Liquid Science: Idea

[ballz0r]

Ive been thinking that watercooling could get a bit more sciency...

By that I mean, there are basic principals of thermodynamics, fluid mechanics and basic electronics that every serious watercooler should know... for example, do you know how to read a pump curve? do you know what head loss is? do you know how to calculate whether the flow is likely to be turbulent or laminar, and even what turbulent or laminar flow means and how it relates to water cooling?

If every serious watercooler took the time to learn the science of watercooling we would get many more innovative ideas.

So... my idea is to run a series of threads on various engineering and science topics related to watercooling.... anyone can start one, just put Liquid Science: in the topic, that way anyone can use the search function to get a thorough grounding in the science behind our hobby.

What do you guys think about the idea?

[smee]

I say absolutely!!
I've really really always wanted to learn all the science to watercooling, and I agree we could get many more innovative ideas through it.

Lets do it!

[coolmiester]

Yeah good idea if you're that way inclined.

I personally don't know 90% of the above questions and care even less but it doesn't stop me from doing what i do and trying what i try, in fact sometimes i think if i did know the science behind some things i might not even try in the first place.

For me, the fun is in the trial and error and take that away and its not fun anymore like many big names have found and disappeared to do something else that's fun.

An example - i didn't like the way the GPU loop hose was routed on my last build so i literally reversed the loop and the not knowing was exiting as i didn't have a clue what was going to happen but i knew it would look better and as it turned out it did diddly squat to my temps but if i'd listened to the science i probably wouldn't of even tried.

Just my :2cents: but if that's what makes you tick then go for it.

[RedRaider]

I think it is a great idea, go for it...:up:

[rubidium]

Ive been thinking that watercooling could get a bit more sciency...
...
...
What do you guys think about the idea?

How can I argue with that? :rolleyes:

[HESmelaugh]

... for example, do you know how to read a pump curve? do you know what head loss is? do you know how to calculate whether the flow is likely to be turbulent or laminar, and even what turbulent or laminar flow means and how it relates to water cooling?


No, no, no, no and... no.

Well, I actually do have some idea of how to read those graphs and I read something about laminar vs. turbulent a while back, but my knowledge is far (faaaaar) from what you might call in-depth.

So I'm all for this idea! :up:

[bianco]

great idea.
i'd be quite interested!

[TedShred]

Great idea.

[RedRaider]

We have 2 engineers and a physicist, this should be good...:up:

I forgot, one of the engineers is an Aggie...:homo:

[mcoffey]

I'd like to see more info on that if it's very well explained in a simple way that makes the topic easy to understand.

I tend to be more practical and experience based myself. In other words, if it doesn't keep the CPU cooler, who cares. I don't have to know, or how to calculate any of the OP topics, to know what works best in regards to keeping a CPU cooler. If I willing to experiment with "hands on" application of the subjects, that's all it takes.

That being said, knowledge can't hurt, and nothing beats a well explained topic IMO.

andyc

[Snyxxx]

ballz0r, I have a background in fluid dynamics, so I will help if needed.

Also, our company does all kinds of pressure drop and flow assurance calcualtions to get the oil from one location to another. I can get help from experts if needed.

[HESmelaugh]

Couldn't have said it better than andyc just has.

I also work based on experience but if someone can make the physics behind it graspable then I'll be a very eager student.

[RedRaider]

Synxxx, you also have a background in Aggieland...

Texas Tech pwns Aggieland....

[coolmiester]

One thing that really makes me curious is heat dump from the pump and as far as i'm aware nobody has really tackled it in any depth so i wouldn't mind seeing the results from that as i have a feeling it plays a much bigger roll that what would be expected :)

And also trying to understand the flow "sweet spot" as again i think this will have a greater influence than you'd expect

[HESmelaugh]

Oh yes, heat dump from pumps would be very interesting!

I did a test once where I changed the System (Q6600, HD4850), from 8mm Tubing and an XSPC X2O 400 to 1/2" tubing and an MCP355 and while flow improved considerably, the CPU-temps only dropped by about .5K...
I'm pretty sure that this was in part because the MCP just runs much hotter than the X2O 400.

[Ricey]

Great if you can do that.

Having said that...

I myself have no patience for scientific data. I often ignore all the graphs and charts in reviews and just get to the crux of the matter.

All I want to know is what works. I rarely have the energy or interest in knowing why everything works.

[zlojack]

I'd be interested in reading this, as long as it's in practical terms as mcoffey says.

When things get too far into calculations and such, my eyes tend to glaze over.

[Ricey]

And its not that I don't understand the calculations. Its not about the math. Its about the applicability of theoretical concepts in empirical observation.

[Kayin]

I'm all for it, if you've read my articles on watercooling...

While not a true physicist or engineer, I've taught myself large amounts of the field in an attempt to design this stuff better.

[Snyxxx]

Synxxx, you also have a background in Aggieland...

Texas Tech pwns Aggieland....

Red, you know what they call an aggie in five years out of school...:D

I am thinking basic concepts would be useful such as pressure drop increrases as a square of flow velocity, the tug of war between flow and pressure drop.

[RedRaider]

BASIC CONCEPTS ?? :homo:

Go BIG or don't go at all....:king:

[andersson.j]

Great idea!

I'm on my third year out of five needed to become an engineer and I'll probably specialize in fluid dynamics, so this is right up my alley.

[CrazyJoe]

I am thinking basic concepts ...

The only thing to consider here then would be ... one mans "Basic Concept" is another mans "In-Depth Analysis".

It's all relative to what a person already knows, what they need to know, and what they want to know.

Still, I personally would love to learn everything I can ! :up:

[ballz0r]

great response to this guys...

The idea is to keep it short and sweet, only enough formulae and calculations to demonstrate how the variables affect performance.

For me science and engineering isnt about doing reams of calculations to come up with theoretical values, its more about understanding the behaviour of things. You'll find (hopefully) that an understanding of the basic principals will change the way you think when designing or troubleshooting your systems all without doing a single calculation.

I dont profess to be a brilliant engineer by any stretch. I studdied engineering 10 years ago and went straight into management. I really enjoyed materials science, fluids mech and thermo dynamics but it wasnt until I got into modding that I actually found a practical application for it :)

I'll start working on my first "Liquid Science" thread :up:

[link1896]

one thing I'd like people to walk away with a better understand of is how much heat we are trying to remove from an overclocked cpu.

we have flow estimators, we know c/w numbers for blocks and radiators, but people often forget, lets take a quad core, doing 4ghz, is dumping a LOT of heat.


On other forums, I've seen too many "my temps are bad when running stress prime" etc etc threads, and as soon as I introduce:

Overclocked Wattage = stock max power * ( OC MHz / Stock MHz) * (( OC Vcore / Stock Vcore )^2)


and I attempt to mention they are not at TDP but max power thanks to their four instances of some core loading app, the silence is normally deafening.

Without knowing our heat load, everything else is a guess IMO

[ballz0r]

Overclocked Wattage = stock max power * ( OC MHz / Stock MHz) * (( OC Vcore / Stock Vcore )^2)

Interesting formulae... Your basically scaling the heat output by the overclock.

is it validated by test results?

[rubidium]

Interesting formulae... Your basically scaling the heat output by the overclock.

is it validated by test results?

I would argue that this scaling relationship was validated experimentally well before any of us were born, when the first makeshift transistor - in the form of a tooth-size hunk of germanium (I guess one would call that centimeter technology :rofl:) - sat on a laboratory bench at Bell Labs.

Real transistors are close to being ideal switches, in the sense that when they are "on" that have very low resistance. and when they are "off" they have very high resistance. The most significant departure of a real transistor's behavior from that of an ideal switch occurs during transitions from off-to-on and from on-to-off. The power dissipated by any electrical component is proportional to the square of the voltage applied across it (thus the right side of the equation above posted by link1896), and inversely proportional to the component's resistance. For a transistor in the "on" or "off" state, the voltage and resistance combination is such that little power is dissipated. During off-on and on-off transitions, the voltage across the device and its resistance are momentarily such as to maximize power dissipation within the device during the transition period. In a processor, the average number of state transitions by the hundreds of millions (soon billions and 10's of billions) of transistors, and hence the average power dissipation, is proportional to clock speed. Thus the left side of the equation.

A related issue is the physical dimensions of individual transistors on the die. As they are made smaller, parasitic capacitances are reduced, allowing transistors to switch faster, spending less time in the power dissipating transition region between on and off states. This is why going from 95nm through 65nm to 45nm and smaller technology reduces power dissipation for a fixed component layout. Of course, as the technology gets smaller, designers are tempted to add more components to the chip, thereby offsetting the savings to some extent.

rubidium