Watercooling Simulation

rudnik68 · Aug 22, 2004

Hey all,

Anyone interested in looking and playing with a WC simulation made in Excel? Right now its pretty basic, but if response is good, I can spend some more time. rightnow, the spreadsheet takes user data for the heat inputs and radiator geometry and iterates to find a steady state loop temperature.

I'm going to be changing to make drop down type selections for the CPU, GPU and NB, and can try to make a calculated heat trasnfer function for the CPU so that the actual die temperature can be estimated based on geometry and flow data. Granted, this uses alot of assumption, but it may be interesting to see what adding an extra 20W to a cooling setup will do to your water temperature.

PM me with an email addy if you are interested in trying.
Thanks,
Aaron

Guderian · Aug 22, 2004

You might get a bigger response on the procooling forums... then again, they usually like empirical evidence, rather than theoretical.

greenman100 · Aug 22, 2004

rudnik68 said:
Granted, this uses alot of assumption, but it may be interesting to see what adding an extra 20W to a cooling setup will do to your water temperature.

a lot of assumption means you're wasting your time

want to know what 20w will do? find out your rad's thermal resistance in C/W, multiply by 20, and you'll know how many C your temps will rise

rudnik68 · Aug 22, 2004

greenman100 said:
a lot of assumption means you're wasting your time

want to know what 20w will do? find out your rad's thermal resistance in C/W, multiply by 20, and you'll know how many C your temps will rise

Depends on your flowrate. At 1GPM 20W is less than a degree. At .1GPM is over a degree. I'm not assuming a radiator c/w tho - I'm dynamically modeling the convection coefficient by empirical correlations for heat trasnfer from a finned tube bundle.

I'm not planning this to be an absolute model - thats extremely difficult. What it can do, if accurate, is to compare changes in a watercooling loop relative to each other and to approximate the magnitude of some changes on temperature.

As a plus, it conclusively proves that low flow is bad for temperatures.

greenman100 · Aug 22, 2004

rudnik68 said:
Depends on your flowrate. At 1GPM 20W is less than a degree. At .1GPM is over a degree. I'm not assuming a radiator c/w tho - I'm dynamically modeling the convection coefficient by empirical correlations for heat trasnfer from a finned tube bundle.

I'm not planning this to be an absolute model - thats extremely difficult. What it can do, if accurate, is to compare changes in a watercooling loop relative to each other and to approximate the magnitude of some changes on temperature.

As a plus, it conclusively proves that low flow is bad for temperatures.

yes, depenedent on flowrate of liquid coolant and air, of coruse

low flow bad for temps has been proven time and time again

http://www.overclockers.com/articles778/

rudnik68 · Aug 22, 2004

Thats no surprise to me. I'm confortable with the physics of watercooling - but at least you can go from theory to numbers and have a dynamic model that will change temps based on coolant flow, pump heat, ambient air temp, radiator geometry and radiator CFM.

Call it a teaching too lfor people who aren't comfortable with watercooling physics.

greenman100 · Aug 22, 2004

yes

but where do you get the data!?!?

do not make assumptions or anything like that, it wil render your time worthless

Guderian · Aug 22, 2004

Not to rain on your parade...

If you did create a spreadsheet to do the calulations for you, assuming you made it detailed enough to be both accurate and useful, who would use it?

On the low end, you'll end up with people who can't or won't measure what needs to be measured, and the high end people will just do their own experiments.

If you came up with some default values for common items ( ie c/w of the jr120, watts added by a 1250) some people might get some use out of it. But the use of default values would skew all the results.

I like the idea of a simple to use "plug 'n chug" calculation spreadsheet, but I'm not certain you could make it both accurate and idiot proof.

rudnik68 · Aug 22, 2004

greenman100 said:
yes

but where do you get the data!?!?

do not make assumptions or anything like that, it wil render your time worthless

bah - there are assumptions in everything you do - but making good assumptions will not affect the outcome of a simulation. I haven't seen anyone ever mention thermal boundary layers in waterblocks - they exist, but are assumed to not matter. This is probably true - the fluid velocity inside the waterblock should be high enoughto minimize the boundary layers at the block. This is why impinging jets in waterblocks can increase heat trasnfer - the decreased boundary layer thickness results in higher heat trasnfer coefficients. One of the points I learned in school is that assumptions are not inherently bad.

This simulation uses a steady state assumption and iteratively calculates to find a steady state answer. It also assumes that the radiator tubes are in a particular gemoetry - round in shape, evenly spaced, in a line.

I don't calculate the die temperature- there are no assumptions for heat trasnfer in a thermal paste, waterblock geometry or secondary heat paths. At steady state, none of these matter - an overall energy balance for heat imput to the waterblock is all that is required. I'm not sure if anyone can measure the magnitude of that assumption - I have no idea how to determine the amount of heat lost to secondary heat paths.

I used correlations for radiator performance. Then I used an energy balance on any heat inputs in the loop (CPU, GPU, NB) to determine the fluid temperature change across the block. There is very little data involved, except for radiator geometry parameters which you can measure (ie tube diameter, fin spacing, fin thickness) and heat inputs from the processor which is listed in engineering data sheets.

rudnik68 · Aug 22, 2004

Guderian said:
Not to rain on your parade...

If you did create a spreadsheet to do the calulations for you, assuming you made it detailed enough to be both accurate and useful, who would use it?

On the low end, you'll end up with people who can't or won't measure what needs to be measured, and the high end people will just do their own experiments.

If you came up with some default values for common items ( ie c/w of the jr120, watts added by a 1250) some people might get some use out of it. But the use of default values would skew all the results.

I like the idea of a simple to use "plug 'n chug" calculation spreadsheet, but I'm not certain you could make it both accurate and idiot proof.

Most of the reason to do it was to prove that I can. I've seen a lot of discussion, but no results. who knows, maybe it can be useful - I don't know to who, but thats alright

I agree - its the accuracy that I will never have. I can't model secondary heat paths, waterblock heat trasnfer coefficients from block geometry and the like.

Its possible to use real data and select a block and model based on actual measured values, but that would mean measuring flowrates and heat trasnfer coefficients for a lot of waterblocks, radiators and the like.

Guderian · Aug 22, 2004

Most of the reason to do it was to prove that I can. I've seen a lot of discussion, but no results. who knows, maybe it can be useful - I don't know to who, but thats alright

If thats the reason, then I'd say do it. There might be some people out there who would get some use out of it, and if people just listen to the nay-sayers (myself included) nothing would ever get done.

Hell, at a minimum, someone would use it, post the results, have those results critiqued, and go back with a deeper understanding of it all.

Its possible to use real data and select a block and model based on actual measured values, but that would mean measuring flowrates and heat trasnfer coefficients for a lot of waterblocks, radiators and the like.

Each item, even the exact same product, will have different properties. A pair of eheim 1250's won't put the same number of Watts in the water. Thats the crux of the problem to me.

I'm not great with the formulas, but accuracy is only useful to a point: Dimishing returns and all. You might have a winner if you could pin-point where accuracy becomes excess and incorporate that.

greenman100 · Aug 22, 2004

rudnik68 said:
bah - there are assumptions in everything you do - but making good assumptions will not affect the outcome of a simulation. I haven't seen anyone ever mention thermal boundary layers in waterblocks - they exist, but are assumed to not matter. This is probably true - the fluid velocity inside the waterblock should be high enoughto minimize the boundary layers at the block. This is why impinging jets in waterblocks can increase heat trasnfer - the decreased boundary layer thickness results in higher heat trasnfer coefficients. One of the points I learned in school is that assumptions are not inherently bad.

This simulation uses a steady state assumption and iteratively calculates to find a steady state answer. It also assumes that the radiator tubes are in a particular gemoetry - round in shape, evenly spaced, in a line.

I don't calculate the die temperature- there are no assumptions for heat trasnfer in a thermal paste, waterblock geometry or secondary heat paths. At steady state, none of these matter - an overall energy balance for heat imput to the waterblock is all that is required. I'm not sure if anyone can measure the magnitude of that assumption - I have no idea how to determine the amount of heat lost to secondary heat paths.

I used correlations for radiator performance. Then I used an energy balance on any heat inputs in the loop (CPU, GPU, NB) to determine the fluid temperature change across the block. There is very little data involved, except for radiator geometry parameters which you can measure (ie tube diameter, fin spacing, fin thickness) and heat inputs from the processor which is listed in engineering data sheets.

You are attempting to build a simulator on theory

one must build a simulator on hard numbers.

the boundary layer is irrelevant.

how it works does not matter for a sim, only how well it works

and that is where good testing comes into play

rudnik68 · Aug 22, 2004

greenman100 said:
You are attempting to build a simulator on theory

one must build a simulator on hard numbers.

the boundary layer is irrelevant.

how it works does not matter for a sim, only how well it works

and that is where good testing comes into play

I agree - will you compare your actual values to the simulation?
Would take 10 minutes.

rudnik68 · Aug 22, 2004

Guderian said:
If thats the reason, then I'd say do it. There might be some people out there who would get some use out of it, and if people just listen to the nay-sayers (myself included) nothing would ever get done.

Hell, at a minimum, someone would use it, post the results, have those results critiqued, and go back with a deeper understanding of it all.

Each item, even the exact same product, will have different properties. A pair of eheim 1250's won't put the same number of Watts in the water. Thats the crux of the problem to me.

I'm not great with the formulas, but accuracy is only useful to a point: Dimishing returns and all. You might have a winner if you could pin-point where accuracy becomes excess and incorporate that.

Its done. Took 4 hours. PM me if you'd like to look at it.

Guderian · Aug 22, 2004

This might sound silly, but I don't have Excel, or any of the MS Office products.

squeakygeek · Aug 23, 2004

Guderian said:
This might sound silly, but I don't have Excel, or any of the MS Office products.

If this ends up being a useful tool, it would be more useful as a standalone program or web page.

greenman100 · Aug 23, 2004

rudnik68 said:
I agree - will you compare your actual values to the simulation?
Would take 10 minutes.

actual values measured by what?

the in socket thermistor?

what a joke

I'd like to see the result, only to see how it works, what numbers you guesstimated

do realize I am not trying to discourage you, only show reality.

Watercooling Simulation

rudnik68

Member

Guderian

Member

greenman100

Disabled

rudnik68

Member

greenman100

Disabled

rudnik68

Member

greenman100

Disabled

Guderian

Member

rudnik68

Member

rudnik68

Member

Guderian

Member

greenman100

Disabled

rudnik68

Member

rudnik68

Member

Guderian

Member

squeakygeek

Member

greenman100

Disabled

Similar threads