to myv65.

However, I'll throw out a suggestion anyways: why not add small fins inside the block to "encourage" the water to take the circular path that you desire? If these are kept small they shouldn't hurt flow too much, and if you keep the leading edges sharp (like a knife edge) it'll minimize stagnant flow... what do you think?

[airspirit]

Cariolis (is that right?) force will interfere with this so I'm going to do it both ways. Half of the rig is complete. When pumping into the bucket from one hose there is definite whirlpooling. Time to rig the second hose and start the draining. To do this right and get rid of any natural tendencies to whirlpool, I'm going to do it both ways (incur spinning in both directions) by repositioning the outlets to first pull clockwise, then counterclockwise. Stay tuned, guys.

[airspirit]

Thanks, Skule. I'm thinking of taking this one step further (design wise). If the circular center chamber was widened it should encourage more spinning. The way its laid out right now I should only get max of 3/4 of one revolution on the water. If I widened it by a 1/2" radius, I should get over 1 revolution, methinks. I really gotta get back to cutting this bucket, though (this is tough ... swiss army knife and heavy plastic ... ).

[gone_fishin]

Here is what the observable flow pattern is like in my cone block when purposely introducing fine air bubbles to the water (partially filled res)
There is no dead spots because the flow going down the exit channels draws the water not directly in line with them with it. A perfect verticle line can be observed down the center of the cone also.