Maximize Pump Efficiency (High Flow)

[airspirit]

Someone in another thread mentioned being able to use a throttling valve to limit flow to a radiator and using a set rate of flow (controlled by the valve) to a bypass loop allowing the radiator to get near maximum efficiency (push it up to the point of maximum returns for flow, where more flow wouldn't really help), and allowing some of the extra flow to go through the bypass rather than get bottled up and stopped in the radiator.

What this would allow is for more flow through the waterblock (which is greatly beneficial in high flow blocks), greater pump efficiency (allowing it to flow at a rate closer to what it is rated for rather than keeping it ratcheted down because of extra restriction), and maximum efficiency of your radiator. You get the maximum use of ALL of your components with little added cost or size.

The diagram below lays out a simplified version of this, where the blue is your block, the light blue is the res, the red is the radiator, the green is the pump, and the yellow is the valve in the bypass line. Theoretically, if, restriction of the radiator notwithstanding, you had 6 GPM flow through your system, you allowed 2GPM through the bypass, and your radiator would restrict the flow down to 2GPM from the remaining 4GPM, you would have net flow through your waterblock of 4GPM, whereas with the radiator in straight series with the pump and block you would be limited, in these circumstances, to about 2.5 GPM.

Now, I used theoretical figures that will be wildly different from system to system. Granted, for low flow systems using a Via Aqua or a low powered Rio, this would be foolish, as the gain in flow would not offset the reduced cooling through the radiator. In a high flow system of 3 GPM or more, you may see gains, since the extra flow through the block would offset the radiator loss, since the radiator efficiency loss above 2-3GPM in most radiators and heater cores is minimal. I suspect that the overall efficiency of a high flow system can be raised in this way.

While a throttling valve was recommended to me to control flow to the bypass loop, I suspect in our situation a simple ball valve like what is on your kitchen faucet will work, since a throttling valve is just a more accurate way of restricting flow (but much more expensive and hard to come by). Literally, you could sit in Prime95 or some other "burn-in" software, and tune the valve to where you get the lowest temps. In some systems, the valve would be barely open. In some systems, the valve might be open quite wide. By tuning like this, I suspect that what would happen is that your minimum (idle) temp would be slightly higher than usual, but you would be able to reduce your maximum temps by a similar amount.

What do you guys think of this idea? It has me kind of excited, and I think this may help people using the Uberflow blocks like Cathar's or the Swiftys that are on the market.

[bigben2k]

Anytime you want to post that diagram...

blue is your block
light blue is the res
red is the radiator
green is the pump
yellow is the valve in the bypass line

[airspirit]

Oh ****, here you go:

[bigben2k]

Repost.

The idea is interesting, but BillA would shoot it down, the same way he did when I suggested the same thing. (that valve might as well be connected to the rad inlet directly).

The problem is that you don't get full use of all the components: if you open up that valve to match the restriction of the rad, you just split the flow through the rad in half. The performance will match accordingly.

Also, if you plan to run that Caravel fan at 10V, what kind of air flow rate are you expecting? Do you have a performance curve for that fan?

I'm going to bug you a little bit more today... what's your expected heatload, in Watts?

[myv65]

For your situation, I think the idea is a good one. For joe-blow looking to cool a CPU (and maybe a GPU) it is probably overkill. You realize that you don't need too much flow through the radiator to get "enough" cooling out of it. You also realize that getting higher flow through your blocks is desireable. If you only had one or two blocks, this whole thing wouldn't really matter.

To me the top two hitters here are:

1) Only one pump required. Fewer devices means fewer things to fail.

2) Lower system resistance. Means more of the pump head is available for your multiple blocks.

Being an engineering gearhead, you also get item 3, which is something additional to play with and tweak. The value of this can't be overestimated for true engineering types.

[bigben2k]

Being an engineer gearhead, you might get a better appreciation with this:
( 2 more to come)

[bigben2k]

Here's the air restriction:

[bigben2k]

and finally, the all important fluid restriction.

[airspirit]

My heatload will contain the following:

1x1900+@approx. 1880 (Palomino)
1x2400+@2400 OR!!! 1X1700+ (TbredA)@2000+Mhz (????)
1x700 Slot A stock
1x110W pump (jeezus pleezus)

And that is for starters. Next year, a dually Opteron is going on this bad boy. I estimate it'll add another 130W one way or another.

All told, I expect an immediate heat load of about 350W. Next year that'll be approaching 500W. That's a lot of frickin heat. That also assumes I never put any GPUs on the line (I probably never will due to the fact that I can live without the latest and greatest GPU ... I have a 8500LE @275/275 128MB now, and my other primary box is running a GF2MX400/64MB). With the radiator with ONE caravel running through it at 10W (I don't have the spec cos I'm lazy, but I estimate 450CFM out of it and dB levels at two feet of 33-34), I expect to have water temps at 4-5C over ambient (typically around 90-100F in the summer and 70-80F in the winter ... my machinery puts off a ton of heat).

[airspirit]

Besides, myv65 is right on one thing. I'd snake the valve to the front of the box right under the rad that would control restriction. The dial for the Caravel would stick out next to it (I really don't want to use 120VAC fans due to the fact that as it is I'm dragging extension cords in from two other rooms because of the amount of power I use ... if I try to plug it in to only two rooms I start blowing breakers ... one room means complete hardware failure and even my UPS' freak out).

Being able to tune my system for summer and winter use and to tune it for temps is going to be a fun addition to the project. All I need is to build in a mini-fridge for the Mt. Dew and I'll be in geek heaven.

[bigben2k]

5 C above ambient sounds about right.

I have to admit, it's really nice to be able to work with the rad specs!

Why don't you just mount that rad outside? I mean, you'll never be able to live with either the heatload, or the airflow, never mind the noise!

A 120 VAC fan will be more energy efficient than a 12V DC fan, because of the ~20% energy loss in the AC to DC conversion.

Your power source is going to be a problem. If you're in an appt, you'll need to tap into the downstairs neighbour's ceiling fixtures

[airspirit]

Good point on the fans. Didn't think of that.

As far as my 500W space heater, it is more fun to have it running a UT2K3 server than to sit in my living room looking pretty, if you know what I mean. When it starts dipping down to 0degF, I'm going to appreciate my computer room ( ).

As far as mounting it outside, I would really like to, but I have a problem. Being in a rented apt, I don't have the luxury of cutting the walls. Being on a ground accessable floor, I can't leave a window cracked. Being in a college ghetto next to greek row and party houses, I can't have ANYTHING that can be taken outside.

Point of example: the guy that used to live next to me had 5 hibachi grills, a bike, a flower pot, and the spare off of his car stolen. The wood scraps from my box that I left next to my porch lasted less than six hours before they disappeared. Since I've window mounted (inside) my 802.11b APs, I've had people try to pry my window open to steal them multiple times. I actually chased one off at 7PM just a couple of weeks ago. If you think I'll leave a window cracked you're nuts. I might as well be living in the slums of an inner city for all of the criminal activity up here.

My favorite is when some guys PICKED UP my old car and slammed it into the dumpster outside my house. It was an 86 Caprice Classic 4dr ... it must have taken quite a few of them. This is what I get for the convenience of living near campus.

[Can O' Beans]

Wouldn't it just be better to just add 1-2 more radiators in parallel (if you have room)?

You would lower the flow resistance in that part of the circuit(one thing you aimed for) and it would also slow the flow down by half in each of the rads (another thing it looked like you were aiming for) alowing a longer time for the water to be in the rad to be cooled down.

I would think that having a bypass line would just give you higher temps. Only part of the water is now being cooled by the rad and the bypass allows warmer water back into the waterblock.

[utabintarbo]

Originally posted by airspirit


My favorite is when some guys PICKED UP my old car and slammed it into the dumpster outside my house. It was an 86 Caprice Classic 4dr ... it must have taken quite a few of them. This is what I get for the convenience of living near campus.

I hope they are on the football team! That's a biiigg car!

[BillA]

not sure I agree with Ben's 'quote' of mine
nor that I understand quite what myv65 is saying
(probably I'm overlooking something ?)

what I can understand is that the valve as shown will permit the manual control of the coolant temp
- but note that with the valve shut the cooling will be 100%,
open any amount it will be something less that that - is too much cooling bad ?

looking at the valve's effect on the flow rate through the wb:
with the valve full open (assuming it is a ball valve with an ID not less that the tubing used), there would be some reduction in the overall resistance . . .
but not really so much as the flow is going through the branches of 2 Ts
(which also add a bit of resistance to the rad circuit as its flow must go through the run of these same 2 Ts)

so the question is actually:
Is the presumed improved convection within the wb due to the slightly greater flow of greater effect than the (CERTAIN as I see it) rise in the coolant temp due to part of the flow not being cooled at all and the other portion being cooled slightly 'better' ?

my experience with flow diversion is that the backpressure the pump 'sees' changes only slightly,
and therefore the flow rate changes only slightly as well

decreasing the flow through the rad is not going to improve performance

EDIT: an additional note
if one wants to 'tune' the CPU cooling system, it is a whole bunch easier to vary the air flow through the rad(s)

[myv65]

Originally posted by unregistered
so the question is actually:
Is the presumed improved convection within the wb due to the slightly greater flow of greater effect than the (CERTAIN as I see it) rise in the coolant temp due to part of the flow not being cooled at all and the other portion being cooled slightly 'better' ?

That is precisely the issue at hand. Without knowing the answer in advance, the valve presents the opportunity to find out. This is a case where the unique combination of components means his answer will be unique to his system.

[BillA]

trying not to rain on anyone's parade here
(just been banned from OCers so I'm on my 'good behavior' - like hell)
but
is there an OCer in the world who can distinguish said effects ?

at the most basic level - the CPU temp as a function of the valve's % open ? (say 0 and 100%)

with a rad limited system . . . perhaps
but no one on this forum would ever have such

I'm rather sure that the typical rad 'over capacity' would effectively hide the effect

airspirit
go for it -> and report back
BUT, you MUST hold your air temps absolutely constant
(QUITE a trick, I cannot do such)

EDIT: spelling is my next life goal

[airspirit]

There will be almost no resistance from the T junctions, as I will be using 1.5" PVC. The radiator will be on the straightshot from the res, with the side branch going to the bypass. It will be the same on the other side. The flow will be optimised to go through the radiator. Everything from the res to the pump will be a minimum of 1.5" in diameter (except the rad, of course) to prevent undue flow restriction. Between the pump outlet (3/4") and the res there will be 3-4 3/8" lines with one waterblock on each. Really, the point of greatest restriction in the loop is the radiator. I'm not saying this will work for everybody. This may help for MP users wanting psycho flow. This will help my system, without a doubt.

Even with an overkill pump like the 1200, I may be restricted to 3-4GPM max system flow if it is straight inline, leaving less than 1GPM per block. That is unacceptable, especially since at least one will be on a very high temp chip (my current XP chip is at 2V and 1880 Mhz, and the one that I will be picking up next month will be worse). In order to pull the heat from the block effectively, I will need more flow that 1GPM. By using the bypass it is obvious that my water temps will rise somewhat. The water that passes through the radiator will be cooling the water more efficiently, however, recovering ground somewhat. Overall, though, it is plain that the overall temp of the water will find equilibrium at a higher temperature than if the bypass wasn't used.

What is the question, is whether doubling the flow through the blocks with slightly higher temperature coolant will cool them more effectively when they are at peak output. I suspect that the answer is yes. When all is said and done, I will do some Poor Boy Benching of the idea. I will hook up all the machines and let them idle with all the coolant going through the rad for one hour. This amount of time should cause all the coolant to go to equilibrium temp. Then I will open the bypass little by little over the course of a couple days to measure changes. Then I will do the same with them all running crazy benchmarks and see what happens. If anything, it will give good (if non-scientific) data to demonstrate if this works or not.

I don't know ... if anything, there will be minimal head loss from the bypass loop, as it will be splitting off on the perpendicular arm of the Ts used, preventing any head loss from the joint (which would be minimal anyway since it will be a 1.5" joint). The worst case scenario is that I just leave the valve closed all the time. I do not think that will be the case, however. I haven't seen any hard data that can prove the concept otherwise. It only makes sense that if you utilise all the components as close to their "Max gain per effort" point that you will get better temps overall.

This will be an interesting experiment. The radiator is on a UPS truck from Spokane and my hose is stuck in Hodgkins IL, and I will probably be ordering the cases and pump later this month (I hope the wife doesn't read this!). Once I pick up a new board/proc/ram next month (Prolly Chaintech NForce2/AXP 1800+ TbredB if avail./2x256 XMS2700), it will be less than a week before I have the data for you.

[BillA]

"What is the question, is whether doubling the flow through the blocks with slightly higher temperature coolant will cool them more effectively when they are at peak output."

whatever makes you think that you are going to double the flow rate through your wb(s) with a rad bypass ?

never happen (you're not listening)
why do you think it will double ?

have you matched your pump's P-Q curve against the summed total of your system flow resistances ? (head losses)

I don't know what a 1200 pump is (probably described in another thread), but you have enough info to actually calculate with fair accuracy (+/-50%) what is going to happen

and that flow rate will never double

BTW, I'm unsure what you mean by "Uberflow blocks"; but Cathar's and the Swiftechs utilize rather different mechanisims, and their head losses are quite different
- many ways to skin a cat, eh ?

[MadDogMe]

(just been banned from OCers so I'm on my 'good behavior' - like hell)

Post a link BillA!, it's been a slow week , hav'nt seen a BillA~BaN for ages!!...

[BillA]

"This is the second time you have flamed another member, much less several members. Many of your posts were both derogatory and insulting."

a mod deleted all the posts, and gave me a new title: "Banned"
sos, nothing new
(my demanding excellence and rigor from the untutored and indifferent)

[airspirit]

Uberflow block = block able to use 5+GPM without MASSIVE head losses. The best example is a straight through pipe with a coldplate.

I don't have enough data to calculate exactly what will happen and that is why I will do the tuning tests. The Pondmaster 1200 is rated at 950 GPH at 6ft head, and 600GPM at 10ft head. Since dH is going to be less than 5ft at the extremes, and at 10ft head I can get 10GPM throughout (w/o resistance), so I feel it is safe to assume that I should be able to get at least 6-8GPM throughout. Since I feel it is doubtful that I can get more than 4GPM to push through that radiator limiting each WB to less than 1GPM, if I can get 8GPM with a bypass, I will be doubling the net flow through each block. At least, that all sounds good on paper.

I honestly don't know what the results of this project will be. My gut instinct tells me I will see higher idles and lower peaks, but I really don't know how it'll come out in the wash. I do know that my radiator will keep flow low, which when split between multiple WBs, the flow in the blocks will be abysmal.

[MadDogMe]

Uberflow block = block able to use 5+GPM without MASSIVE head losses. The best example is a straight through pipe with a coldplate.

That rules Cathars block out then! ...

***************
O/T, why do they delete the posts?, is the mod sleeping with masked geek?(hmm~maskedgeek?, is that the guy/homo that's tied up in the closet with a squashball in his mouth?, who'd choose that name?), mind you anyone who puts a photo of themselves in an avatar needs to sleep with a moderator to stay unflamed . is Safemode MaskedGeek's alter ego?, pre smeglocked? ...

I take it it WAS mg who you flame/corrected?, no one else has a "Please kick me BillA" notice stuck to the back of their head!...

[bigben2k]

Here's the pump chart
(you can click to change the chart type, line is preferable)

I calculated that the 1200 model would push 2 gpm through the rad (alone), which results in a pressure drop of about 14 feet. That's about all the pump can do (i.e. near shut off). Nasty rad!

[BillA]

4gpm to me says you have the wrong rad for your app
look at the fluid side graph Ben posted
you're pissing up a rope

change rads, or get more of them in parallel
do look at all those 180° bends, do you know that these rads are used with pressure pumps ?
- not aquarium pumps

by your definition there is only ONE "Uberflow block" - the Swiftech 462-UH