Water blocks for amd 64

[BrianW]

Well then you would want a CPU block that would be able to keep flow non laminar at low flow. I imagine a grid array would be able to do this. Although it is obviously all conjecture.

BrianW

[Althornin]

Originally posted by superart
good point, good point, but look at it this way.

In a rad, the deltaT is not very large, in relation to the deltaT of a cooler or instance, so the flow rate of water really doesn't make much of a difference since high flow or low flow, the efficiency will be more or less the same. The reason you would want high flow on a rad setup is because it is better for your specific block, since some blocks perform better with higher flow.

In a cooler, however, the deltaT is much greater, therefore the more "time" the water spends in contact with the cooler, the longer it will be exposed to the "greater deltaT", the more efficient and effective the cooler will be. Because of this lower flow rate, you would probably want to pair the setup up with a block that performs better at lower flow rates.


Am i thinking along the right track?

I'd argue with a couple of things:
"some blocks perform better with higher flow" is false. All blocks perform better with higher flow, all other things remaining constant. (this is refering to cpu blocks, of course).

And remaining in the chiller for longer results in lowered efficiency, as delta T drops as more time is spent in chiller. Now, that said, if one imagines a loop with no heat source, and only a chiller block, slow flow through the chiller will deliver really cold water faster to the cpu block (assuming it is right after the chiller) than high flow, but high flow will chill ALL the water faster. At least, i think so.

I just dont see how lower flow can help - if X watts of heat can be removed from the water via the chiller per delta T per second, then maximizing the delta T maximizes efficiency. This is how water cooling works! Feel free to explain the wrongheadedness in this, because from a mathematical/heat transfer perspective, i simply dont see it.

Feel free to get as technical as you want, i am familiar with PDE and the general heat transfer equations.

[#Rotor]

The pin-grid is indeed the best known ,to me anyway, for chilling a sluge.... It is the reason Why I came to my design in the first place. The other benifits was just cherry on the cake,

do keep in mind, that Delta-T is inversely proportional to the time spent in any particular location.... thus, the longer the substance stay in that location, the smaller DT will become... this is true for any initial DT value, and therefor high flow will always get you better thermal transfer, by virtue of the fact that DT is maximized.
Do, however not link this, with the ability to sustain "high flow".... and it is In there that lies the reason why chillers tend to revert to a "low-flow" senario, only because it's more work to push a sluge through a pipe.... [M2C]

[redleader]

And remaining in the chiller for longer results in lowered efficiency, as delta T drops as more time is spent in chiller. Now, that said, if one imagines a loop with no heat source, and only a chiller block, slow flow through the chiller will deliver really cold water faster to the cpu block (assuming it is right after the chiller) than high flow, but high flow will chill ALL the water faster. At least, i think so.

The chiller is not a rad. Its an active heatpump. If you leave water in it longer, it will chill that water cooler and then reduce its temp as well. In theory at least, since TEC performance doesn't vary vastly over the narrow range of temps we're considering, the deltaT is the same no matter the flow (within reasonable flows that is).

[#Rotor]

Actually a chiller is a radiator.... Just that a chiller has external propulsion to enhance the thermal capacity and lower the operating temperature at which it still can opperate with a DT. A radiator is reliant on the temperature of the air being moved through it, for it's operational temperature. Both are, in principle exactly the same thing, they move heat from one substance, to another... In fact... all the heat transfer devices in a system work on exactly the same set of rules, they are all radiators or "heat transducers" as I like too call them....

[Althornin]

Originally posted by redleader
The chiller is not a rad. Its an active heatpump. If you leave water in it longer, it will chill that water cooler and then reduce its temp as well. In theory at least, since TEC performance doesn't vary vastly over the narrow range of temps we're considering, the deltaT is the same no matter the flow (within reasonable flows that is).

wrong.
Heat transfer follows some basic rules - the chiller cools the waterblock, which cools the water. Delta T matters for this thermal interface. Period.

[superart]

superart is on the right track, as are several others

good, glad to know that I haven't gone completely mad......only partially

[BillA]

lol, all kinds of wierd science goin' on here

Althornin
focus on the attainable temp, not the efficency

#Rotor
please; wbs, rads, and 'chilling chambers' are all heat exchangers

[Althornin]

Originally posted by unregistered
lol, all kinds of wierd science goin' on here

Althornin
focus on the attainable temp, not the efficency

I dont see how it will achieve lower temps.
Pls explain.

If at some deltaT, the chiller is able to "extract" say, 100 watts out of the water, and the cpu puts in say, 80 watts, then the water will eventually cool down to a lower deltaT and a rough equilibrum (swings lower thru chiller, then high again thru waterblock). slower flow will result in lower water temps in the chiller, but higher water temps over the waterblock (more time to absorb heat). I'm just not sure what gains you could see, if any, especially due to the lowered Waterblock efficiency at lower flow. I'd take some real convincing that you'd see a good difference, and i'd tend to believe that the solution is sub-optimal anyways, and that a higer flow, higher efficiency setup could be created that would "win".

[BillA]

keep going
and . . . . .

so the temps at flow rate X are A and B (coolant and cpu)
what will the temps be at flow rate 2X ?
and at 4X ?

the cpu/coolant gradient (°C/W) will be less; due to 'h' generally speaking;
but this is more than offset by the rise in the coolant temp

all heat exchangers have a limiting function (one side/conditions wrt the other side/conditions),
in this case it is the extraction of the heat from the 'cold' coolant
(do not ask me why it is more difficult to cool than to heat, I do not know - anyone ?)
- just as with liquid/air radiators it is the air side that is limiting
(which is why slowing the coolant down can result in lower temps, though in a 'big' rad such is not perceptible; with two smaller rads in parallel it is quite apparent)

as to the 'real convincing',
apparently you are not aware of the very long history that TEC chillers have in the overclocking world
- with uniformly poor results

my humor, eh ?
this course is called: How Things Work

[pHaestus]

Received it yesterday. Some issues with packaging. Styrofoam made its way into all of the barbs. I will be forced to completely disassemble the unit to remove it.

[BillA]

crap !
sorry
would suggest blowing out with air, then flushing

you will have little success reassembling it (gasketing and torque setting)

if it cannot be flushed suggest sending it back and let us do it

damn

yours was the first 'retail' packaging as we have been shipping these in cases of 20
another lesson

[pHaestus]

Tape over the barbs or better yet parafilm would be suggested. I don't remember off the top of my head if there was styrofoam shavings in the smaller central barbs or if it is isolated to the 1/2" barbs on the outer blocks. Those are easy enough to deal with; the central chamber was my concern.

[cristoff]

I've always kinda figured it this way....

(Just side thoughts. IF you stick your hand in the water and take it out quickly it might get a little colder and then get warmer again...
IF you stick you hand in the water and keep it there it might get colder but then after a while if the water is cold enough to overcome your warmth your hand will freeze, if not your hand will warm the water around it, by very little mind you.
IF you stick your and in the water and take it out quickly, and do so repeatedly, your hand will become cold if the water is cold enough to regenerate its "coldness" in that area, if not, your hand will become cool and overcome the coldness there and exhange it with heat... )

The problem is for each is that you have energy that is constantly replenishing itself... Therefore you must think of the exchanging as constant.

Becuase copper itself has a rate of exchange, the amount of exhange taken place will be affected by the design of the block and how much of the copper is in contact with the water and the chip. For the water has a rate of exchange as well, the amount of exchange determined by the flow and impingement.

Many people think of the water since its cold will cool the copper down, which in actuallity is sorta true. The water in which its being cold is drawing the heat away from the copper at a certain rate of exchange. Many people think of the copper as being warm and warming the water... Either way the water needs to get away from the copper... but it also needs to contact as much of the copper as possible. But the myth is that the water "needs" to stay in there to take the heat away for X amount of time in order to pull the heat away or cool the copper down. It doesnt need to although it might help, the water will just impinge the colder water coming in to exhange heat with the copper and thus make that water warm as well...

IN all actuallity, for me at least and my theories, is that the faster the flow and the least impingement, but the most contact with copper is the best idea. You need to have water block inside the same volume as a hose the same length....
IF you have more volume inside, then you have dead spots and spots that just take more heat.... but dont exit to give time for the other water entering to take its place....

Although all of this can be calculated to create pretty much the perfect water block....

I believe at least that a very thin base, with fins inside very much like a HS, with an inside volume the same as the tube the same length, with the fins over the whole chip only, will create almost the best environment for the chip to become cool...

How much wattage do the chips produce? I am trying to build a pelt system soon...

[Althornin]

Originally posted by unregistered

as to the 'real convincing',
apparently you are not aware of the very long history that TEC chillers have in the overclocking world
- with uniformly poor results

my humor, eh ?
this course is called: How Things Work

The uniformly por results are, imo, a result of poor efficiency waterblock usage.
Plenty of people have made the "water spends as long as possible int he block" approach, because it seems so logical if you dont know about heat transfer efficiencies.
Look, it may be that you can achieve, WITH YOUR WATERBLOCK (designed for low flow) better temps with low flow through the chiller, but you havent even come close to addressing the idea that low flow is the "key" to chiller performance. I may be wrong, and am of course thinking aobut this in my spare time here at work, but right now, i cant see why a system designed with waterblocks that work more efficiently at high flows might not easily be able to offset the resultant higher coolant temps. And the temps will be less higher than simple logic would dictate due to increased efficiency in heat transfer. You are right, i am sure, for your waterblock - but i still remain convinced that the solution is sub-optimal anyways, and that a higer flow, higher efficiency setup could be created that would "win".

cristoff:
"IN all actuallity, for me at least and my theories, is that the faster the flow and the least impingement, but the most contact with copper"
how do you think you get mroe contact with copper? Let me tell you, impingment is one way.....

[Les]

Originally posted by unregistered
lol, all kinds of wierd science goin' on here

I think we may be in complete agreement.
In fact you are being excessively polite.

[Since87]

Originally posted by Les
I think we may be in complete agreement.
In fact you are being excessively polite.

Ah, c'mon Les. It's your turn to be rude for a change.

[Zhentar]

Here's my impression on the issue...

We all know that higher flow in the waterblocks gives better performance. That will be true regardless of whether the coolant is 0C or 100C. Whats different here is the heat exchanger. Logic says we want the water to stay in the heat exchanger as long as possible to get the water cooled down as much as possible, and therefore want low flow rates. However, billA's testing shows that this isn't entirely true in heatercores as you will actually see an improvement in performance up to about 4 GPM thanks to the added turbulence at these flow rates exposing more water to the copper of the heatercore. But now we're not working with a heatercore, but instead a waterblock with peltiers on either side. This is a different environment with different characteristics, most notably the drastically reduced surface area, which has a new sweet spot; instead of the 4 GPM of the heatercore we have a much lower flow rate that billA is using here for a balance of the high flow the waterblock wants and the low flow the heat exchanger wants.

Hopefully there is some accuracy in all that. Some people (my physics teacher, the AP tests) tell me I know physics, but I know their full of shit and I actually have no idea whats going on.

[Since87]

Originally posted by Althornin
Look, it may be that you can achieve, WITH YOUR WATERBLOCK (designed for low flow) better temps with low flow through the chiller, but you havent even come close to addressing the idea that low flow is the "key" to chiller performance.

I'm thinking along similar lines.

I'm wondering if the attached picture doesn't somewhat describe the pattern of flow through the block at low vs high flowrates. (left and right respectively) Obviously the inlet and outlet on the chiller are different than on an MCW5002, but I'm just trying to convey a general idea.

Maybe such a change in the pattern of flow could be consistent with die simulator test results as well as chiller test result?

[Les]

Originally posted by Zhentar

...................

I actually have no idea whats going on.

Exactliy.

[Les]

Originally posted by Since87
I'm thinking along similar lines.

I'm wondering if the attached picture doesn't somewhat describe the pattern of flow through the block at low vs high flowrates. (left and right respectively) Obviously the inlet and outlet on the chiller are different than on an MCW5002, but I'm just trying to convey a general idea.

Maybe such a change in the pattern of flow could be consistent with die simulator test results as well as chiller test result?

Looks rubbish to me.
The science is somewhat iffy.

[BillA]

those without an 'historical' (sp ?) perspective of these devices may not appreciate this question, but

anyone, please:
provide a link to a successful TEC chiller that had a high flow rate in the chilled loop

now this is not for lack of trying, also by some clever minds

effort might be better spent determining why this scheme works,
than in an attempt to disprove it with words

Since87
I'v done a rather lot of work manipulating the flow path(s) in the MCW5000,
it will not change in that fashon, at least not with the flow rates we can create

[Althornin]

Originally posted by unregistered
those without an 'historical' (sp ?) perspective of these devices may not appreciate this question, but

anyone, please:
provide a link to a successful TEC chiller that had a high flow rate in the chilled loop

now this is not for lack of trying, also by some clever minds

effort might be better spent determining why this scheme works,
than in an attempt to disprove it with words

well, thats why i am posting here. I'd like an explaination, rather than vagueness. I have also said that i understand your point about lower water temps offsetting lower efficiencies - i just disagree that a high flow setup is crap.

Mind, just because it hasnt been done, doesnt mean that its wrong/impossible/etc.

And as for an example - nope, most TEC based chillers have fialed miserably. Of course, what is "high flow"? I'd like to see some more of cathars work with his "hydra" style pelt block - he was working on a pelt setup that i would call "high flow", compared to this swiftech model.

Ah yes - please stop assuming i have no "history" of following this, eh?

[Zhentar]

Les, I was hoping for a bit more, like why I'm wrong.

[BillA]

Originally posted by unregistered
. . . . .
so the temps at flow rate X are A and B (coolant and cpu)
what will the temps be at flow rate 2X ?
and at 4X ?

the cpu/coolant gradient (°C/W) will be less; due to 'h' generally speaking;
but this is more than offset by the rise in the coolant temp

all heat exchangers have a limiting function (one side/conditions wrt the other side/conditions),
in this case it is the extraction of the heat from the 'cold' coolant
(do not ask me why it is more difficult to cool than to heat, I do not know - anyone ?)
- just as with liquid/air radiators it is the air side that is limiting
(which is why slowing the coolant down can result in lower temps, though in a 'big' rad such is not perceptible; with two smaller rads in parallel it is quite apparent)
. . . . .

vagueness ??
no, I would say you did not think through the scenario I posed
don't get your panties bunched
several of us here have been at this a long time, I simply do not recognize your 'handle' as a long term, ahem, contributor

perhaps Cathar will share his experiences with his late chiller project

Since87, see what happens when you give Les some slack ?
next he'll be pounding on me, lol