Made my own waterblock!

[bigben2k]

Most puzzling!

As I remember from Les' graph, there is an intersect between 1mm and 6 mm bp (or am I remembering it all wrong?), which would indicate that it's possible to get the same result, but that the performance would differ at different flow rates.

:shrug:

[Graystar]

Well, les did say he was guessing...

I actually used the Ahanix Iceberg setup just because I figured the low flow would give the most noticable results.

Hey, I have material for lots more blocks. If you have any other ideas you want to try just send me more bases

[theetruscan]

Originally posted by Graystar
The only way to do that is to increase the height of the channel, but I don't see how that would benefit. I've yet to make a version with a wider channel. That would increase the area. I'll probably try that next.

The reason I asked is because I think there would be a benefit. Whether it was from a wider channel, or a taller channel it would be the same effect. My reasoning is pretty simple, you create significant, unnecessary pressure drop by having the cross-sectional area of the block significantly less than that of the barbs.

It seems to me that you would get much better flow rates with the larger channel. And, as this is laminar flow, it seems equivalent to some of the maze/spiral design blocks (i.e. laminar flow etc). Those blocks, looking at BillA's testing (specifically the GeminiCool Tsunami), benefit much from increased flow rate.

One indication that it may be useful is that D-Tek, Danger Den, etcetera have all designed their blocks with the channel cross-section at least as

Originally posted by Graystar
I actually used the Ahanix Iceberg setup just because I figured the low flow would give the most noticable results.

I think this is entirely wrong. All of the waterblocks tested by BillA benefit to some degree from increased flow rate, but the amount of benefit varies greatly between blocks. It may well be that your extroardinarily "tight" block creates such pressure drop that both of your current blocks are performing well below where they could, and that differences would begin to appear if you redid them either taller or wider. I think those components could in fact be minimizing the differences by slightly compromising the blocks functioning.

Also, I think one of the arguments for a thicker base (I know, I'm beating a dead horse) is that with other laminar flow blocks, the manufacturers use thicker bases. Another argument is that some of the makers here have tested their blocks, keeping everything the same but reducing the thickness of the base and found that there is an ideal point, and it's not necessarily as thin as possible.

Anyway, hope this wasn't offensive, I do that sometimes. I'm impressed regardless, I just think you can get a better block without too much work.

[Graystar]

Originally posted by theetruscan
The reason I asked is because I think there would be a benefit. Whether it was from a wider channel, or a taller channel it would be the same effect. My reasoning is pretty simple, you create significant, unnecessary pressure drop by having the cross-sectional area of the block significantly less than that of the barbs.

It seems to me that you would get much better flow rates with the larger channel. And, as this is laminar flow, it seems equivalent to some of the maze/spiral design blocks (i.e. laminar flow etc). Those blocks, looking at BillA's testing (specifically the GeminiCool Tsunami), benefit much from increased flow rate.

Well, why speculate. Next chance I get I'll make one with 4 channel pieces and give it a try.

[Volenti]

Originally posted by theetruscan
The reason I asked is because I think there would be a benefit. Whether it was from a wider channel, or a taller channel it would be the same effect. My reasoning is pretty simple, you create significant, unnecessary pressure drop by having the cross-sectional area of the block significantly less than that of the barbs.

It seems to me that you would get much better flow rates with the larger channel. And, as this is laminar flow, it seems equivalent to some of the maze/spiral design blocks (i.e. laminar flow etc). Those blocks, looking at BillA's testing (specifically the GeminiCool Tsunami), benefit much from increased flow rate.

One indication that it may be useful is that D-Tek, Danger Den, etcetera have all designed their blocks with the channel cross-section at least as



I think this is entirely wrong. All of the waterblocks tested by BillA benefit to some degree from increased flow rate, but the amount of benefit varies greatly between blocks. It may well be that your extroardinarily "tight" block creates such pressure drop that both of your current blocks are performing well below where they could, and that differences would begin to appear if you redid them either taller or wider. I think those components could in fact be minimizing the differences by slightly compromising the blocks functioning.

Also, I think one of the arguments for a thicker base (I know, I'm beating a dead horse) is that with other laminar flow blocks, the manufacturers use thicker bases. Another argument is that some of the makers here have tested their blocks, keeping everything the same but reducing the thickness of the base and found that there is an ideal point, and it's not necessarily as thin as possible.

Anyway, hope this wasn't offensive, I do that sometimes. I'm impressed regardless, I just think you can get a better block without too much work.

I'll just say one thing, in all the blocks that I've made, and I've made dozens, the blocks that perform the best, are the ones with the highest pressure drop.

[Graystar]

Originally posted by Volenti
I'll just say one thing, in all the blocks that I've made, and I've made dozens, the blocks that perform the best, are the ones with the highest pressure drop.

That's what I was thinking. Getting more water through probably won't make a difference unless that water is touching something hot. Certainly in my block I'm not gonna get any more water touching the base than I already have. But, you never know...I'm willing to try.

[theetruscan]

Originally posted by Volenti
I'll just say one thing, in all the blocks that I've made, and I've made dozens, the blocks that perform the best, are the ones with the highest pressure drop.

I think I misspoke, or you misinterpreted me. Probably the first. With complex blocks, or jet impingement blocks, the pressure drop serves a distinct purpose. With blocks that rely on laminar flow (such as maze style blocks or Greystar's), the performance improves with less pressure drop. Like the Maze 3, which performs better with 1/2" barbs than 3/8" barbs. Whether it would improve with increased barb size if the internal channels had a 3/8" cross section, I don't know. But the two are quite different as far as I can tell.

Also, you use an Iwaki MD-20RZ, no? That might have a major impact.

I think Greystar's design will show a benefit from reduced pressure drop because of the kind of block it is.

[hara]

Originally posted by Les
Dunno,but maybe the different configurations could behave like this:-


The flow-rate being controlled by the pressure-drop (dP).Estimating the dP is fraught with difficulties, but a summary of my guesses:-

Where did you get the data

[Cova]

Why does everyone assume this block depends on laminar flow? It's really not all that small inside (total cross-section area), there is a very sharp bend upon the coolants entrance to the block where it slams into the base and then is pushed out of the way by more coolant coming down. Do you really think that the flow smooths out that much in the 1/2" it travels before it's over the core?

I would try making it a bit wider (say the inside-width of the channel should be about 4mm wider than the CPU core - so a 2mm + sidewall width overlapp on each side), perhaps score the bottom parallel to the flow (scoring perpendicular will likely just create tine deadspots in the bottoms, and I don't think the turbulence around the peaks makes up for it), perhaps score the top perpendicular to flow to create turbulence and dead-spots in the top don't matter.

[Graystar]

Originally posted by Cova
Why does everyone assume this block depends on laminar flow? It's really not all that small inside (total cross-section area), there is a very sharp bend upon the coolants entrance to the block where it slams into the base and then is pushed out of the way by more coolant coming down. Do you really think that the flow smooths out that much in the 1/2" it travels before it's over the core?

I guess because there's nothing specifically designed to create turbulence. I know that water never comes out of the block smoothly. Sputters all over the place.

[gone_fishin]

Originally posted by hara
Where did you get the data

He uses flowtherm or something similar. They are high priced engineering software packages. I doubt the technically inclined (les and BillA for example) will further this discussion since the attitude was get out. It's a shame that the only truly usefull piece of information posted in this thread is now even in question.

Remember the graphs show a trend not absolute values set in stone. The curves relation to each other should be noted. The bp thickness has less impact than the total area (as Les already stated).

[Graystar]

Originally posted by gone_fishin
He uses flowtherm or something similar. They are high priced engineering software packages. I doubt the technically inclined (les and BillA for example) will further this discussion since the attitude was get out. It's a shame that the only truly usefull piece of information posted in this thread is now even in question.

Remember the graphs show a trend not absolute values set in stone. The curves relation to each other should be noted. The bp thickness has less impact than the total area (as Les already stated).

I always wondered if it was valid to use the freeflow rates to predict flow in a restricted system.

[hara]

Originally posted by gone_fishin
I doubt the technically inclined (les and BillA for example) will further this discussion since the attitude was get out.

Why is this? I was just asking cause I'm interested in the subject.

[gone_fishin]

Originally posted by hara
Why is this? I was just asking cause I'm interested in the subject.

You have participated a lot in this thread so you must have read the part where the author told someone, quote:
"But if my process offends you then feel free to never post in this thread again. I assure you, we'll survive without you." end quote

Not the best way to present and seek knowledge.

[Les]

Originally posted by hara
Why is this? I was just asking cause I'm interested in the subject.

I did not reply because I thought my previous reply to Graystar covered. I am only a dabbler/dilettante and use free calculators/programs.
My reply to Greystar:-

"quote:
--------------------------------------------------------------------------------
Originally posted by Graystar

1)I'm sorry, but I just can't figure out what those charts are showing me. I don't know what "30x10x1.1mm" means.
2) The calculation for Convection Coefficient considers area only. 3)What's that third dimension? Is that the channel height? Is that implying that my channel should be taller?

4)Also, as Convection Coefficient is a property of the boundary, it is unaffected by base thickness, and can be considered a constant when trying to calculate the effects of base thickness. Therefore, I am confused by its appearance in the charts.

5)Finally, there is absolutely no explanation whatsoever as to how the C/W is calculated. "C/W" is Thermal Resistance. The thermal resistance through a solid increases with the length of the solid. That is a fundamental. In the matter at hand, base thickness is the length. For any given area, it is impossible for a solid with a longer length to have a lower thermal resistance than a solid with a shorter length. So I simply don't understand what that first chart is demonstrating.

6)I also have no idea what the second chart is showing, or why it's there.

Les, can you explain the charts and how you arrived at these numbers?
--------------------------------------------------------------------------------


1) "30x10x1.1mm" is a reference to the water chamber dimensions(lenght x width x height) of your wb.
2) The Convection Coefficient is calculated by Kryotherm*. Kryotherm uses the Sieder-Tate relationship. Have calculated at various flow rates for two Channel sizes(10x1.1mm and 40x0.8mm).It is plotted against flow-rate(LPM)
3) Third dimension ( "1.1mm and 0.8mm"?) is channel height. Not syuggestions on channel height made. Used 0.8mm for larger bp area(ed) wb since had calculations to hand.
4) Only plotted Convection Coefficient against flow-rate.
5) C/W(for 10x10mm heat-die) is calculated by equating the Thermal Conductance at the bp/water interface with the Film Coefficient in a Waterloo Spreading Resistance calculator** . The Thermal Conductance through the bp/water interface(Edit****) is taken as the reciprical of dimensionally corrected "Kryotherm K/W" for a 0.1mm bp. The "Total Resistance" from Waterloo is added to an estimate(0.1C/W)*** of the TIM Thermal Resistance to give a C/W value.
6) The second chart shows the predicted Pressure Drop(dP Total)) v Flow-rate (LPM) for the two considered wbs and th P/Q curves for three pumps. The intercepts of the pump and wb curves gives the predicted flow-rates - from which graph1 gives the predicted C/W for the different different pumps.


* http://www.kryotherm.ru/soft.htm
** http://www.mhtl.uwaterloo.ca/old/on...rce/strip2.html
*** From Billa [url}http://www.overclockers.com/articles654/index02.asp[/url] making an allowance of 0.05 C/W for die RTD offset.

**** EDIT: "through the bp/water interface" did read "of the wb". "

[Alchemy]

Originally posted by Les
2) The Convection Coefficient is calculated by Kryotherm*. Kryotherm uses the Sieder-Tate relationship. Have calculated at various flow rates for two Channel sizes(10x1.1mm and 40x0.8mm).It is plotted against flow-rate(LPM)

Heh. "High priced engineering software packages" said gone_fishin. Riiiight.

This goes against some things I said last year regarding BigBen's Radius block, but I would be extremely careful using a S-T correlation for a waterblock. It's a quasitheoretical relationship that is intended only for uniform flow fields. Inlet effects, macroeddies, stagnation, etc. for a block will probably drown out the effects of the linear flow field.

You can *maybe* get heat transfer coefficients within a degree of magnitude or two using S-T, so you can make very vague comparisons between designs as long as they are relatively simple. But if you want to know anything at all about the block performance, you really need a discrete numerical analysis.

Alchemy

[hara]

I refrained from replying to the thread earlier because I thought I was missing something.

I'm not understanding, what is so special about this block? :shrug:

The surface area to volume ratio is poor. There are no fins or channels inside the block. I'm afraid that comparing it to a comercially available block will yield results that will not be great at all. I think good aircooling would beat your block (by good I mean reeeeally good, and a bit noisy).

One might argue that this block is simple to make, but so is a #rotor block.

With a hacksaw and drill, I've demonstrated that a cathar clone could be easily done (will not perform as good as cathar's, obviously).

Graystar, with a flat design, you will hit the wall too early. Optimisation will only yield a couple more degrees. Listen to the experts, don't expect to know everything from the start, they know what they're saying.

[Les]

Originally posted by Alchemy

.........This goes against some things I said last year regarding BigBen's Radius block, but I would be extremely careful using a S-T correlation for a waterblock. .............

From
http://forums.procooling.com/vbb/sho...&threadid=5918

"Alchemy
Which wb's cooling charactaristics have you found to be describable by Sieder- Tate?

I have yet to employ Sieder-Tate successfully for any wb.
E.G.

Used Billa's data* and obtained Convection Coeff using my own Waterloo and Kryotherm Excel plots.
Dunno but possibly seeing a corruption of the Womac** correlation

* http://thermal-management-testing.com/white%20water.htm
** Referenced here: http://www.electronics-cooling.com/...001_may_a2.html "

The "Graystar wb" is a side-entry wb and maybe,just maybe, it can be described by Sider-Tate.
Likewise maybe the WW in side-entry configuration can be - would love to see a Billa test in this cofiguration.
Predicted diffences using my simple model* :-




* http://forum.oc-forums.com/vb/showth...hreadid=161563