Cooling My Athlon64 Waterblock!

I think you have hit the nail on the head here.

This is definitely the best way to go, and will also be easier to machine, while using a cheaper material (1/2" polycarbonate will be cheaper than 3/4")

8-ball

 

Interesting stuff

 

I could explain in more detail, but I'm sure I've already done that in the forums some where, and I'm a little short of time.

8-ball

 

After a long wait finally an update!

I got back the final polycarbonate tops from my machinist. This seems like the wining final design. The tops are 1/2" thick (roughly 13mm for you British folk). They also sport new mounting holes for P4 and A64 processors. Here are some pictures...





Now that I have finally perfected the top portion of the block I can’t help but think the bottom can be perfected as well...
Currently the "pattern" on the copper base plate is as follows...


I have been thinking about these designs to replace it...


and...


Id like to know what you think of each performance wise. Especially you 8-ball! If you have any other ideas id like to see them as well.

 

Nice to see the finished top. No expert on the internal design, but perhaps a spiral design will create a better movement of water?

 

very nice looking block you got there now, cant wait to see some reviews of it as i'll be switching to A64 shortly.

 

Those lugs looks a whole lot stronger now Ferrari

In fact they are starting to look much more like the finished article and I don't doubt that they will give a good return in both $$$$££££$$$$ and temps.........any plans on who gets to test them????

Good luck

 

@[GiB] Spawny : the spiral design is good but it has a laminar flow. (the water goes in a straight forward orderly fashion through the channels). In a laminar flow the outer part of the water (the molecules...) that's in contact with the side of thew channels (the copper itself) stays on the side. Thus the molecules in the center also stay in the center of the flow throughout the whole block. Therefor those don't get heated as much. You get what we could call a layer effect. In a turbulent design ( such as the ww ) shape are added to cause turbulences (ie:stire up the water..). Ferrrari's current design is a turbulent one. The rules for thermodynamics are different in such a setup. I don't know them but I can tell you that good turbulent design have been known to outperform the spiral design by quite a league.

 

I see the errors of my thoughts thanks TetarZ for the explaination

Perhaps they should all be made as testing prototypes and after testing the winner stays? I should imagine they will be very close?

 

Ferrari,

The poly top is looking considerably better now thanks to the full thickness lugs. I do like the slots rather than the holes to allow for fitment to the P4 as well as the athlon 64 mount. Mind you, I think Intel are changing to the LGA775 spec soon, though with the popularity of athlon 64 and the introduction of the relatively low cost 3000+, that shouldn't matter in a business sense.

Anyway, back to the design of the block.

As I said at the beginning of the thread, I don't think the cascade style dimples are a good idea without the corresponding nozzles to facilitate easy flow of coolant out of the dimples.

What you have is a large volume, but not too high a velocity, jet impinging on the surface above the core. What you need to think about is a "furniture" design below the central inlet, which will increase both the surface area above the core AND the turbulence of the, BUT it must be "self clearing".

By this I mean that it shouldn't have pockets which, when the block is filled with water and flow applied, will "probably" not receive much fresh water. In effect, you want to minimise pockets of stagnant water, which is what I think would have happened with your dimple design.

However, what you need to be wary of in trying to meet these requirements is compromising the strength of the base to the point where it bends under the immense clamping pressure required by the the athlon 64. (make sense)

I hope this helps you in your quest, and good luck

8-ball

 

I have to first thank you all for your criticism in the past, it is what has made the block what it is today.

I’m still struggling with the design on the copper base plate. I realize that the cohesive properties of water will some what "stick" it to the insides of the dimples. Thus little water will flow in or out. I was thinking about making the dimples larger and have them overlap so there is a cavity where water can more easily flow. But this design creates little turbulence and in my mind wouldn’t improve performance significantly. Thus I have come up with other designs listed above.

For you to understand the flow of the water you must understand that there is no specific channel the water must flow through in this block, much unlike other blocks where the water must snake its way through a maze of channels. The inside of the block is open creating a somewhat thin cavity where the water must "squish" down allowing maximum surface area contact to the copper base plate. To increase performance the bottom of the block needs some sort of design milled into it to create turbulence and increase surface area. The biggest problem I face is areas where water will become stagnant thus becoming useless for cooling. Let me post 2 pictures of more finalized designs then I can explain how I came to each and you can, yet again, help me decide which is best.

1.


Here is my logic behind this design. The overall flow of water through the cavity of the block is lateral ( <----------> ). By introducing geometries perpendicular to the flow the water must "leap" over the peaks and spiral around (turbulence) in the valley much like you would see wind do going over the ridge of a house. (Thanks 8-ball for the fluid dynamics link).

The problem I face is the width of the channels and there depth. If I make the channels to deep water in the bottoms will become stagnant and essentially dead weight as far as I’m concerned. Too shallow and I loose turbulence. The width of the channels is also very important. Too wide and they aren’t much of an improvement over a flat bottom too skinny and water wont be able to pass into the valleys.

Finding the balance is what I’m struggling with. I am wondering if someone can enlighten me on a best case scenario. Might end up being, make the channels as deep as they are wide? I dono. I must say at the moment I’m leaning towards this design.

2.


The idea behind this design is; place fins parallel to the flow of water and the water will easily flow along and through the fins. The only enhancement these fins have over a flat bottom is they increase the surface area, there downfall is they create little to no turbulence. I wouldn’t have to worry so much about water becoming stagnant inside these channels due to the direction of the flow of water. This design seems like the easy way out of the "dead water" problem, but sometimes you have to keep it simple right stupid

 

Ho well ferrari has just given out a better explenation than mine. I hope it is clear for everyone now...

@Ferrari: I vote against the second wb: If there are no turbulences you are in a laminar flow model. In such a scenario the length of the maze is essential (yes length, surface is important but the length is really what you are looking for) and therefor spiral is just about as good as it gets. You seem to be raising some very technical points> I think procooling is a more appropriate place to have such a discussion. I myself cannot help you as I'm only a poor little computer science undergradute (and no physics phd) . However I do encourage you to keep posting here as I am very interested by the conclusions you willl come to.

Good luck.

 

I don't entirely agree.

It is less likely in the second case that there will be stagnant coolant, particularly if the fins extend out of the range of the inlet, as the coolant entering the fins can get out more easily than in the first case. and stagnant water is worse than laminar flow.

However, I think that in a design where there is a large volume above the furniture, encouraging the coolant to flow through the furniture rather than over it will always be difficult.

Ferrari, if you could alter the design slightly such that the cavity in the base of the poly top were shallower and the copper base thick enough to machine a good design, then the coolant would be more encouraged to flow through the dooling furniture than over the top of it where is is doing little good. I hope that makes sense.

As for the design it self, do you have access to conical endmills, which could be used to machine a "V" profiled groove rather than a square sectioned channel. I would imagine that a diagonal (imagine a pattern based around an "x" rather than a "+") "criss-cross" pattern of such grooves, combined with the (much) shallower cavity might perform quite well.

Other than that, the only thing I can suggest is to get a few bases made up with different designs and try them out, to see which performs best. I would also suggest constructing a simple die simulator for said testing, as it would prove much more accurate and reliable, further aiding the quest for perfection.

Remember though, stagnant coolant is the enemy, and it is no good having an elaborate design if it is filled with stagnant water.

Hope this helps.

8-ball

 

Definatly some nice work here Ferrari. Looks like your block has really made some improvements. When can we see some reviews on this thing, it would be great to have a new waterblock on the market.

 

I got back some demo Cu bottoms with a new pattern I came up with milled in them. Here are some pix.





These are just demos but you get the idea. Im thinking about increasing the depth of the fins and possibly adding more, what do you think?

 

ooo nice stuff gonna try and invest in water cooling soon as just got new upgrades of athlon64 3200+ and radeon 9800XT and need something big too cool these mofo's down while playin games

 

Ferrari, it is more important to have "furniture" above the core, than it is to have it around the rest of the block.

8-ball

 

Fluid dynamics, and thermodynamics are very complicated subjects, which I know only the basic concepts of. When I was working at a university I had dealings with a MEMS group (mechanical electrical micro systems). They basically investigated incorporating mechanical, electrical, fluid, thermal, and chemical systems onto chips. Some of the examples of mems stuff I've seen are amazing: turbines on a chip, 128bit mechanical locks etc (you should all do a search for info on google, it's quite interesting ). They were looking at fairly simple (and boring stuff) that used electro mechanical actuators to filter particles from fluid in a cavity between a channel. To simulate designs, they used computational fluid dynamic software. I think that the software that they used was called CFX:

CFX

It would be really great if water block designers could use this as it would make debates about how to get turbulant and laminar flow redundant. I believe that this software also simulates thermodynamics, so in theory, you could completely simulate your water cooling system. This software and others like it are used in multiple engineering disciplines from aerospace to power generation. Things like jet engines don't get designed without software of this kind. Trouble is, my company had a copy of it. It's uber expensive (makes Solid Works, or AutoCAD look cheap), and the liscence only lasts a year (I believe). It's the kind of thing that they won't directly state a price for, you have to e-mail, or call, then they come round to see you to "discuss your needs" (which means they want to know how much your willing to pay for it).

Another option might be to fill the block with cleaar water, and then pump dyed water through it (or vice versa). You could video the process and observe the flow of water. Just a thought .

 

CFX seem to have some interesting software. Do they have evaluations to download?

 

I think, though I'm not sure, a few of the guys at procooling have used the waterloo software.

8-ball