Need knowledgable advice

[shaft01]

Originally posted by 8-Ball
Shaft01

As Skulemate has said, this topic has been brought up over and over.

If you look through this thread, I think I did a pretty good job of explaining the situation. It's one of the things I just seem to be able to visualise perfectly in my head. I'm good like that.

::HERE::

If you're still unsure after reading this, ask and I'll try again.

But not today as I've got a load of work today, and unfortunately, this is something I don't seem to get! :shrug:

8ball

Ok......thanks for the reply.......per your statement in that other thread it looks like we -agree-

quote from 8-ball

""""I think what people are trying to envisagewhen they say fast through the block and slow through the rad is,

a. cold water into the block to maximise the temp difference between core/block and coolant.
b. lots of time in the rad so as to get rid of all the heat.

The best way to do this is to have either a wide radiator so it moves slowly through the rad, or a long radiator, so it moves at the same speed just in there for longer. Both of these mean having a bigger radiator. That's the only way of having each unit volume of water spend more time in the radiator for each unit of energy it picks up from the block.""""""

end quote

Based on your statement above, since the loops would be separate in this case with separate pumps (I assume) then adjusting the flow without affecting the water block would be easy.......

[bigben2k]

(Dang it, power glitch!)

Dave is right, I didn't take the timing of it all into consideration.

Someone recently posted a link to a thermal probe, with a NO (Normally Open) connection. If you stuck that in your res, you could have all PSUs shut down when the water temp reaches say 45 deg C.

[8-Ball]

Shaft01,

did you read my post?

If not, then please do and read the linked thread aswell.

If you have and don't get it, let me try and explain.


If you consider a single unit of water, 1ml for example, travelling through a radiator at X ml/m(millilitres per minute), it could transfer Y Joules of het to the air flowing through the radiator, agreed.

Now consider the same 1ml of water travelling through the same radiator, but this time at X/2 ml/m , ie half the flow rate. We can "assume" that this will now be able to transfer 2Y Joules of heat to the air, as it is in the radiator for twice the length of time.

Agreed

(NB I put "assume" in inverted commas as the actual heat transfered would be a little below 2Y. This is because heat flux is a function of temperature gradient. After losing Y Joules of heat in the first half of the radiator, the temp of the water will have dropped, lowering the temperature gradient, therefore lowering the flux of heat from the water to the air.)

Now here's the key part, taking each case in turn, how much heat will be dissipated for a constant flow of water.

Case 1

X ml of water per minute.
Each ml of water transfers Y Joules of heat to the air

Total Heat transferred per minute = X x Y = XY

Case 2

X/2 ml of water per minute.
Each ml of water transfers 2Y Joules of heat to the air

Total Heat transferred per minute = X/2 x 2Y = XY (2's cancel)

So would you agree that in a world where variation of temp gradient does not change the heat flux, and turbulence does not exist, both of these cases would have the same heat output.

Now, we all know that turbulence does play a part, and will allow the faster flowing water to dissipate a little more heat due to the break down of the planar flow at the interface.

Therefore, case 1 will actually be able to throw out X x (Y plus a little bit)

And,

As I said before, temp gradient does affect the heat flux so in fact, case 2 should dissipate X x (Y minus a little bit.

Ergo, by this logic, the argument of "keeping the water in the radiator for longer allows it to give out more heat" is not valid.

Someone please tell me if I'm spouting a load of horse sh*t

8ball

[8-Ball]

Sorry Shaft, it appears you have read the other thread,

though, I'm still not sure we agree, for reasons listed above.

Having the rad in a parallel loop with lower flow would be exactly as outlined above. where by lower flow does not necessarily mean more heat output, as you're moving less water through the rad.

Hope this makes sense

8ball

[myv65]

Airspirit,

A "throttling" valve isn't a specific type of valve, though I'm sure you've already found that out by now. It merely imposes a restriction that you can adjust. Look at it from an electronics point of view. At the point where you split the flow to go through the radiator and bypass line, the pressure will be the same. At the point the flows rejoin, the pressure will be the same. The valve is like a variable resistor. If you close the valve (infinite resistance), then all current goes through the radiator. If you open the valve fully, you will not have zero resistance, so a lot of flow will go through the bypass but some will still go through the radiator. What you want is a valve that is full port and stays in you position (percentage opening) you select.

To anyone doubting that slow flow through a radiator is bad, all you need to do is examine the extremes. At zero flow, there will be zero steady-state heat transfer. Slow flow is not the answer to better heat transfer in a radiator. Recognize that getting the water to the same temperature as the air is not the same thing as getting a lot of heat out of the water. Heat transferred is basically flow rate multiplied by delta-T in the water. Low flow means you need a heckuva lot of delta-T. Assuming you can get down to near ambient, this means that your inlet temperature (the stuff leaving your CPU) must get progressively warmer as the flowrate drops. As you increase flow rate, the delta-T in the fluid will decrease. In the real world, you reach a point where the added pump power required to get more flow through a radiator becomes counter-productive. Hence airspirit wanting the ability to artificially limit the flow rate. Honestly this approach will do very little to the effectiveness of the radiator, but by keeping the total resistance lower (combined resistance of rad + bypass line) the total system flow rate will be a little higher. This is a good thing for the blocks.

[bigben2k]

Originally posted by myv65
There is no reason you require a second pump simply to run a parallel loop for a radiator. All you need is an ample pump and a throttling valve in the "bypass" line. The throttling valve works a little nicer than separate pumps as you can then split the flow between the radiator loop and bypass in whatever ratio you wish. The downside is that you burn some energy in the pressure drop going over a valve, so overall efficiency takes a minor hit.

Actually, the reason for the second pump is to avoid having to use one massively large (and expensive) pump.

As I've stated above, that radiator is so restrictive, that even a Supreme 1200 can only push 120 gph through it, and that's at the cost of a "14 foot of water" pressure drop (!). Since that pump deadheads at 15 feet, it really doesn't leave any room for anything else!

BTW, that pump already costs $100. The 2400 model costs about $125. I think that's already a bigger dent in Airspirits budget than he expected (but hey, he got a nice rad for dirt cheap!).

[airspirit]

The cost of the pump is no big deal, honestly. If I'm going to support thousands of dollars of equipment on one loop, I want it as rock solid as I can get it. The deal with the 1200 is that I want to use that crazy flow for the blocks, not the rad. By using a bypass, I can let quite a bit of that flow through, and tune the system to maximum efficiency. I suspect I can keep all of my systems (since mine peaks at 46C now) under 50C in this setup, if I had four on it. The only real restrictions (I'm going to carve the PVC channels and 90s out of 1.5" PVC to reduce flow issues) will be the waterblocks, the sheer height and length of the tubing, and the rad. By lessening the impact of the rad, I can save more fun for the blocks.

[airspirit]

As I stated in another thread when this project started to eat away at me, I fully expect to spend up to $1000 on this ... I'm only up to about $250 at this point. $100 for a pump is chump change in the big scheme of things. I'm more pi$$ed about the caravels costing so damn much ($70 after shipping, but oh-so-sweet).

[Alchemy]

Originally posted by myv65
Airspirit,

A "throttling" valve isn't a specific type of valve, though I'm sure you've already found that out by now. It merely imposes a restriction that you can adjust.

Actually, *all* valves do this.

airspirit, the idea is pretty clever and certainly has potential. But my concern is how you plan to account for the temperature variations in the reservoir. You want to make sure the rad is drawing off the hottest water and returning it, while the line(s) to the CPUs is taking the coldest water. And unless you're using very high flow rates or some way to mix up the stuff in the reservoir constantly, you could see differences of several degrees from one point to another.

Alchemy

[shaft01]

Originally posted by 8-Ball
Sorry Shaft, it appears you have read the other thread,

though, I'm still not sure we agree, for reasons listed above.

Having the rad in a parallel loop with lower flow would be exactly as outlined above. where by lower flow does not necessarily mean more heat output, as you're moving less water through the rad.

Hope this makes sense

8ball

I agree that there is a point in reducing flow in a rad makes things start heading in the other direction......ie heat transfer is reduced.

However the point I am trying (actually asking about) to make is that since we are dealing with metal to water heat transfer in a water block and water to metal to air transfer in a radiator would you not agree that while a happy medium (sweet spot) can be reached regarding velocity between the block and the rad that due to the principals involved regarding heat transfer that a block is more efficient at higher velocities while a radiator is more efficient at a lower velocity than you would have in the block?

Would not having separate loops for each allow you to tune each to maximum potential? Also, would you not find that the happiest velocity for the radiator would be lower than the one for the block?

[bigben2k]

Actually, if you want to introduce the airflow as a factor, you really can't compare it to a waterblock anymore.

The fluid velocity is going to depend on the size of the opening. In a rad (heatercore), multiple passes are common, making them less restrictive. A waterblock is most often the biggest restriction (Airspirit is now the exception!).

You're right though, a rad will perform very nicely at a low flow rate, lower than a block. Why would you want to reduce the efficiency of a rad though?

[airspirit]

shaft01:

The point of this project is to have one WC system for multiple computers in a rack style setup. I want to only monitor one pump, one rad, one set of fans, and one flow loop. Naturally, the hotter the machine, the closer to the bottom it'll be (since the lower on the stack the faster the flow if everything else is the same due to less restriction). I don't want to have a super-tuned machine ... I've given up on the whole OC to OC aspect, however entertaining it may be. I'm going for a decent overclock on most of my machines, and I will leave the crazyness for whichever hobby box I DON'T have on the rack. This rack is being designed for nothing more than stability and reliability. Most people don't think a WC system can be as reliable as a standard air system, and I want to prove that notion wrong. With the heat of that back room in the summer, I was having problems with overheating on my Athlon 700, not to mention my old 1700+ (may it rest in peace), and I don't need to add more boxes to the room and watch it get worse. The easy solution would be to add AC, but that isn't an option I like. I'd rather set up the hardware in a way that is bulletproof ... plus, this is going to be a hell of a lot of fun. I might even see if I can get some of the are locals (CDA, Spokane, Pullman, etc.) together when I do the final construction, since this will be the only time they'll get to go to these lengths of insanity ... plus, the added experience will make the final construction turn out better. Once I've got it together, I never want to have to rearrange it again. I want to be (*gasp*!!!) DONE.

[8-Ball]

Shaft, I can understand where you're coming from, though I think you're argument is still flawed. I shall try and explain.

I have shown above (at least I think I have) that higher flow would give you greater capacity to remove heat from the water stream (NOT per unit water) per minute. As such, much heat will be transferred into the fins of the radiator. Now, provided you are still blowing air of the same temperature through, the Temperature difference between the air and the radiator fins will be greater. This in turn will create a larger temperature gradient which will induce a greater flux of heat from the rad to the air.

You have to try and visualise the whole system, rather than individual packets of water.

If you're still unsure, just keep firing questions, cos I'm pretty sure you won't be the only one. It might develop into a nice thread to refer people to.

Airspirit, would you like me to conitnue this in a new thread, as we seem to have hijacked your discussion?

8-ball

[shaft01]

Originally posted by bigben2k
Actually, if you want to introduce the airflow as a factor, you really can't compare it to a waterblock anymore.

The fluid velocity is going to depend on the size of the opening. In a rad (heatercore), multiple passes are common, making them less restrictive. A waterblock is most often the biggest restriction (Airspirit is now the exception!).

You're right though, a rad will perform very nicely at a low flow rate, lower than a block. Why would you want to reduce the efficiency of a rad though?

For the purposes of this discussion the airflow can be a constant......non-varying. If I read you right you are saying his Lytron is more restrictive than 3 to 4 water blocks? THat makes me wonder how restrictive my micro-channel radiator is.......14 channels per flat tube in this one.......Made by modine......16x9x3 inches. Full load temps on a 2200+ @2100 mhz with a 226w pelt range from 0c to 3c depending on ambient. Two 172mm 24v rotrons @12v = quiet



And all setup.....

.

As to reducing the efficiency of the radiator I dont see it that way......I am saying the most efficient velocity for a radiator is slower than the most efficient velocity for a water block. Unfortunantly I dont have the math/thermodynamics background to prove it on paper (or for that matter follow some of you that do) but I do have the findings of my own trial and error results that lead me to believe I am right.

[shaft01]

Originally posted by airspirit
shaft01:

The point of this project is to have one WC system for multiple computers in a rack style setup. I want to only monitor one pump, one rad, one set of fans, and one flow loop.

I can see the advantages of keeping it simple......primarily the less components the less likelyhood of problems.....

Sounds a a MAJORLY FUN project........Maybe you should look into an Iwaki MD70RXT pump......You can find them on Ebay all the time, high flow, high head, reasonably quiet and VERY reliable and dont introduce much heat to the system. I recently sold two of them in new condition for $125 each......I dont recall the GPH but I think it was over 1000 GPH.......

BTW sorry didnt mean to highjack you .......We can certainly move the discussion to another thread if you like.......

[bigben2k]

You use the term "most efficient velocity" when really there isn't one. There is however a "point of diminishing return". Maybe that's what you mean?

Theoretically, the heat dissipation would increase indefinitely, along with the flow rate. There is a point however, where a very high flow rate doesn't provide much more cooling, than it's slightly slower counterpart.

Then there's the cost of putting a high flow rate through a rad (expensive pump), and there's also physical limits (pipe burst), but that last one is way up there.

Heatercores I found, are tested to about 30 psi. You certainly can't count on even half of that, if it's used.

I don't know about your radiator. From your measurements, it would be comparable to the Lytron 6220 (copper, 4220 in SS), but the Lytron looks like a single pass design, where yours looks more like a 14 pass design.

The 6220 is by far the most restrictive of the Lytron copper rads. If yours is multipass, then you could probably compare the thermal properties, with some adjustment in the numbers.

[shaft01]

Originally posted by bigben2k
You use the term "most efficient velocity" when really there isn't one. There is however a "point of diminishing return". Maybe that's what you mean?

Theoretically, the heat dissipation would increase indefinitely, along with the flow rate. There is a point however, where a very high flow rate doesn't provide much more cooling, than it's slightly slower counterpart.

I don't know about your radiator. From your measurements, it would be comparable to the Lytron 6220 (copper, 4220 in SS), but the Lytron looks like a single pass design, where yours looks more like a 14 pass design.

The 6220 is by far the most restrictive of the Lytron copper rads. If yours is multipass, then you could probably compare the thermal properties, with some adjustment in the numbers.

I think there is light at the end of the tunnel now.......so you are saying that the same exact principals apply regarding the convection of the heat out of the water into the air even though it is a slower process due to air being a much less effiicient conductor.......I can buy that.......However that still leads me to the question of the time factor...... due to the air being less efficient at removing the heat than water is would it not hold true that the longer a given volume of water resides in the cooling medium (ie radiator, within reason of course) the more heat will be convected out? Hence multipass radiators are inheirently better since it takes longer for that given volume of water to make the trip thru?........

I really dont know if my modine is multipass or not......what I do know is the flat tubes when looked at in cross-section view consist of 14 small channels inside resulting in greater surface contact area.......

[myv65]

OK, I *do* have the background and I'll try to give this one a final shot.

Ask yourself a question. When I stand in cold air or jump into cold water, do I feel colder if the air/water is still or blowing/flowing past me? The issue in question is one of convection and the medium, whether air, water, or any other fluid/gas, follows the same rules.

The confusion in radiators seems to stem from the fact that there are three main heat flows that must occur. Note that all three *must* transfer the same amount of energy, as energy can't simply continue to build up indefinitely. OK, so the first is convection from fluid to tube. The second is conduction from tube to fins. The third is convection from fins to air.

Look at these items individually. If you want to get heat from a liquid to a solid, do you want high fluid velocity or slow? The answer is high. In getting the heat from one area of a solid to another, do you want high conductivity or low? The answer is high (aluminum falls behind copper). In getting heat from a solid back to air, do you want high velocity air or low? The answer is again high.

Lots of confusion seems to come from the fact that the air has so much lower density (hence lower convective properties) that people seem to think that keeping the fluid around longer gives the air a greater chance. Fact is that the only thing tying all these things together is that a uniform amount of heat must get through all the barriers. The air really has no idea how fast the fluid is travelling through the tubes. Likewise the fluid has no idea how fast the air is travelling. All the water knows is that it requires a certain delta-T from it to the tube based upon flow rate and heat load. Higher flow rate equates to lower delta-T.

What further muddies the water (sorry for the pun) is that there are at least two other things that can occur here. The first is internal flow enhancements within the tubing. These break up boundary layers to enhance convection. There is evidence in BillA's testing to show that these indeed may have a "bump" in heat transfer versus flow velocity. Note that as a whole, heat transfer goes up with flow, but some radiators have a bump in the curve that I can't personally explain definitively. Second is that it takes energy to drive flow through a radiator. This means that a radiator must get rid of both the thermal energy from the fluid as well as the energy dissipated in the form of a pressure drop over the radiator. This latter term is extremely small for most cases; however, if you start putting massively powerful pumps onto systems that only have a CPU heating the water, then the pump energy can become a dominant contributor. In these rare, rare cases, you can see an increase in CPU temperature in spite of increasing flow. By this time, the pump is probably dumping 200 watts into a system heated by a 75 watt CPU.

Helpful at all?

[airspirit]

Let me lay some groundwork here that the thermodynamics and hands on people both can appreciate. My radiator provides little extra cooling between 2GPM and 4GPM. This means that after 2GPM, I've hit the point of diminishing returns. There is no real tangible advantage to pumping faster than that. This equates to 4xCPU blocks at .5GPM which is sh!t.

You will also notice that as you go from 2GPM to 4GPM the resistance provided by the radiator goes up dramatically. Since there is no real benefit to the heat transfer ability, there is no sense forcing flow through it.

By allowing a bypass, I will tune my system to allow for higher overall GPM in the system, to let the blocks move more efficiency. Since I've already tuned the radiator to the point of diminishing returns (2-2.5 GPM), I might as well allow the CPU blocks to work more effectively (more like 2 GPM EACH). That extra flow has to go somewhere, so I will need the bypass. What will happen in the system is that the overall water temperature will go up. This is a given, since the water will not all be circulating through the radiator. This will also raise the idle temps of all the CPUs in the array. This will also, since the temp. differential is going to be higher between ambient and the coolant, allow the radiator to be more efficient at its point of diminishing returns, flow-wise. Since the water will also act as a large buffer (there will be a total of over 5 gallons in the entire system, maybe more if I thread the internals of the cooling box exclusively with 1.5" PVC like I'm planning), and the higher flow through the blocks themselves make the BLOCKS more efficient at their jobs, I suspect the peak temps of all the CPUs will be lower than they would be if the rad was exclusively in series.

In the end my the dT between idle and peak will be lower, with the idle higher and the peak lower. Also, with the sheer water mass in the system, I could crank for quite a while before my water temps will rise appreciably. I'm thinking of putting a digital temperature display on my box for the water temps (measured in the res), but I don't know how to do this without generating leaks. This would also help me "tune" the system to its max efficiency. I suspect that in this particular case, with the original three CPUs on the array, that I will find peak efficiency with 3GPM through the rad and about 4GPM through the bypass. One thing to note is that I will probably use 3/8" blocks on my CPUs (Black Edge blocks from BECooling). Since the pump (1200) will have 3/4" output, I can use 4x3/8" easily.

[airspirit]

Oh, and if anything, this will put this whole argument to bed, since I have definitely decided to do this. The worst that'll happen is I'll just close the bypass permanently. It'll be easier to do that than do multiple retrofits of PVC (ugh).

I'm going to document the whole process (maybe Joe can post it as a WHAT NOT TO DO article), so you will all see the success/failure of it (I'm already starting the documentation process now ... some is at http://users.adelphia.net/~jnsholcomb/Page1.htm ). I also invite anyone in the Eastern WA/Northern ID area who wants to help to come over and put in their expertise when I assemble this. I'll supply the beer and pretzels, and we can have a jolly drunken time putting the pieces together. The piping will all be assembled and the various quick disconnect loops will be built, it'll just be a matter of seating the blocks, stacking the units, filling it, firing everything up, and doing some tuning on the "AQUABUS [tm]" up front.

[shaft01]

Originally posted by airspirit
Let me lay some groundwork here that the thermodynamics and hands on people both can appreciate. My radiator provides little extra cooling between 2GPM and 4GPM. This means that after 2GPM, I've hit the point of diminishing returns. There is no real tangible advantage to pumping faster than that. This equates to 4xCPU blocks at .5GPM which is sh!t.

Aha!.......and therein lies exactly what I have been trying to say.......Your radiators sweet spot is 2gpm.......and we all know that most if not all water blocks will perform their best at a good bit higher gpm rate.......

BTW I just bought one of those same radiators from the same guy you did tonight on Ebay.......I have not liked running the modine because its aluminum and I have a copper block. I have yet to see any adverse effects but I will feel better with copper

I wont tell ya what I won it for as it might tick you off

BTW folks there are two more of them listed for auction........

[airspirit]

Naw, what did you get it for? Oh, and those bastids are really slow to ship (a week and a half).

[shaft01]

Originally posted by airspirit
Naw, what did you get it for? Oh, and those bastids are really slow to ship (a week and a half).

51.00

He wont like the feedback he gets if it takes him that long on mine.........

[airspirit]

Maybe I need to get a second and lay the flow concerns to rest? Heh, my wife would shit her pants.

[UnloadeD]

why not place a small submersible pump in your res tuned for ya rad? this seems a much simpler solution to me.