anyone with fluid dynamics background? need some help with waterblock design

[tarui]

ok i am designing a waterblock. not too keen with the  current designs with parallel tubes and curved pipes.

Does the top part of the board need to be actively cooled as well? if so then i will extend the pipes there.

anyway

which flow design is better, what are it's properties like pressure loss, effects on flow rate, etc





for the 2nd pic, if it's better i intend to add fins along the way instead of just curved pipes.

[1715597629]

1715597629

[1715597629]

1715597629

[1715597629]

1715597629

[1715597629]

1715597629

[The Bitcoin Catalog]

Hey there,

Mechanical engineer with fluid dynamics background. There is no simple answer to your question it depends on too much factors. (i.e. pipe diameter, total length, fluid velocity, pipe material, etc..)
You should consider making finite elements analysis for optimal results. If you can't afford this, you'll have to make tests.
Here are the things you should consider:

Try to get as much as possible pipe contact with the heated area AND the cooling area.
Smaller diameter pipe combine with longer length configuration are often more optimal.
Try to get as much velocity as possible in the pipes. At a certain point, fluid is moving in a turbulence mode where heat transfer is optimal.

My personal advice would be the second option. The first one can create areas of stagnate fluid which can cause overheating on that area.

[crumbs]

Which board are you designing the block for?
As a place to start,  the parallel (ladder) approach is theoretically better, since the serial (snake) approach does not offer the same cooling on the outlet side as it does at the inlet (the water picks up heat as it travels through the snake).  It's also much more restrictive, which can be a problem unless you're using a true positive-displacement pump (you're not). The Bitcoin Catalog pointed out the possible problems with the parallel approach.  
If it helps, most high volume/high viscosity systems tend to be parallel (car radiators), while low volume/low viscosity serial (Air conditioner heat exchangers).
Have fun.

[dogie]

1) adds so much minor losses *read 'frictional losses due to geometry* and isn't viable
2) 2 is better, but again reduces flow quite a lot. The most commonly used layout in waterblock is a large cavity with cylindrical vertical pins. You want to maximise surface area WITHOUT reducing flow rate.

^^^Highly simplified because these are the only things you need to know.

[tarui]

Which board are you designing the block for?
As a place to start,  the parallel (ladder) approach is theoretically better, since the serial (snake) approach does not offer the same cooling on the outlet side as it does at the inlet (the water picks up heat as it travels through the snake).  It's also much more restrictive, which can be a problem unless you're using a true positive-displacement pump (you're not). The Bitcoin Catalog pointed out the possible problems with the parallel approach.  
If it helps, most high volume/high viscosity systems tend to be parallel (car radiators), while low volume/low viscosity serial (Air conditioner heat exchangers).
Have fun.

designing it to cool the avalons



ok where can i get rubber gaskets to fit the waterblock?

[crumbs]

Which board are you designing the block for?
As a place to start,  the parallel (ladder) approach is theoretically better, since the serial (snake) approach does not offer the same cooling on the outlet side as it does at the inlet (the water picks up heat as it travels through the snake).  It's also much more restrictive, which can be a problem unless you're using a true positive-displacement pump (you're not). The Bitcoin Catalog pointed out the possible problems with the parallel approach.  
If it helps, most high volume/high viscosity systems tend to be parallel (car radiators), while low volume/low viscosity serial (Air conditioner heat exchangers).
Have fun.

designing it to cool the avalons

https://bitcointalk.org/index.php?topic=211825.msg2221230#msg2221230

[tarui]

Hey there,

Mechanical engineer with fluid dynamics background. There is no simple answer to your question it depends on too much factors. (i.e. pipe diameter, total length, fluid velocity, pipe material, etc..)
You should consider making finite elements analysis for optimal results. If you can't afford this, you'll have to make tests.
Here are the things you should consider:

Try to get as much as possible pipe contact with the heated area AND the cooling area.
Smaller diameter pipe combine with longer length configuration are often more optimal.
Try to get as much velocity as possible in the pipes. At a certain point, fluid is moving in a turbulence mode where heat transfer is optimal.

My personal advice would be the second option. The first one can create areas of stagnate fluid which can cause overheating on that area.

what if the edges and corners are curved for first diagram?

[dogie]

what if the edges and corners are curved for first diagram?

Reduces minor losses slightly, but still pretty crappy.

[tarui]

anyway

having problems drawing just the basic diagram

bloody sketchup giving me so much problems/

thought i was done with the mountinh holes then i realised sketchup placed the holes floating in the middle of nowhere or it somehow thinks a column has a line that cuts it into 2

till this day they dont allow users to create a line or a circle by inputting the dimensions instead of dragging it to size.

[ultrix]

Water is so 1998.  You want a working fluid that will undergo a state transition as it effectively pumps itself in the right configuration and it captures more heat.   To do this with water it would take a fairly strong vacuum.   Maybe look into R134A or other HFC's?  There is a reason they tend to be the working fluid of choice for cooling operations.   Additionally, you can run all the fluid dynamics simulations through OpenFOAM, http://www.openfoam.com.

[tarui]

are there any issues with design a waterblock such that the rubber gasket/seal is above the edge of the asic chip?



the black area is part of the seal/gasket

red squares are the chips.

as u can see, the seal/gasket is "overlapping" the corners of the chips. will it cause problems?

[tarui]

does it matter if fins in the waterblock that faces the water are flat surface or round surface?