Ultimate block? Theory.

[myv65]

@ quickmcj,

Any straight conversion of an air-cooled heatsink to liquid cooling will not be optimal. Air cooling requires a lot higher ratio of surface area simply due to the low convection coefficients that you get with gas versus liquid. In other words, the fins/pins of an air-cooled heat sink are a lot longer/taller than they need to be for use with liquid cooling.

It will work just fine, but will not be the most efficient of designs.

If you wish to create a system designed for extremely low flow rates (low convection coefficients) then this may be a good place to start. I'm talking on the order of 5 gph (~14 lph). It would result in a greater temperature rise in the water, but still perform on par or better than using a high-rpm Delta fan.

[Quickmcj]

naaa... dont want that then

We are talking ultimate block

but stille i think it is quite a good ide, if some pine wer removed, and the height of the block was smaller. In that design there will be ALOT of turbulence...and alot of cobber surface.

[bigben2k]

Originally posted by myv65
Hey BigBen,

I finally took a look at that thread you referenced about varying base thickness versus flowrate. I'll tell ya this: The "best" baseplate thickness will depend on how the block is designed. If it relies on direct impingement over the die, then a thinner baseplate would be best. If it relies on transferring heat over an area much larger than the die, a thicker baseplate would do better.

This sort of thing recently popped its head up when I started looking at the new 0.13 micron XP chips. They have a die area of ~80 mm^2 versus the "old" XP die of ~128 mm^2. Even though total wattage is lower, heat flux is quite a bit higher with the new chips. This is one of the reasons why people aren't seeing drops in temperatures despite much lower power requirements.

The new XP chips will do better with heat sinks (air cooled) that have a little extra baseplate thickness. This is because the air cooled heat sinks rely on getting the heat spread out to the fins/pins that populate the entire surface of the baseplate's upper side.

Water blocks fall into two very broad camps. One's like Michael Westen's would rely on getting the heat spread out over a large surface area. This requires a thicker base than ones that use direct impingement. (On a side note, Michael's block with channels 1.2mm wide is going to challenge the strongest of pumps. You start getting into gaps of 1.2mm (roughly 0.042") and you dive into a whole 'nother realm of fluid mechanics. Fluid surface tension and adhesive properties begin to take on significance not seen in larger channels. I suspect he'll find this out soon enough if not already.)

These two broad camps work on different premises. Michael's block sacrifices surface velocity in exchange for copious surface area. Direct impingement blocks do precisely the opposite. I couldn't tell you which is better with a strong degree of confidence, but I lean toward the direct impingement approach personally.

I agree with there being two basic block designs: the maze types, and the "direct impingement".

Actually, I didn't mean to put you out there, looking at Michael's block, sorry. I was just pointing out the baseplate thickness part. I also lean on the direct inpingement approach, just because I believe that the baseplate should be cooled in the exact reverse heat dispersion from the CPU, or at least as close as possible. In other words, cool it the opposite way that it heats up.

On that note, assuming that 5mm would be good in a direct inpingement (aka center inlet) design, I'm thinking about reducing the baseplate thickness, maybe as far down as 1mm, around the core itself: it would produce a copper island, slightly larger than the core itself (the new AMD core of course), and would allow the coolant's effect to be concentrated in the hottest area of the copper. Here's a rough pic.

[bigben2k]

Just to confuse everybody... here's a re-post of a block design I posted in another thread.

Ok, here's my design.

The first one is to be capped with a clear acrylic piece. It is a single channel straight through, with rows of square pyramids over the area of the core (actually they're offset, so that the water doesn't flow straight through between the rows).

An alternative would be to make a second top, where a center inlet would exit either side, but I'd like to see the channels shallower.

The second design is a 3/8 or 1/2 barb that fits inside a radial pattern of fins. The water exits the block through a 2 inch tube, which is attached to the 2 in barb that's mounted on the block. Of course the small tube is inside the big tube, and the flow is seperated through a special res.

The center barb needs to let the water out, so it can't be screwed in all the way down: it's held in place by the fins. As for the large barb, well, I don't have one handy, I haven't even found one, but any way it can be mounted will do, even if it involves soldering a short ~2" copper pipe, and using clamps, whatever will hold the 2" tube.

The fin design is irrelevant actually...

[bigben2k]

Combining the 2 posts abose, design#2 and the copper island theory, the fins would have to extend over the copper island (they can be brazed or soldered into place) to hold the center barb.

BTW, the center barb only needs to be held so that it doesn't flail around, it really doesn't have to be sturdier than that. Actually, the fins could have pins that point straight up, to hold the tube in place.

Comments?

Volunteer CNC'ers?

[myv65]

Hey Ben,

Don't forget about structural integrity in all this. A 1mm thick base isn't liable to support the 12-24 lbf of clamping pressure (AMD spec) without excess deformation.

As to removing the heat the same way it goes in, it doesn't really work out that well. It gets in via conduction through a thermal interface of some finite thermal resistance. Interface resistances tend to be small compared to the convection coefficients we can generate. This means that even direct impingement coolers require higher delta-Ts on the water side than the "inlet" from the die. Isolating yourself on an island around the core would also be detrimental. Yeah, you'd force all heat to get out through a small area, but so what? All you're doing is losing out on what small percentage would otherwise escape outside the "island". Given an alternative between a good conductor the size of the die only and another with the same conductance but much larger area, the larger area will do better even with direct impingement.

Oh yeah, and no problemo taking a look at Michael's design, I just didn't have time at the moment I read your post earlier.

[Volenti]

On that note, assuming that 5mm would be good in a direct inpingement (aka center inlet) design, I'm thinking about reducing the baseplate thickness, maybe as far down as 1mm, around the core itself: it would produce a copper island, slightly larger than the core itself (the new AMD core of course), and would allow the coolant's effect to be concentrated in the hottest area of the copper. Here's a rough pic.

my athlon block is similar to that (posted here a while back, the one I "carved" out) it's the best performing block I've made to date.

[Limeygreg]

Here is my block that I finished recently. It would have been done 6 weeks or so ago but I went on vacation for 2 weeks and when I got back the milling machine had been moved to our helliport facility (hellicopter maintenance). I had been waiting on a rotary table so I could mill the o-ring channel and it arrived just as I was going on vacation. The block had to be effective but simple to mill as the milling machine is not CNC equiped.

[Limeygreg]

Here is one of the finished assembly. I made a clamping mechanism that used the P4 plastic frame that is attached to the mobo as I didn't want to have to dis-assemble my motherboard from the case in order to mount the WB.

[ECUPirate]

good call with the mounting system... Looks sweet. Please post temps...

[Dix Dogfight]

Limeygreg
Nice job on the surface of the WB. How did you manage to get it so rough?

Edit: I'm talking about the inner surface not looking smooth. Otherwise excellent job

[bigben2k]

Originally posted by myv65
Hey Ben,

Don't forget about structural integrity in all this. A 1mm thick base isn't liable to support the 12-24 lbf of clamping pressure (AMD spec) without excess deformation.

As to removing the heat the same way it goes in, it doesn't really work out that well. It gets in via conduction through a thermal interface of some finite thermal resistance. Interface resistances tend to be small compared to the convection coefficients we can generate. This means that even direct impingement coolers require higher delta-Ts on the water side than the "inlet" from the die.

Isolating yourself on an island around the core would also be detrimental. Yeah, you'd force all heat to get out through a small area, but so what? All you're doing is losing out on what small percentage would otherwise escape outside the "island". Given an alternative between a good conductor the size of the die only and another with the same conductance but much larger area, the larger area will do better even with direct impingement.

Oh yeah, and no problemo taking a look at Michael's design, I just didn't have time at the moment I read your post earlier.

I know what you mean about the structural part. I'm going to have to run a few numbers on that, but the concept remains the same. Alternatively, the copper island could be mounted in plexi, but then the structural integrity of the block would be in far greater danger.

As for the heat conduction, I understand that, given a solid block of copper, the heat would spread in essentially a "spherical" pattern, from the source. So I'm thinking about reproducing that "sphere" in order to apply the cooling evenly. Also, this "sphere" can't be too thick, otherwise I'm defeating the purpose, so I used the 5mm measurement (although still preliminary) as a guide, to come up with what I did (design #2B)

I hear you about using a larger mass as a baseplate, but I'm exploring the idea of minimizing the copper mass, to its absolute smallest (structural integrity included though). That's how I came up with design#1 (center inlet option). I don't believe that the extended perimeter of the copper baseplate is necessary, other than for structural purposes. The core is only 80mm^2 in area, so why would I otherwise use a 2 in. by 2 in. plate, especially if the heat spreads from the center out?

[myv65]

Originally posted by bigben2k
The core is only 80mm^2 in area, so why would I otherwise use a 2 in. by 2 in. plate, especially if the heat spreads from the center out?

Convective heat transfer = h * A * delta-T

h is mainly a function of fluid properties and for a given fluid, velocity is the dominant variable. If you fix A to 80 mm^2, then delta-T becomes inversely proportional to h. h has practical limits based on the maximum velocity we can develop. This means if you limit your block to 80 mm^2, you have a minimum achievable delta-T (block to fluid) that you'll approach asymptotically (sp?) as velocity heads toward infinity.

In the case where the block matches the die size, you can calculate the delta-T required for conduction across the block by assuming unidirectional heat transfer. This is simply q = k A / l * delta-T. Say q = 75 watts, k = 401 w / m - K, and A = 80 mm^2 (80 * 10-6 m^2). Then delta-T / l = constant. At l = 5 mm, delta-T = 11.7 K (same scale as °C). At this point, something should sound fishy. You know darn well that a copper heat sink with a Delta fan will keep a chip within 12°C of the air temperature, yet your "die sized" block can't possibly do as well because there's nothing left for block-to-fluid delta-T.

You absolutely require some spreading of the heat to take place.

If 2/3 of the heat goes to your die-sized block and 1/3 escapes to the 2" X 2" periphery, then your block's peak delta-T (which WILL occur over the die) drops from 11.7 to 7.8 °C. That's still not so good, but now in the ballpark of the best (read: Loudest) air cooled solutions.

Now you should understand where baseplate thickness comes into play. If the baseplate is really, really thin then not much spreading will occur. In this case, you best have some serious direct impingement going on. As baseplate thickness increases, the percentage of the thermal energy that gets dissipated "outside" the core region increases. You need to start balancing direct impingement against maintaining flow over a wider surface area.

IMHO, the best designs rely on direct impingement for a substantial portion of the heat transfer as this provides excellent flow instability (read: minimal boundary layer thickness) while carrying that flow through to the perimeter where it can absorb additional energy. You still need to maintain decent velocity outside the "die box" in order to keep the convection coefficient high, but you do get to trade some of the initial velocity of the impingement for the added surface area around the periphery.

I'll tell ya what, I really liked Paul's block that he posted a few days back (and wants $120 US to ship). Not only is it a work of art, but it also takes a solid (solid from the engineer's perspective) approach to maximizing heat transfer from direct impingement AND the outer perimeter. On a side note, it has a strong similarity to what I posted a while ago as my idea of a good low-flow block. Paul's is better than what I showed as he still maintains the direct impingement that I did not and he also has a better exit than I showed.

As a final footnote, I'll repeat what I said earlier about the 0.13 XP chips. Their smaller die size is a bigger pitfall than people realize. Before you had ~72 watts going through 130 mm^2 whereas now you have 62 watts through 80 mm^2 (numbers from AMD's tech docs on XP2100 chips, 0.18 and 0.13 construction). The heat flux (watts / mm^2) is over 37% higher on the new XPs. This means you have less total heat, but a lot higher concentration of what heat remains. Baseplate thickness should increase a little to compensate for the higher concentration of heat.

I haven't seen anything concrete from any heatsink or block manufacturers on this point. Probably because (A) I haven't looked and (B) they probably haven't thought too much about it. Anyway, this is the primary reason why people aren't reporting lower temperatures for the new XP chips and their lower power consumption.

[dream caster]

a really interesting and informative thread!

I have an idea that could suggest something more to a guy involved in this stuff: what really interest us is not turbulence as such but thoroughly mixing water and having the thinnest posible boundary layer. I've thought of a kind of "ordered turbulence" having vortexes the size and orientation you want and where you want them.
Mostly I thought about a hard disk cooler but that could be used for other blocks; in this case I would use triangle cross section fins parallel to flow and between them short fins at 45º to flow to make water spin. the small fins would go one direction in one row and at 90ºdegrees in the next row so where vortexes touch they would have the same flow direction. (i'll post a drawing later i have to go to work now).

[ECUPirate]

*spelling police*


dream caster, you might want to check your bio... "Desert" has one 's', unless you live in the middle of a bowl of ice cream in Chile, in which case two "s"s is correct.
....remember, dessert makes you fat, so it has 2 's's. No food in the Desert, so just one 's'.


oh, and welcome to the forum.

[Skulemate]

Dream Caster... nevermind ECUPirate... this spelling help is coming from an accountant who can't count

[bigben2k]

Do you mean Paul_Vodrazka? I guess I don't understand your definition of "(direct) inpingement". I thought you meant "central inlet", like design#2.

I'm guessing that you're talking about the narrow channels, but not too narrow (like what Michael Westen posted in the OC forums.

I understand that the baseplate should be of the appropriate thickness, according to the design. In a maze type design, the baseplate is best left thick, since the cooling will occur over a significant portion of the baseplate (including the channel walls).

But then you say that with direct inpingement, it should/could be thinner. If you're talking about a center inlet, then I agree, hence design#2. I also tend towards design#2 (aka direct inpingement, aka center inlet) as performing better.

As for the calculation that you just presented, I understand them, however, the island being so close to the core would get hotter, would it not? This would increase the delta T, no? With direct inpingement (my favorite), it would work better, no?

I understand the limitation (as described by #rotor, i think) where if the baseplate is too thin, the coolant may not be flowing with sufficient speed to pick up the heat.

I also understand that *some* spreading is required, otherwise directly cooling the core would be very popular by now. I know that fins will do that (and really, that's what the baseplate is), but on an 80mm^2, with normal tools, unless you can put something together like Volenti did (i.e. the copper wire brush), there are not too many options. Even bandsaws are thick.

So I figure that there's got to be a balance between the heat dispersed by the baseplate/fin, and what the water can take away.

Maybe another approach... How much heat can a coolant (water) absorb, at reasonable speeds(i.e. flow rate, we can design a channel for a speed)? I mean what's the maximum rate, given a delta-T of say 10 degrees?

Knowing that, it might be possible to optimize the baseplate and fins.

[bigben2k]

Another note, since the issue of boundary layer has been brought up... I found this link last night, while searching for a coolant that would perform well under water's freezing point.

What it is is dimethyl polysiloxane or silicone oil.

[myv65]

Hey Ben,

I'm going to backtrack for a sec a clear up a couple of things I said. First, my long post above was referring to your "island" design, not design #1 or design #2. Second, I had to take another look at Paul's block. Really the only difference between it and what I posted was his exit. I mistaken thought he had a center-inlet, aka direct impingement. (That WOULD be a nice upgrade ).

The big point I was getting at is that the delta-T between chip and the side of the block that meets the fluid depends on many things. If you constrain your cooling to an area equal to the die area then you are required to have a higher delta-T across the baseplate then if you allow spreading. When I read you idea for removing heat in the same way it enter, ie through a small rectangular area, I wanted to make clear that this is not an effective solution.

The absolute worst case scenario for the baseplate is having to pump heat through a small area. Per my example, pumping 75 watts through an 80mm^2 area 5mm thick requires a delta-T of nearly 12°C. Allowing 1/3 of the heat to spread outside that 80mm^2 region drops the peak delta-T by 4°C.

Now, you ask about having a hotter baseplate in contact with the fluid. Sure, this will raise the baseplate to fluid delta-T, but since the fluid temperature is practically fixed (say at ambient for ideal sake), then raising the fluid-baseplate delta-T requires a higher CPU temperature for a given baseplate thickness. This is obviously not good.

There's a whole lotta stuff to think about and this single thread is probably as good a discussion as I've seen.

[ECUPirate]

Originally posted by Skulemate
Dream Caster... nevermind ECUPirate... this spelling help is coming from an accountant who can't count

I just found out today, I passed all four parts of the CPA exam w/ high marks, and on my 1st try. I'll post a topic RE this in Random Nonsense, as I don't want to hijack this thread...

[Limeygreg]

Congratulations ECUPirate - my tax package is on it's way to you, can you have it ready for Aug 18th ?

I'm a little upset as I just spent 20 mins or so typing a reply, then I attached an image but when I posted it I was told that it was too large and of course my whole post dissapeared, so here is a shorter version.

All images of block can be found here :-

http://www.geocities.com/limeygreg/E...aterblock.html
http://www.geocities.com/limeygreg/F...aterblock.html
http://www.geocities.com/limeygreg/W..._Assembly.html

But please be patient as they keep closing it due to excessive d/l as I previously posted this on OC.com where I spend most of my time (same handle). Any questions, ask away.

I don't have my radiator yet, still hunting - if anyone knows where I can find a cheap Serck aviation H/E I'd be very grateful. What rad can I shoehorn into my Antec 1200 ??? Temps will probably be meaningless unless I can get at the cpu with a thermocouple (which I don't have), I have a Soyo Daragon P4S and I wouldn't trust the temp monitoring on that any further than I could piss upwind in a hurricane. I do have something coming though that may be of interest to you so hang on a while until I get it up and running and then I'll post a few revealing pictures (now, now, minds out of the gutter please).

The combined CSA of the channels is a little more than the CSA of the 1/2" tube that will be suppliying it, although the barbs are 7/16" ID (my next ones will be true 1/2" ID). The block should operate in the high velocity/turbulent domain but I want to reduce the flow to see the effects, the roughness was to disrupt the boundary layer if the flow was reduced to the point where laminar flow started to establish itself. I was going to shotblast the block but the only machine available was a beast and very indiscriminate and would have required a sheet metal mask to protect the upper surface. The answer to how I achieved that finish is in the picture - it took about 15 minutes but i found it to be very controlable, and whats more, it's readily available to anyone. The finish on the upper surface is from the shotblaster and it has also seen some wire wheel action. (a dremel engraver - that's my first block, I used it to experiment on )

[dream caster]

the drawing took me hours and it is no good :there should be two parallel plates with longitudinal fins one just above the other and the 45º to flow fins in one plate at 90º to the other. Ihave some thoughts to use this kind of design for waterblocks even if there is only one plate

[bigben2k]

Ok Limeygreg, what's your baseplate thickness? How deep are the channels? what flow are you getting out of it (i.e. how restrictive is it?)?

dream caster: do you mean fins that are bent over?

[Limeygreg]

Hi BB2K

Copper stock was from a grounding bus plate so it's only 1/4" thick. The channels were made with a 3/16" milling bit (smallest I had at the time) and are a little under 3/16" deep leaving a little more than 1/16" at the thinnest part. I had calculated the CSA and SA figures but now I can't find them - I suppose I could sit down with a calc and do them again. The CSA is slightly larger than the csa of a 1/2" ID tube but the ID on the barbs was opened up from 3/8" to 7/16" so they are probably going to be the limiting factor for now. I designed it to not be the bottleneck in a 1/2" system, I wanted it to be just a tad larger CSA so that I would get maximum flow velocity through the block, the limiting parts will be the interconnect tubing - although at present it is the 7/16" ID barbs that will limit the flow.

I am planning two other poly tops. One will have true 1/2" ID barbs and hopefully will be at about a 45 degree entry angle (although I will probably have to change the clamping mechanism if it works significantly better than the perpendicular entry barbs). The second top will have three barbs, 1/2" in the center and a 3/8" outlet at either end. I want to see just how much difference the different inlet/outlet methods affect the performance.

The system is not built yet as I am still working on the rad, but I am actually leaning toward putting together a test bed with a copper slug and resistive heat source (I have plenty of ceramic resistors at work as well as a Motorola Variable PSU that will give me 20v at 40 amps - so I got the heat source covered ). I'm going to try and build a manometer to measure delivery presure. Flow rate will have to be by timing and weigh the coolant (water) method.

I have access to some interesting temperature measurement equipment and hope to use that to give me a idea of which inlet/outlet configuration works best - so don't hold your breath waiting as I will be doing it during my breaks at work so it may be a little while. I am also planning another block base but with 1/16" wide x 1/8" deep channels, the tops will be interchangable.

Several people have asked what the rate of flow is, I can't answer that yet as I have not tested it - but the problem I have with that question is that it's not meaningful if you don't reference it to a pump delivery pressure. One of the other aspects I hope to investigate is how much heat the pump puts into the water. The way I see it is that a good tuned system will use a high flow low resistance block with a low resistance radiator capable of bringing the coolant close to ambient using minimal air flow (read fan noise here) and a pump just big enough to circulate the coolant (with a little to spare) - so, as a system it will be efficient and quiet.

[Brad]

http://xjinn.procooling.com/block%20...ess/block6.jpg

comments?