Storm/G7 Prototype + Piccies

[bobo5195]

yeah i do believe there has been a mix up in terms mainly as the post was the result of coming out of the pub and been shouted at to come out of my room and go back to drinking. I assumed solid for some reason.

My theoretical muscles are tiny as a matter of fact. Just nice to talk about this stuff without it being a tutorial sheet from hell. Why i did engineering in the first place.

My argument is far from philosopical at all its basic engineering definition (although it is outside the level of this board a little bit). the thermal boundary layer size can be estimated by some formula. If you look at the wolfram page (on the first post here) the thermal boundary layer is ONLY CONDUCTION that is a basic defintion. Maybe a little jump in terminology but i think this board can take it. You assume (for right or wrong) that no convection occurs.

Your philosopical stuff is making my head hurt so ill leave it for the moment or till my insomnia wears off at least my gut feeling is its wrong but then im often wrong,

Oh and congrats on the design it looks really good pity i cant afford it as im a student. Never seen anything like that in the textbooks so you could be on to a winner.

[Cathar]

Hey, it's all good. Like all science once you dig down deep enough, it trascends the simplistic secondary-school-room level and becomes pseudo-philosophical.

What you're calling the thermal boundary layer, more commonly referred to as simply the "boundary layer", is a basic given in the convectional thermodynamics, being the boundary between static conduction of the liquid-metal interface and convective motion. Could argue for days over the strict definition of what's going on at this level with respect to potentially mobile liquid molecules interaction with the metal surface. Some trains of thought suggest that the molecules that touch don't move at all, while others suggest they do move, albeit very slowly, but can be effectively treated as if they were static for purposes of considering the thermal boundary transfer layer as being primarily, or solely, conductive.

This level of theory doesn't interest me too much though. I'll leave that to the university lecturers. I'm more interested in assessing gains and understanding the theory to a necessary point to predict whether some design change is going to have a desirable impact. I do my best to understand just as much as I need to in order to forge ahead, which is probably why I ended up dropping advanced physics by third year university because it got too hand-wavingly theoretical. I do admire the guys though who spend their lives breaking it all down and analysing it to the nth degree and formulating the theories for it, but I learned long ago that my interests were more founded in understanding the theory just deep enough to allow me to experiment to push things along.

What I think now may be beneficial is looking at some ways to further disrupt the boundary layer. There's been research done on pulsating flows which seem to suggest that this can work well, but the implementation of such within a waterblock is the tricky bit, not to mention the potential for additional stresses on the pump and other components.

Perhaps a spinning disc with a scimitar slit pattern located above the jet intakes providing a constant pressure drop resistance upstream to the pump, but giving localised pulsating flow input into the jets would be a way to implement such?

[bigben2k]

I can see "conduction" within the boundary layer; I would think that we can all agree on that one.

I'm at the point where I believe (along with Cathar apparently) that the next step up in performance is in the pumping action; I think we've exhausted mag drive solutions.

[bobo5195]

i know i said no more tonight but oh well.

We're engineers dont bring philosopy into this :P . At the end of the day though engineering becomes making up graphs out of nothing.

my definition is boundary layer is the fluid movement boundary layer, while the thermal boundary layer is outside/inside this as defined by the nuselt number? (too late at night to be sure and the nu was never explained to us as anything more than a hand wave)
Both are the boundary layer but since your talking about heat transfer striclty its better to keep them seperate for clarity.

Personally im less keen in performance but more in bring costs down alot and eeking out a little more heat transfer. The pulsing jet thing suggests to me costs (motors etc) but i was thinking you could just set up a fan to power something (under water fan connected to something to block off holes). This sounds like a gimmick and in away it is but as billA says blings sells. Im not sure what you mean by scimitar though, need a pen and paper really.

For disrupting the boundary layer roughness is always a good bet but that requires alot of science stuff i would of thought.

Oh and im like the A team if you can hire me as i dont have a 4th year project proper yet.

[Cathar]

No motors. Some angle on the edges of the slits would be enough to provide rotational motion as the water passes through.

Scimitar. Best way to describe it is to ask you to picture what shape a slit would have to be if it were of fixed width but has to provide a fixed time unit of opening when the disc rotates across the range of radii.

I believe in the Formula 1 style of advancement philosophy. Explore the limits at the outer edges of practicality and affordability, and then decide what can be folded back into a commercialisable low-cost (but lower performing) product, but still have that low cost product benefitting from the advanced work.

Of course, if you build a Formula 1 car, everyone wants one though. $300 seems extreme, but in relative terms there'd be very few people who wouldn't at least consider buying a Formula 1 car if it only cost 4-6x as much as a sporty family sedan, even if the track times don't improve by a great deal...

[mwolfman]

Originally Posted by Cathar

<0.1mm channels clog very easily despite best efforts

Since your block isn’t "normal", have you tested it to see what happens if you use a liquid with a lower viscosity, like alcohol?
The same goes for pumping power, since your block is restrictive, what gains do you get for mounting two pumps in series?

Originally Posted by Cathar

Perhaps a spinning disc with a scimitar slit pattern located above the jet intakes providing a constant pressure drop resistance upstream to the pump, but giving localised pulsating flow input into the jets would be a way to implement such?

I see what you mean, however the cost in pump pressure can be immense unless you just want to block 10 % of the jets or so. The problem is to make the disk spinning in a way that it don’t wobble or begins to “hum”…
In theory this will help move the water throw the narrow channels (the same method is used for making of steel wires, it helps the deformations process to go faster, about 20kHz pulses)

Originally Posted by bobo5195

We're engineers dont bring philosopy into this

Here I have to disagree; we do so, due to the fact that we can’t see what is actually happening in the block.
I've done loads of simulations that should contradict that, however my workstation needs about 125 GB more ram to go as close as I want and even then I'm far from the real truth in a block as the Storm (the resolution has to be extremely high compared to normal fluid calculations).

[Les]

Originally Posted by mwolfman

Since your block isn’t "normal", have you tested it to see what happens if you use a liquid with a lower viscosity, like alcohol?
.....

Some aspects of the influence of Viscosity are discussed, from Post 159 to Post227, here

[bobo5195]

engineering approximation are guesses they are not philosophy, philosophy is based on logical reason. When have you been able to see alot of the stuff in engineering. Most of materials science is based around guestimation of what you cant see.

You most certainly dont need 125gb of ram for this simulation eek. Then chem engers run full multiphase analysis on 80gb of ram. Run some simplifying assumptions. I know for a fact one of my lectures has run CFD on a water pump and i dont think he used a supercomputer to do it.
Aximsymetric flow at the outlet of one nozzle. This makes it 2d and you can add thousands of cells if you want your going to get near enough the same answer if you use the right cell geometry. Problems are going to occur when you get down to microscale roughness that cant be modeled but im guessing thats beyond your abilities leave the phds to it.

What program are you using to model stuff?

[Cathar]

Originally Posted by mwolfman

Since your block isn’t "normal", have you tested it to see what happens if you use a liquid with a lower viscosity, like alcohol?
The same goes for pumping power, since your block is restrictive, what gains do you get for mounting two pumps in series?

Alcohol (ethanol) has a higher viscosity than water for temperatures above about 10C.

Acetone is one such liquid which is significantly less viscous than water, while apparantly having fairly decent thermal conductivity (as liquids go). Haven't been able to track down large quantities of acetone though, but then again I can't say that I've tried that hard. I'm sure the local trade-supply paint shop would be able to get some for me in 10litres quantities with little trouble.

The blocks are of middling/moderate restriction only, and I mean that in terms of comparing to some of the various super-restrictive euro-blocks, and in comparison to the more open-flow US designs.

I don't design blocks specifically for high pumping powers though, and I certainly did not do so with the G7. This is all carefully balanced through base-plate thickness choice. I actually chose a base-plate thickness for the G7 that best suits use with something like a single Laing D5 (~2.5W of hydraulic power acting on the block directly in a loop consisting of the block, pump and a radiator). Once you get to Laing D5 levels the gains beyond that are in the order of a 0.7C improvement per 100W by the time you get to 6W of applied hydraulic power.

It's much like any block, the more flow you push the better it performs. The issue eventually becomes one of managing the pump heat dump though.

[mwolfman]

Originally Posted by bobo5195

engineering approximation are guesses they are not philosophy, philosophy is based on logical reason.

…and in fluid dynamics combined with heat transfer the line is very narrow. (haven’t you started to make your approximations with logic)

Originally Posted by bobo5195

You most certainly dont need 125gb of ram for this simulation eek. Then chem engers run full multiphase analysis on 80gb of ram. RUn some simplifying assumptions. Aximsymetric flow at the outlet of one nozzle. This makes it 2d and you can add thousands of cells if you want your going to get near enough the same answer if you use the right cell geometry. Problems are going to occur when you get down to microscale roughness that cant be modelled…

Nope, look at the thumb... simulation of an entire block…
Why simulate the entire block? In the Storm block the mass flow isn’t constant throw all the channels, the amount of heat that each jet varies a lot, the amount of water surrounding the exit from each cavity, even the density will differ between different jets (due to different water temperature and pressure)…
The assumptions that you want to make will hopefully tell you in witch direction that a small change in a
And I'm down at micro scale, and I think Cathar is in the same region (0,45mm)…

Originally Posted by bobo5195

...but im guessing thats beyond your abilities leave the phds to it.

You are only a second year student… consequently, EOD…

Les: Thanks for that link, my concern was rather on the fact that Cathar uses 0.45 mm hols, that means that the amount of wall friction will be huge compared to a normal block, an empirical test should easily provide with an answer… Since we are talking about minor gains, even a gain in 5% of cooling performance would be considered a significant gain. The question should rather be, what’s the optimal cooling medium for the G7? In Cathar’s (and Swiftech)case this could lead to a cooling additive that should be mixture with battery water…

[mwolfman]

Originally Posted by Cathar

Alcohol (ethanol) has a higher viscosity than water for temperatures above about 10C.

Sorry, dident have the books in hand...

Originally Posted by Cathar

Acetone

And will probably melt most of the sealings...

I got the books in hand now...
I find two liquids interesting
1. Methanol (CH3OH), 0,30 (10^-3Ns/m^2), 0,023 (N/m), 2,50 (kJ/(kg*K))
2. Dietyleter (C2H5)2O, 0,23 (10^-3Ns/m^2), 0,017 (N/m), 2,30 (kJ/(kg*K))
As reference:
Water, 1,00 (10^-3Ns/m^2), 0,073 (N/m), 4,18 (kJ/(kg*K))

Originally Posted by Cathar

The blocks are of middling/moderate restriction only, and I mean that in terms of comparing to some of the various super-restrictive euro-blocks, and in comparison to the more open-flow US designs.

The thing is that most of your pump losses happen in the channels towards the cooling block

[bobo5195]

I'm a 4th year student, one year away from a phd and my specialisations have so far been into CFD/FEA and manufactoring technology. My 4th year project is on comparisions of two FEA packages for use in the FEA course.

Modeling the entire block is exceptionally wasteful on resources, im not sure if such a model would complete is sufficent time (less than a week on HPC grade systems). 2d approach gives you approximations which can then be used in a full scale model, without a large reduction in accuracy, such an analysis could be completed in a day on desktop grade hardware. EFD zooming and mesh refinement at the nozzle exits may help in your approach as each nozzle is essentially the same equations and the only the area around the nozzle exit is complicated fully turblent flow while in other regions the flow could be modeled as your standard incompressible fluid, using possbily different cell types.

[Les]

Some data, from Incoherent(who else!), using an ethylene-glycol additive in a SwiftechMCW6000.
A beermat model is included in the subsequent discussion

[Cathar]

Originally Posted by mwolfman

The thing is that most of your pump losses happen in the channels towards the cooling block

Rough breakdown of total pressure-drop of block as percentage for each section:

The following are all calculated pressure drops for a certain flow rate, and applied as a percentage of the total pressure drop of the block at that flow rate:

2.9% due to tubing->barb transition
0.15% due to barb length traversal
-1.1% due to plenum spread (pressure gain due to sudden expansion)
2.8% due to plenum->jet intake transition
60% due to jet acceleration stage (constriction transition into jet tube proper)
10% due to jet tube length friction

From here on with the remaining 25% of the block's pressure drop it gets a little murky as to the exact breakdown of the following, and these are just my estimates:

~15% total pressure drop due to jet exhaust back-pressure due to proximity of jet nozzle discharge to base-plate (best practical example of why can be felt when holding a garden hose at full blast and bringing it close to a wall - you can feel the hose pushing itself away from the wall) - basically this is the focused jet impingement pressure-drop cost
~1% due to discharge of flow from cup (squishing between cup and jet tube walls - includes transition from cup base up past bottom of jet tube)
~9% total pressure drop due to exit/discharge of water-flow through jet nozzle array and out the discharge ports to the exit barb

Of all the various pressure drop transition points in the block, I don't feel that there's a whole lot more that could be done, even across a selection of points, to significantly improve the overall picture. I broke down every transitional point and optimised every one that I could in the G7.

[Les]

Originally Posted by mwolfman

The thing is that most of your pump losses happen in the channels towards the cooling block

As I belatedly accepted here , reduced pressure drops are an important feature of the Storm design.
Guess, have been further optinised in progession from G4 to G7.

[mwolfman]

Originally Posted by Les

Some data, from Incoherent(who else!), using an ethylene-glycol additive in a SwiftechThanx for tMCW6000.
A beermat model is included in the subsequent discussion

Thanks for these links as well… And for the mixture well thank god I'm not using glycol in my system (using DD super cool, about 3%)

Originally Posted by Cathar

60% due to jet acceleration stage
10% due to jet tube length friction

I got about the same results for the acceleration stage, however I expected a higher friction in the jet tube…
I might do some test on the G7 when I have some spare time (and the license server must be running…)

OFF TOPIC!{ from http://forums.procooling.com/vbb/sho...5&page=2&pp=25

Originally Posted by Cathar

I'm not particularly fond of GPU blocks due to them being so fiddly and non-standard.

If the new ATi card (580 if I'm not mistaken) has the same mountings as the older ones, then there shouldn’t be a problem with the “non-standard”, you just simply do a ATi block…}

[GlassMan]

If the new ATi card (580 if I'm not mistaken) has the same mountings as the older ones, then there shouldn’t be a problem with the “non-standard”, you just simply do a ATi block…}

Hey, how about those of us that plan on getting an authentic (and higher performing) GPU.?
Seriously, although there are many variations, the mounting scenarios are, (like cpu's) variations on a theme. Memory can be ignored as ddr3 runs very cool.
If you want a challenge Cathar, how about a Universal nF4 ultra-sli chip cooler. An effective design that can fit all the boards available will keep someone occupied a long time.
I know water cooling the north bridge is passe, but it may be time to be reconsidered. My chip (probe to side ) hovers at 50C moderately (255, 3x) OCed, the board sensor reads mid-30's, which is misleading, to say the least.
Of course if you are doing a graphics coooler, the nF4's will be obsolete, but the nF5 might need more help!!

[mwolfman]

Originally Posted by GlassMan

Hey, how about those of us that plan on getting an authentic (and higher performing) GPU.?
Seriously, although there are many variations, the mounting scenarios are, (like cpu's) variations on a theme. Memory can be ignored as ddr3 runs very cool.

The reason why I ask is due to the fact that my Sapphire card sounds like a electric grass trimmer...
I can’t say that DDR3 is cool, I have active cooling (air), and they definitely wouldn’t survive without that... (45-50° after 20 min in DoW)...

They (Watercool.de) use modules, one cooler but many diffrent configurations for diffrent cards...

[bobo5195]

got my online journal password working and typed stuff in, minutes later my brain stopped working due to the information.

Got a paper in the post which might as well be titled "how to design a storm block" aka heat transfer of confined jet impingment, to digest with my hot chocolate.

Some work suggests that you get alot better mass flow with an increased number of jets per block hole ie, have many nozzles per hole.

One paper im reading now did some analysis on oblique jets it took 6hrs to run an analysis on a single jet hitting a plate. It also showed an error of plus minus 7.4% so computational methos obviously arent going to help so much here

[Les]

Originally Posted by bobo5195

Some work suggests that you get alot better mass flow with an increased number of jets per block hole ie, have many nozzles per hole.

In beermat fashion was briefly considered here , with influence of no of jets on "direct cooling of a heat die" broached here

[Cathar]

Am not wholly convinced in the multiple jets/hole thing. May work well in theory or in limited practise but unconvinced about it scaling down, in a machinable sense, to the levels being dealt with here.

In my experience multiple jet arrays work well provided they are being driven with sufficient pressure, and by that I mean >>10PSI, to ensure sufficient jet integrity when the jet outwashes start interfering with each other. I played with multiple jet arrays with decent results when giving 20PSI, but down at 2PSI levels performance seemed to fall away more quickly than anticipated.

I wonder what pressures are being considered in the research paper?

[Eddy_EK]

Cathar, how did you test your G water blocks. I mean about the holes diameter in copper and jets holes.
I suppose you tried with different diameters and test them all to come to conclusion what seems to be the best compromise. :shrug:

I wonder how thick is the base in storm WB...

[bobo5195]

they werent concerned with jet back pressure there conditions were flow rate for a fixed heat transfer. They were concerned with angle of attack of the nozzle and nozzle shape as well. The idea is that the best heat transfer occurs in the center of the jet so using many jets increases the amount of high cooling areas as shown by how g1-.> g7. This increased cooling ability means you need less mass flow per watt disipated.

I'm goign to read more, but this is turning into another fun litrature review.

[Cathar]

Pressure and flow are linked. Can't have flow without pressure.

Okay, how about I ask it another way. What sort of jet sizes are they dealing with, and how much flow are we talking about?

The one jet-array paper that I saw was this one, but they were dealing with 59mH2O pressures and 237LPM flow rates across a 50x60mm area.

[bobo5195]

Originally Posted by Cathar

Pressure and flow are linked. Can't have flow without pressure.

Okay, how about I ask it another way. What sort of jet sizes are they dealing with, and how much flow are we talking about?

The one jet-array paper that I saw was this one, but they were dealing with ~50mH2O pressures and 330LPM flow rates across a 50x60mm area.

Its was numerical analysis and they didnt record the results. Also re reading the papers there reynolds numbers are alot higher than the ones for pc cooling so its less useful for us. The fluid was air which at high reynolds numbers is going to behave like water. Our low numbers of Re (under a 1000, using what should become know as the bobo equation for reynolds number in a round pipe with flow of Q lpm; Re = 21.2*Q/D) this not the case so all im after now is lamina jets.