My micro-channel concept block

[BillA]

as you wish Peter
your last two posts are illustrative of what I speak

[myv65]

Originally posted by GeminiCool
if you had a 1" diameter hose and reduced it through a 1/4" hole, then back up to 1" then back down... where will the backpressure be GREATEST. You are such an ass! The resistance felt at the 1st reduction will be ORDERS of magnitude greater.

I'm not going to bother defending Bill's mode of operation. Everyone knows his style. I also know nothing of your history with him. All I'll say is if you wish to make him look like he doesn't know his stuff, make certain what you say is correct. Your statement above is flat out wrong. I'll do you the courtesy of explaining why.

Flowrate throughout a single-path system is a constant. This means that regardless of where one places a given component, velocity and volumetric flow through that component is a constant. This means that pressure drop across a component is a constant.

If you have one inch tubing and place a single 1/4" restriction in the line, you'll have a specific total flowrate with the greatest pressure drop occuring at the 1/4" restriction. If you now place two runs of one inch tubing, each with a 1/4" restriction, in series with one another, your total flow rate will drop and your total pressure drop will be greater. The pressure drop at each restriction will be the identical, though lower than the pressure drop when only one restriction existed.

Rather than "orders of magnitude higher" at the first one, the pressure drop at each will be the same because each sees the same flow rate.

Personally I commend your previous post stating that you would take this outside the forum. It's entirely unrelated to the original topic and benefits no one to continue an argument such as this within a thread. I realize it isn't easy maintaining one's composure when you've been insulted, but jawing about it here will not improve the situation.

[GeminiCool]

myv65 -

Ok let me think about this.. I may actually have to try this experiment.

If I have a pump with a 1" outlet and measure the backpressure seen by the pump I'll see.. X

Now I add one .250 fitting, I'm sitting at the pump I will see some new backpressure X250

Now I add a 1" pipe behind the 250.. the backpressure seen at the pump should not go up.

Finally I add another .250 fitting. My pump is pulling water from a pond or something. When I add the second .250 fitting, your telling me that the pump will see twice the backpressure as it did with a single 250 fitting? My backpressure seen at the pump is 2*X250?

That seems very counter intuitive. And I believe it is a valid question to ask how the pump could possibly see 2x the resistance??

The intuitive answer is that when the second 250 fitting is added there will be some drop in flow and some increase in backpressure. HOWEVER that drop should not be near as significant as the 1st.

Please explain this to me

[bigben2k]

All right ya'll... take it down a notch or two!

Cathar: That pic was just a baseplate prototype, not a complete block. If couse it would have to have fins (of one type or the other).

The problem here bigben is that the statement is first made that we need thin base-plates to decrease the thermal resistance of the copper perpendicular to the CPU die, and then it's now stated that we need to increase the copper thickness to decrease the thermal resistance laterally.

I missed that completely. GeminiCool said:

I'm working on a base design that will vary the thickness of the base in the channel as a function of the radial distance from the core. The end result should be a better transfer of heat at the center do to a thin base and better transfer of heat in a radial direction do to the decreased thermal resistance.

I took that to be talking about a baseplate design, as I illustrated. Techno babble aside, the point was about the design, and although Peter might not have described it technically correctly, the design stands.

I don't see what all the huffing is all about. From my perspective, I see a block designer that's trying to use technical language to describe a design. Obviously he's struggling with the words (as the spelling alone suggests), so why can't we just help him along?

GeminiCool: I'm really glad that you stopped by. For the pump heat issue, I'm going to refer you to this thread where you can see me going through the process of figuring out exactly what the heat is, where it comes from, and so on.

For the BP design, I would hope that I was close, with the pic I put up. I do foresee a few problems with it though:
1-very hard to lap a recessed area
2-block becomes "core specific"
3-it must have enough clearance to make sure that it doesn't contact the surface of the CPU (shorting the bridges).
4-A mounting scheme would have to be redesigned.

[myv65]

OK, you are on the right basic track, but ya haven't quite arrived.

If you have a pump with an open discharge, you'll get maximum flow at what is basically zero discharge pressure.

If you add a 1" line of some length (even if the pump discharge is only 1/2") you'll add some finite resistance. Flow will drop slightly and you'll have some small, measurable discharge pressure.

If you add a 1/4" fitting, your flow will drop dramatically. Your pressure will go up dramatically. You will have gone from the "far right" end of the P/Q curve to the left region.

If you add another 1" line to the end, your resistance will go up slightly again. There won't be much change at all in flow or pressure, but some very small change will occur. Your flow will be ever so slightly lower and your pressure ever so slightly higher.

If you add another 1/4" restriction to the second line, your resistance will go up again and your flow will drop again. The changes will not be anywhere near as dramatic as that first major restriction, but they will be real. And here's the point that your earlier post missed: With both 1/4" restrictions in place, each will present precisely the same resistance and each will yield precisely the same pressure drop. Each individual pressure drop will be lower than that caused when only a single 1/4" restriction existed, but the total drop due to two tubes and two restrictions will be higher.

This is really a case where a graph might help. I suppose you can think of it this way: Pressure drop across an orifice is a function of velocity. When you have a single orifice, it will have some flow vs pressure curve. If you put two in series, that same flow vs pressure curve applies. However, since there are two, the total pressure required to maintain a fixed flow would go up a lot. Since a pump actually reduces flow as pressure goes up, we know this can't happen. What does happen is that the flow goes down (total overall pressure increases) while the pressure drop across each individual resistance goes down (flow rate drops).

[utabintarbo]

Originally posted by bigben2k


For the BP design, I would hope that I was close, with the pic I put up. I do foresee a few problems with it though:
1-very hard to lap a recessed area
2-block becomes "core specific"
3-it must have enough clearance to make sure that it doesn't contact the surface of the CPU (shorting the bridges).
4-A mounting scheme would have to be redesigned.

What about something along these lines (as applied to cathar's micro-fin block). The baseplate gets thicker as we get further from the core (radially). It would alleviate all the issues dealing with non-flat bases and still apply the stated principle (unless I'm missing something here).

Pardon my cheezy MS paint skills!

Bob

edit: How will the integrated heat spreader work on the new AMD offerings?

[bigben2k]

I like it Utabintarbo, but the point was to basically limit the effect of the baseplate to the "radial" pattern, or in other words, keep the same thickness of copper at exactly the same distance from the core, which is how the heat spreads, roughly.

It may not be relevant though: the heat will spread (mostly)equally in all directions, but as it spreads sideways, who really cares, since that heat will have a larger mass of copper to allow it to be dissipated, right?

Wrong.

The added mass adds to the buffer zone between the cooling solution, and the core. The bigger the buffer, the higher the core temp will be.

Mind you, my baseplate prototype isn't of much help either. What we need is for the coolant to flow at a level below the top of the core, and that just goes into the "physically impractical" realm.

Oh well, back to the drawing board!

[GeminiCool]

myv65 -

Thanks for the explanation..... I guess this is where I have problems communicating, especially with ppl like BillA.

So you now see and maybe even agree with my original statement? I believe I think in very logical terms. I had not pondered the pressure drop being the same across both reducers, but I will keep that in mind. Basically, I was trying to make the point that if you had a 1,000,000 of something and took away 100,000. The dominating number is still 1,000,000. Yeah yeah, you have to sum it all up, but that's all in the "three sigma" band. I understand now that the 1,000,000 in this example as seen by the reducers is 500k + 500k but as seen by the pump, it's all just 1M.

So I hope you see my point. My logic may be primitive, but it's has served me well so far. I have problems communicating, and many times find myself having to re-explain what I was really trying to say. Oh well, it's a short coming I guess.

I just hate it when I see others making posts directed at forum goers that can be misleading. Some statements are "basically" correct and if you take a step back, you see that the one little piece which is being nit picked is really insignificant. Or was perhaps simply a mis-interpretation.

Thanks,
Peter.

[utabintarbo]

Originally posted by bigben2k

What we need is for the coolant to flow at a level below the top of the core, ....

OK I guess I dont understand why this requirement, given the (essentially) non-interface of the block with the core except at the top of the core. Would you not need a water-cooled IHS (ie. total core contact with baseplate) to really get to what you're shooting for?

Sometihing like this:

[myv65]

@utabintarbo,

AMD's heat spreader will work well, for air cooling. For good watercooling it will almost certainly do more harm than good.

@ben,

Actually, increasing the thickness outside the core area may not be bad. Yeah, more thickness normally means more resistance, be delta-T is not a function of resistance alone. Delta-T is resistance and heat flux. Outside the core, heat flux drops dramatically. The fins outside the core will not get rid of as much heat because most of the heat gets handled directly over the core. For a given amount of heat and a given distance to travel, the delta-T in that material will be inversely related to cross sectional area. Translation, more material means lower delta-T.

Over the core it is a different situation where the distance to travel is the material thickness. Outside the core, the distance to travel is roughly the square root of the thickness^2 + radial distance^2. As you get farther from the core, the "thickness" portion takes on less meaning in determining distance. For radial heat dispersion, thickness is proportional to cross sectional area meaning that resistance is inversely related to thickness.

[bigben2k]

Right on Dave, as usual

UT: that would be about right, but the problem is that it's not so easy to make a block do that, in fact, I'd bet that it wouldn't fit every core.

'tis probably best to stick to a flat baseplate.

[gone_fishin]

Originally posted by GeminiCool
myv65 -

Thanks for the explanation..... I guess this is where I have problems communicating, especially with ppl like BillA.

So you now see and maybe even agree with my original statement? I believe I think in very logical terms. I had not pondered the pressure drop being the same across both reducers, but I will keep that in mind. Basically, I was trying to make the point that if you had a 1,000,000 of something and took away 100,000. The dominating number is still 1,000,000. Yeah yeah, you have to sum it all up, but that's all in the "three sigma" band. I understand now that the 1,000,000 in this example as seen by the reducers is 500k + 500k but as seen by the pump, it's all just 1M.

So I hope you see my point. My logic may be primitive, but it's has served me well so far. I have problems communicating, and many times find myself having to re-explain what I was really trying to say. Oh well, it's a short coming I guess.

I just hate it when I see others making posts directed at forum goers that can be misleading. Some statements are "basically" correct and if you take a step back, you see that the one little piece which is being nit picked is really insignificant. Or was perhaps simply a mis-interpretation.

Thanks,
Peter.

If you add a hypothetical restriction that causes a 50% reduction in flow, then logically when you add the same size reduction ahead of the first, the flow will not result in zero flow, but 50% of the remaining flow.
100gph +50%r = 50gph + 50%r = 25gph is the logical line of reasoning I think you were trying to convey?

[utabintarbo]

Originally posted by bigben2k
Right on Dave, as usual

UT: that would be about right, but the problem is that it's not so easy to make a block do that, in fact, I'd bet that it wouldn't fit every core.

'tis probably best to stick to a flat baseplate.

Oh, I have no illusions about actually trying to produce such a thing! I'm wondering exactly how AMD is going to make it work. Any links I might follow?

Bob

[GeminiCool]

gone_fishin -

Yeah I guess that's about right, but I think I felt the drop would be a little less drastic, since 50% is not ignorable. I did not think in terms of pressure drops across fittings. I simply looked at it from the perspective of the pump.

If I were at the pump looking out, when that 1st fitting was added, I'd see a HUGE increase in pressure. A second fitting would not make a huge increase... yeah you'd see something, but from the perspective of the pump, the "big hit" was when the 1st fitting was added. Based on this, is it not 'reasonable' to approximate the drop in such a system as close to the drop of this one fitting, weather there be two or even three reducers a little farther down the line?

That's the spirit of my theory. I don't see why you can't approximate the pressure increase/flow rate drop, in this situation as very close to the same with one reduction as two or even three reduction, from the perspective of the pump. If forced to think numbers, let's say the reducer cuts flow by 75%. Now a second reducer is added and flow is 73%. The pressure difference between each individual reducer is 1/2ed (thank you myv65), but the pump still sees the same "total" pressure or nearly the same and you have "about" the same flow rate.

I must say that myv65 comment's were very informative and the more I think about it the more I seem to recall that being taught in my thermo courses, especially, in phase change type systems.. but I'm getting old, a 30 something body with the memory of an 80 something.

[bigben2k]

It seems counter-intuitive, when explained in detail, doesn't it?

I look at it like this: there's a flow rate that's dictated by the total restriction and the pump.

The flow rate is the same at any point, so it logically follows that the pressure drop for the same restriction, is also the same.

[GeminiCool]

hey Ben -

yeah it definitely messes with your head.

I guess in the same spirit, removing one of the 1/2 restrictions doesn't give you twice the flow, just twice the drop across that one restriction. Same thing as I stated prior, but in reverse.

Perhaps I should try to put some real numbers to it. I'll do the fill a bucket.. that should get people talking. :-p But it would serve the purpose of demonstrating a 2 fold reduction or increase in flow. If it takes "about" the same time to fill the bucket, the flow seen by the pump could be approximated using one restriction.

[BillA]

talk about weasel words, and in copious quantity
Peter, come to the party

"The pressure difference between each individual reducer is 1/2ed (thank you myv65), but the pump still sees the same "total" pressure or nearly the same and you have "about" the same flow rate."

one of two choices here:
Dave's words were inadequate to convey the information
or
you Peter, are up to the same old bullshit of trying to 'make' your half-baked ideas into descriptions of what I]you think/speculate[/i] about what's really goin' on

please spare us the imagined view from the discharge port of a pump
speak of facts, cite some references, titles, page numbers

YOU ARE WRONG
and now you will expend 10 pages of inane circular posts to defend the indefensible
simply because YOU are unable and unwilling to say "I was wrong, lame idea, sorry for being so stubborn"

there are readers here who do not understand much more about fluid dynamics and thermo than you do, and you can always find support from them
give it up
science is not based on popularity, nor disproven by filling a bucket
(jesus christ, and you have an MS !)

they also do not appreciate
- that a wb does not need a 'good design' to function well, it needs only to be made moderately well
(the rationale bringing it into existence is not essential to its performance)

- that the design description you gave on OCers was absolute (scientific) garbage so your 'qualifications as a designer are the product of circumstance - not engineering,
and fortunately Fixitt had a good model that you could improve upon

some references (but I gave these same ones to you on OCers, eh ?)

Mechanics of Fluids by Shames, Chap 8 Incompressible Viscous Flow through Pipes, 8.9 Minor Losses in Pipe Systems

Flow of Fluids by Crane Tech Paper No.410, Chapter 2 Flow of Fluids Through Valves and Fittings

and ck your e-mail for what I really think

EDIT: removed information given 'in confidence'
sorry Peter, did not recall

[bigben2k]

[myv65]

Originally posted by gone_fishin
If you add a hypothetical restriction that causes a 50% reduction in flow, then logically when you add the same size reduction ahead of the first, the flow will not result in zero flow, but 50% of the remaining flow.
100gph +50%r = 50gph + 50%r = 25gph is the logical line of reasoning I think you were trying to convey?

This one is definitely getting way out of hand. Here's my last comment on this thread.

Right idea, but wrong final answer. Pressure drop and flow don't quite work so easily as summing up inverses. In general, pressure drop varies with velocity squared. P/Q curves for pumps are never flat (unless you've got a positive displacement pump, then who cares). The only way to really know is to measure the results.

[rant warning]

The following is directed to no one in particular:

This whole relationship between flow and resistance is probably the biggest misunderstanding people have when trying to determine how to plumb multiple blocks and/or radiators. There is no single method that is guaranteed to yield the better result. I've been dragged into numerous "discussions" on this topic and regardless of whether I take the "simplified, analogy-filled" approach or the hard-core theoretical approach there are always those that insist it must be best one way in all cases. It simply isn't true.

I've long since resigned myself to the conclusion that not everyone will ever "get it". I do not lose sleep over those that would insist their will in the face of overwhelming contrary evidence. I try very hard to form one of two opinions about people that I "meet" on the forums. I either conclude they're on the up-and-up and doing their best ("best" whether learning or sharing knowledge) or I conclude it is not worth my time to bother with them.

[gone_fishin]

Originally posted by myv65
This one is definitely getting way out of hand. Here's my last comment on this thread.

Right idea, but wrong final answer. Pressure drop and flow don't quite work so easily as summing up inverses. In general, pressure drop varies with velocity squared. P/Q curves for pumps are never flat (unless you've got a positive displacement pump, then who cares). The only way to really know is to measure the results.

[rant warning]

The following is directed to no one in particular:

This whole relationship between flow and resistance is probably the biggest misunderstanding people have when trying to determine how to plumb multiple blocks and/or radiators. There is no single method that is guaranteed to yield the better result. I've been dragged into numerous "discussions" on this topic and regardless of whether I take the "simplified, analogy-filled" approach or the hard-core theoretical approach there are always those that insist it must be best one way in all cases. It simply isn't true.

I've long since resigned myself to the conclusion that not everyone will ever "get it". I do not lose sleep over those that would insist their will in the face of overwhelming contrary evidence. I try very hard to form one of two opinions about people that I "meet" on the forums. I either conclude they're on the up-and-up and doing their best ("best" whether learning or sharing knowledge) or I conclude it is not worth my time to bother with them.

I was trying to get a line of reasoning of what he really meant from the long posted paragraphs down to a single short description of what was being discussed. What was failed to be said in discussion is that the same percentage of reduction in orifice size would need to be placed as the second restriction in order for the same percentage of flow restriction to be imposed as the first.

In short, I was just trying to find out what GeminiCool was trying to say. And this was his answer to the querry,
quote by Geminicool, "Yeah I guess that's about right, but I think I felt the drop would be a little less drastic, since 50% is not ignorable. I did not think in terms of pressure drops across fittings. I simply looked at it from the perspective of the pump. " end quote

myv65, you are of course correct. Vagueries should be questioned to clear up long rambled roads that lead to nowhere

[pHaestus]

Wow Bill I guess Peter must have peed in your Corn Flakes before. Relax; have a beer.

The point you made today somewhere on this forum about integrating the same principles into every new question and not starting fresh every time is a good one though.

As is Dave's comment that pressure drop is proportional to velocity squared. That's an empirical relationship though, correct? Too much of this sort of thing (fluid dynamics and heat transfer) is; not the same as the thermo that one gets as a chemist.

Understanding that the curves (resistance vs flow rate) for all the components are interactive is also key I think.

[Cathar]

Wow. I go to sleep, wake up, and find that I feel like my house has been used for a wild party....

Good morning all...

[BillA]

Cathar
now you know what I think about the cavorting on OCAU while I’m trying to catch a bit of sleep
sorry (not really) about messing up 'your' thread

pHaestus
"Understanding that the curves (resistance vs. flow rate) for all the components are interactive is also key I think."

well, not really
(a shame Dave gave up in disgust, I am presuming - as a major contributor)

this was generated as part of calibrating the flow bench:
(these are wb connectors/ID restrictions)



the value shown is for a single reduction from 0.625in. ID
if there were 2 such in line the pressure drop would be doubled if the flow rate was held constant
- but this is NOT the 'normal' pump setup, eh

as 'pipers' deal with flow resistance, everything is translated into 'equivalent lengths' (of straight pipe of the same type and size as the system or branch) and the head loss calcs are run then on the total system 'footage'

no interaction, simply additive resistances
see Crane

EDIT: this is an old graph and is innaccurate in an absolute sense for several reasons,
none of which affect its use to illustrate the above example

[Cathar]

It's okay. I'm not really fussed at all.

I'm surprised that people could read my thread at OCAU and all they get out of it is a friendly caricature of yourself, and miss all the discussion on physical design constraints, machining constraints, thermal transfer, nozzle experimentation, pump flow experimentation, Les's excellent analysis of `h' vs various designs vs flow rate, etc, etc. This list goes on, and yet the lasting impression is a picture of a guy who smokes?

[pHaestus]

Yes I did not define my boundary conditions there eh?

Sorry about that. I meant for a given pump, since that seemed to be the topic for discussion. Not for a given flow rate.