I swear...this maze design makes no sense to me!

I mean why? I've seen SOOOO many home made blocks trying to immitate DD's Maze design, and I'm aware it works, but WHY??? I mean correct me if I'm worng: The ONLY part of the processor that gets hot is the CORE, so have the water move around the rest of the CPU?

I know, I know: It works, so why ask!!!! But, really now, WHY does this design work?

it has something to do with the turbulance it creates and therfore it creates more surface area for the heat to dissapate through hence better cooling. Someone else would and will probably explain it better then i can.

well if you feel a waterblock when the computer is on you will see that the whole waterblock gets warm, so by creating a maze design the water is cooling off the whole maze, not just the center because water doesnt absorb heat that well so it needs more surface area to work good.

The heat is acually disipated threw the copper this is what makes it work by the heat being spread so it can be cooled by the watter

cheezboy had the right idea and i will pick up where he left off... surface area is the key in this concept. the heat from the die dissipates throughout the entire block. the maze design on their blocks is designed primarily to maximize the surface area of the inside of the block with which the water comes into contact. i said primarily because the maze design also serves a second purpose which is slightly harder to visualize. the second purpose it serves is that it creates turbulence so that the water churns more as it passes through the block, therefore more water molecules come into direct contact with the block (direct contact between water and block is more efficient than heat dissipateing through water molecules due to greater temperature difference). Both these design concepts serve to increase the contact between all the water molecules and the inner walls of the block.

when designing a block, the benefit of turbulence needs to be balanced with flow rate, as the more turbulence introduced the harder it is to move water quickly through the block.

[only important part]the goal of a good performing waterblock is to allow maximum contact between individucal molecules and the block, through a perfect balance of flowrate and turbulence, in order to achieve maximum heat transfer from block to water. [/only important part]

side-note: as heat dissipates through water it moves relatively slowly as compared with block to water heat dissipation. the water touching the walls of the block is the hottest, the water in the middle of the channel is the coolest. you want the greatest temperature difference between block and water molecules. so turbulence causes more "cooler" water molecules to come into contact with the block. if you consider only heat dissipateing through still water you will realize that as the heat disipates it will be transfering from a warm molecule to a slightly less warm molecule and so on. turbulence skips some of the middlemen and introduces more "cooler" water molecules. (i say "cooler" because the actual temperate difference between nearby molecules is almost infintesimally small, however that tiny difference is multiplied by a larger almost infintesmal number of water molecules and therefore does present an advantage.)

It's basic astrophysics that I think can be best summed in If-Then form. If said body A6 Hunk of copper is larger has more surface area and on average creates more turbulance than said body B6, C6, or D6; Then said body A6 will cool better.

-Toysrme

BATTLESHIP SUNK!

that is a good correct explanation for a general understanding, however it leaves out flow rate which has the identical function as turbulence - introducing cooler water molecules to the block. his explanation sums it up best however and is much more fun to read.

The Maze design is an incredibly inefficient way to design a waterblock. It is flawed in a number of ways. The opening poster is on to something when he says about "Why cool around the core when the core is the only thing that's hot".

This is totally correct. A block that's 1/2" x 1/2" x 1/2" and designed properly is all that's required to significantly outperform a Maze design.

It performs pretty good for a commercial block, is pretty easy to make, is highly available and, in general, looks good. It does exactly what they want it to do - Generate sales, especially when people new to watercooling beleive the "commercial hype".

OF course, theres way better designs being made (Cathar's for example), and some in planning.

figures, right when I go to phase-change cooling something new has to pop up making watercooling even better. oh well I like the sound of -20c temps better

cather, mind if I use your design for a direct die block lets see how it does with liquid propane running through it

Mazes work by first spreading the heat into a large surface area, and then using that surface area to transfer heat to the water. This allows for good performance even with relatively low flow rates.

Warlord2 said:

cather, mind if I use your design for a direct die block lets see how it does with liquid propane running through it

Click to expand...

Go for it. That'd be great to see.

There are really two functions of a waterblock

1. Get water from cpu core spread out through the block itself

2. Get water from the block into water which will be taken away and replaced etc etc etc

1. is really based on the same as heatsinks are. A material which is a good conducter of heat (copper) and as much contact with the core as possible.

2. Has seen alot more innovation because Well its alot easier to carve copper differently than to dish out the cash for a material that conducts better. (Theres what? Silver, single diamond crystals, and some manmade stuff probably) And silver probably corrodes really damn fast or something (I didn't take chemistry, or if I did I dont remember it).

More surface area = more heat transfer
More turbulance = Faster heat transfer (You maintain a greater temperature difference between copper and water)
More flow = Faster heat transfer (You maintain a greater temperature difference between copper and water)

(both turbulance and flow have essentially the same effect as a better radiator... but they do it very differently and they rely on an effective radiator (Introducing more water to the blocks surface isn't very effective if the water is almost as hot as the block itself).

Cathar's statement is correct, however, it does do size a bit of injustice. A maze-sized block that is as well designed as the 1/2" cube would be Much more effective, assuming their was enough heat to make need for it.

There are several variables involved and I'll list as many as I can think of.

1. Good transfer to block (Entirely the block and its mounting systems concern)
2. Good transfer to medium (water) (Block and pump, and (due to restriction of flow) the entire systems concern) This includes terbulance, and flow rate and #3
3. Temperature of medium (water). Radiator's concern however other factors such as pump heat output and ambient temperature are involved.
4. Medium (There are mediums that would work better than water, however most are somewhat impractical.)


The device needing cooling is also important.

An Athlon is a better candidate for watercooling than a pentium. Why?
Lets say you can safely run an athlon up to 70* C A pentium to 50*C
Assuming the water temperature stays constant, A hotter core (with effective heat transfer to waterblock) would make for more heat transfer to the water.... thus a more effective system (assuming the radiator can handle the heat)

This is the biggest bottleneck in watercooling in my opinion.

With a good block, pump and a good radiator with enough airflow With NO active cooling element (unless you count the airflow and waterflow). I've seen on die temps within 5* C of ambient. If your ondie temp is 30* C, I doubt you Can improve much more because that's only a 1-2* imperfection at each transfer.

What could be bottlenecking you?

1. Transfer from cpu to block. Not much to do here. Lapping, as3... that sort of thing. (Direct die removes this variable I believe... But thats just an assumption based on the name.)

2. Transfer from block to medium (water). Alot here. Surface area, turbulance, flow rate, Medium temps, Mediums thermal properties, maybe more Im forgetting

3. Transfer from water to radiator. Again not much. There may be some improvements to be made on heater cores but I dont know enough about them to think of them. You can increase surface area (bigger radiator) The temperature of the heater core itself is another thing that can be changed by improving the...

4. Transfer from radiator to air. More airflow, better directed airflow (To the whole radiator, higher flow concentrated where the most heat is congregated (Probably the area closest to the inlet)) This can also be improved by improving...

5. Air temperature. Crank up the damn AC or in winter open a window. Maybe even create a ducting system running from window to window and put your radiator in there. Put the radiator on the floor (unless whole system is there, this will lessen flow)

All very valid points, all, thanks for posting.

This brings up several questions of mine (I'm a pretty big idiot): turbulance would be, well, how much the water gets smashed around inside the block, correct? So a smooth channeled block, the maze for example, would have very little turbulance (other than that made by routing the water through the spiral design)? So a block with something to interrupt the smooth flow of water through channels (ie spikes or something) would cause more turbulance, and therefor better heat transfer, correct?

Just to sum it up for me: from what I understand, the factors influenceing the efficiency of a water block are: 1) the speed of which the heat can be transfered from the core, through the block and into the surface water; 2) the surface water itself: by which I mean, the amount of time certin molecules of water stay in contact with the edges of the WB, creating turbulance shortens this time; 3) flowrate, etc.

I'm going to read up on direct die: this sounds real interesting to me. I've been trying to fudge together a WB real cheap, with limited success. thanks again.

a good WB ill perform as well if not better than direct die because when you do direct die, there's alot of heat in one smaller spot, thus the water has less surface to pick heat off of. Whereas a WB will pick all the heat up, and spread it out over the whole of the copper thus allowing easier heat dissipation.

any gains made by not having that core-> WB transfar layer (in the WB case) are negated by the lack of surface area


then again, I could be wrong


direct die is prety much permenent, remember that.

Nick C said:

a good WB ill perform as well if not better than direct die because when you do direct die, there's alot of heat in one smaller spot, thus the water has less surface to pick heat off of. Whereas a WB will pick all the heat up, and spread it out over the whole of the copper thus allowing easier heat dissipation.

any gains made by not having that core-> WB transfar layer (in the WB case) are negated by the lack of surface area


then again, I could be wrong

Click to expand...

Volenti over at OCAU has probably one of the best direct die implementations I've seen recently. Hard to quantify it exactly but from what he's seeing and been experimenting with it would be a very close race between the White Water and Volenti's Direct Die setup.

Lithan said:

Lets say you can safely run an athlon up to 70* C A pentium to 50*C
. . . . .I've seen on die temps within 5* C of ambient.
. . . . .

Click to expand...

what is the basis for your postulating a 20°C (max ?) operating difference between an Athlon and a Pentium ?
- links ?

you second statement is flat wrong
'instruments' can be made to read whatever number is desired,
what you (may) have 'seen' means nothing without substantiation
- links ? (and be quite prepared to have to defend other's work if thats what you are going to cite)

I have made thousands of such measurements (ck out some old articles) and can assure you that a 5°C differential will limit your CPU to ~ a 15 or 20W output
- is that what you meant ? (I think not)

your assessment of wb size is not correct
wbs are large 'cause its easier to make them that way, not because the CPU's heat generation necessitates such

your assessment of 'bottlenecks' is wrong
-> there is quite a lot to be gained at the TIM joint (relatively speaking)

need to sort out a bit your opinion from fact

be cool

exactly correct there bill, I would be making my blocks 1/2inchx1/2inch if I could, but it just doesnt really work that way... Its really difficult to get center inlet and corner outlet for 1/2inch blocks in 2x2 material, but I have managed to do it, but I MUST solder these blocks together, or glue them with some sort of good aheasive, I wish there was a better way.....

Lithan,

The fact of the matter is that the temperatures reported by CPU's are mostly junk.

What I find amusing is that whenever a motherboard comes along that is calibrated somewhat close to reality, everyone immediately jumps on it and dismisses it as "reading higher" than the rest, and is therefore inaccurate and is dismissed.

Question is, of course, who defined accurate and how was it defined?

P4's? They all read about 10-15C lower than reality. Volenti has pulled the heatspread off his, and doing a rough plot of die temperature vs "predicted" watts and charting a line through the data points finds the P4 intersecting the 0W line at some point WAY below ambient.

Same with many Athlons and their motherboards. I have the Asus A7V333. Everyone, including many on these forums, believes that the A7V333 reads temps about 10C higher than accurate. I have a fair amount of anecdotal evidence that suggests that it still doesn't read temperatures high enough by 10C. I can get it to read sub-ambient temps, yet everyone insists it is a "high reading" motherboard.

What you're basing your opinion on with respect to C/W is around a collective ego massaging by motherboard makers (Intel and AMD alike) to fool customers into thinking that their CPU temps are accurate, and that 0.05C/W's are possible, which they aren't, well not for <300mm^2 CPU dies they aren't (P4/XP = ~127mm^2).

BillA is coming at this from the engineering point of view. He has the equipment to measure and validate what are accurate C/W's. He knows what is possible and what isn't because he is using calibrated equipment.

You are basing your opinion on what totally uncalibrated motherboards are telling you, which near as I've been able to establish, ALL of them lie by dramatic amounts, and even the highest reading boards on the market STILL don't read high enough, although some come close, but when they do come close everyone dismisses them as being horribly high.

Supplementary to this is the actual watts that the CPU is dissipating. Every single thermal calculator in existance greatly overstates the actual wattage output of a CPU, and this leads to people reporting even lower C/W figures.

It's a case of garbage in means garbage out when it comes to the C/W figures that people report.

BillA said:

be cool

Click to expand...

when doesn't he sign his posts like this?