There's a lot of spare parts in the typical o/cer and watercooler's closet. If you "upgrade" to the part of the day you'll go broke. Bill's comment on total system performance is well noted, and I will add my (anecdotal) test results here. They differ from Bill's in that:

1)My flow rates are +/- 6%
2)My temperatures are less precise
3)I was not isolating the waterblock so there is a heatercore for heat exchange with air rather than a chiller.

Take the numbers as illustrative rather than quantitative, please.

http://phaestus.procooling.com/gpmvary.jpg

The test setup was a Maze2 (newer rev) 1975 Caprice heatercore with 2 Panaflo H1A fans, Eheim 1250, 1/2" silicone, and a ball valve to adjust flow rate. Overall conclusion was that you must keep flow rates over 1 GPM; with the pressure drop of the flowmeter and the 1250 I couldn't get to the higher flow rates where Bill sees no improvement. I would suspect, however, that if the wb performance isn't improving but the radiator's efficiency is decreasing with ever higher GPM, then the delta T may actually rise again.

Thanks for that pHaestus, I think your results fall within BillAs, if one look pretty close up...

What do you think about adding a bypass to the rad?

Ya'll know that we talked about pump efficiency, but we didn't say the words: if you're shooting for high GPH, you should get a higher pressure rated pump, so that the pump runs in an efficient range. What I'm having an issue with is that there are few, if any centrifugal pumps that seem to have those specs, so maybe we should review the whole big pump thing, and start looking at other pump types.

The Eheim 1250 is not all that great for pressure, but no one questions it, and since Eheim won't put out P-Q curves for them, then this thing starts falling the way of the rumors with statements like "get an Eheim, it's the best". Shame!

"efficiency'' is being used a bit loosely (and several different items possibly confused)

other than for several round-tube radiators (graphs here),
gross dissipation will always increase as the flow increases (remember both wbs and rads are heat exchangers)
[invert a rad curve - guess what ? its just like a wb curve with different units)

rad efficiency has to do with the dissipation/in.^3/°

nikhsub1
posing an 'intellectual' question to ponder -

could not the thread have been titled 'Swapped pumps and a crappy wb install boosted my temps' ?
- using exactly the same data

all I'm saying is that what 'seems obvious' may not be the case

Ben
I had (for 2 days !) a 1060, and it had a P-Q curve on the side of the box
do not the others ?

Bigben you are right about Eheims being "the best". Although I do think they are the best in terms of failure rate and leakage. Although your valve idea sounds interesting, I don't really want to be the guinea pig. All I am after is the best possible temps with H20. Like I said, the 1250 "seems" to be the pump of choice and I figured there must be a good reason, seems I shouldn't go figuring. To add even more complexity to this issue, I swapped the Spiral for the TC-4 and I am seeing somewhat better results thus far. All I can say is that I will be putting the 1048 back in when I get some time, only this time, I WILL NOT remove the TC-4 during the process so temp differences HAVE to be based on the pump alone. Am I correct to assume this?

Iwaki (MD motors only, made in Japan)
much better
cost effective ? - here we go again

Is this the P-Q curve?

That Iwaki is one of the best pumps I've spec'ed. The reality is that most people will use a cheap pump, and it is going to be used inefficiently. Hey, that's watercooling... we gotta go with it, or stand alone.

Thanks nikhsub1, yes, that'll work. Just for kicks, you could try removing the turbulators in the TC-4, it seems that Viperman gets 2 or 3 deg C better without them. (I know, it's dissapointing to get the TC-4 and run it like a maze1).

Bill I've actually taken a liking to your crass attitude, and yes, I suppose that COULD have been the thread title. I may not have as much experience in testing as you but I DO know how to install a WB and thermal grease correctly, as a matter of fact, I reinstalled the WB 3 times to make sure.

Woot! there it is!

Thank you!!!

Can someone explain this P-Q curve to me? How does the 1250 stack up?

I like BillA too. You know though, some people really shoot themselves in the foot: if you know how to install a WB, why did you have to do it three times?:p

I got this one.

If you run the pump with no restriction, then it's giving out 100% of its effort to move water, and since this pump is centrifugal in nature, which means that it's about 70% efficient in converting electrical energy into a waterflow (which can be calculated), then you can easily see that if you achieve half of the max flow rate once this thing is installed, your efficiency (aka energy efficiency)is somewhere around 40%.

I didn't HAVE to, I wanted to be absolutely positively 100% sure the higher temps were NOT DUE TO a crap WB install. Needless to say I got the same result each time.:(

OK but how to read the chart I scanned? WTF are those #'s?

On the horizontal axis, you have the flow rate. On the vertical axis you have the head. Head can be converted to pressure. The higher the flow rate, the less head/pressure there is.

If you could measure your flow rate, you would be able to tell the pressure that the pump has to fight to get to that flow rate.

If your coolant is mixing together then you'll have warm water... the "cool" water returning from the rad would mix with the "hot" water returning from the block, and when they mix in the reservoir, voila, you have warm. This would hurt the effectiveness of both the block and the rad since you're lowering the delta T in both for no good reason. I could see this working in a two chamber reservoir setup, but then you'd have to match the flow in both loops or one reservoir would drain into the other. If I'm missing something, please set me straight folks.

Its just a thought, but with the pump upgrade , I see a reduction in tube size going to the vid block. I wonder if going to a larger sized pump would reduce the flow thrue the smaller tubes (Hence the whole "Water will always take the path of least resistance" there for introducing warmer water that does excape the vid card loop. Just a thought. Also having the pump at a higher location would increase the heat output due to it having to lift the water higher. maybe im totally off base....


Also I want to note that I am running a very small rio pump, and I can still get temps in the low 40`s C

Well, you're right about one res draining into the other one, that's for sure.

We covered this in another thread here

I suggested sticking the rad loop outlet into the CPU loop inlet. It'd be near impossible to match the flow rates though, they're always going to be different.

That's the exact same thing as having two pumps in series (i.e. it's one loop, not two, unless I misunderstand).

This one is specifically for Ben, but for all to consider.

It is a myth that radiators have a sweet spot. Perhaps "myth" is too strong a term. Maybe misunderstanding is more accurate.

When you're considering fluid flow through a radiator you are concerned with convection from a fluid to a solid. Sound familiar? It's the same concern with a water block. And just like a water block, higher velocity will always improve heat transfer.

Take a gander at the equation for convection again. Q = h * A * delta-T. Both "h" and "delta-T" are continuous functions that vary throughout the radiator. Nonethless, "h" is largely a function of velocity with higher velocity increasing "h". For a fixed "Q" and "A", this means higher velocity equates with lower delta-T.

So why the misunderstanding? Just like with waterblocks, pushing more flow through a radiator requires more pump power. Power tends to go up as flow^3. When flow and power are both low, it doesn't take a whole lot of power to increase flow rate "significantly". As a result you get better results from a flow increase. Once you get to a certain point, it requires a greater increase in power to generate more flow than the improved "h" can compensate.

There's a double-whammy here, too. As "h" gets better, delta-T drops toward zero. You start getting delta-T too small and it requires massively more flow to make a dent toward yet lower delta-T values.

So practically, yeah, if you get beyond a certain flow (which varies by radiator) then you'll see a drop in "performance" versus increased flow. This drop is largely imaginary, as the radiator is really dissipating more and more energy as the flowrate goes up. The only way to truly judge this is to decrease heat input from a CPU (or simulator) as the power to the pump goes up such that total power input remains constant. Even this is a crapshoot, however, as pump efficiency varies according to flow rate.

Anyway, more flow will always make a radiator more effective. It just won't always make your fluid temperatures better.

Should also note that water flow is only a part of the overall picture. Airflow is every bit if not more important.

Am I not reading this graph right?

http://www.overclockers.com/articles481/dissvsflow.gif

Granted it is at a rather typical 120mm fan pressure and not something more typical of a serious fan, but for the application of these heat exchangers to water cooling the 0.05in H2O is pretty tpical. Doesn't this figure imply that for some radiators there is a "sweet spot" where dissipation is improved (Dangerden Cube) and for most others that lowering flow rates increases heat dissipation to some extent?

In the absence of additional information, the graph doesn't mean a whole heckuva lot. I know that BillA has done a lot of work in the area, but I do not know the specifics of how he has performed his tests.

All I'm telling you is that convection from a liquid to a solid will improve with increased velocity. Whether this results in an overall improvement depends on the change in energy input to the system versus the change in convection efficiency.

From this article: http://www.overclockers.com/articles481/

Bill controls inlet water temperature to 0.1C with a recirculating water chiller, so no effects of pump power or throttling here. Surely he'll pop in soon to elaborate :)

I think the difference in convection efficiency must be what is observed here.

the conditions were explicitly described in the article;
flows, temps, pressure drop and static pressures were all controlled and recorded (to generate the graphs);
and the tests were run a number of times


yes pHaestus, you are reading them correctly;
and interesting is that this 'anomalous behavior' occurs only with a particular type of radiator,
the round tube configuration having many short lengths with 180° bends
- it is not a testing 'artifact', something real is occurring

my speculation as to what is the cause revolves around the transition zone between laminar and turbulent flow regimes
-> I suspect that turbulent flow is attained at a lower than 'normal' velocity due to the bends in conjunction with the tube length;
then as the flow is slightly increased the regime reverts, causing the dip, and then reverts back to the 'typical' turbulent regime as the flow increases some more

for the specific rads shown, and ONLY for those, the bump is quite real
(I cannot count the hours passed incrementally shifting the flow rate up and down observing this phenomenon)

for any and all other rads; the sweet-spot is a fiction
(from a systems perspective however, the notion is not so farfetched - though technically not correct)

just noticed you were linking to the old article
(which sorely needs revision in view of my excellent and ever improving 20-20 hindsight)

the graphs have been corrected for a liquid flow rate calibration error and are posted here, the revised graphs

Speak of the devil :)

Thanks for clarifying that Bill. Seems reasonable I guess. Still a little confused as to how the flow regime would revert back to a more laminar type when flow was increased. Also the Serck is not the same type of rad as the others with this maximum; why do you think you see it with that one (but not with the other heater core types)?

am inclined to make an analogy with the tuned stacks on race engines
in a certain velocity range the pressure pulses are 'in phase' (wrt the manifold vacuum) and their throughput increases
- but will be worse above and below this range

with the fluid changing direction, there will be an entrance effect and the flow will not become uniform for some distance (depending also on the velocity and diameter)
the same factors that promote a turbulent flow regime can also serve to inhibit it
note that the 'bump' occurs at different velocities for the different sized round tube rads
(disregard the very small rad data, high variation due to the small temp/flow rates)

I am quite sure myv65 can describe this with the correct terminology
(I'll ask another fellow also, but this is not a good forum for the uninitiated)

I just found out yesterday that the Serck is an oil cooler with pleated turbulators in the flat tubes
(as are Mocal rads, also from the UK)
the BeCooling 'dual' rad is a very exceptional performer (not in the original article) which also has fluted turbulators in the straight sections

turbulators are excellent devices, but make the rad atypical

Tons of good info; thanks.

In chatting this over with Bill and looking at his revised graphs, I'll say the following.

As a general rule, you will see improved heat transfer with increased flow rates. There exists the possibility that due to a given radiator's internal configuration, there may be a performance bump at a specific, and typically low flow rate. At the high-flow end of this bump, performance should be better than at the low-flow end of this bump. Performance outside of this bump will improve with increasing flow.

Why some radiators exhibit this performance is open for debate. I don't personally believe it is related to laminar flow. I think it's likely a result of secondary flow, mainly vortices, and how their shape varies across a narrow flow range. As I told Bill, this is the la-la land of CFD. It's also possible, though unlikely, that this is a fictional artifact due to the way Bill is testing and measuring his results. You can easily dismiss such a result in a handful of trials. Dismissal is not so easy as the number of trials grows.

I lean towards the vortices myself, but I have no intention of cracking open a heatercore to prove it.:D

You know that it's an aspect that's relatively new. There was an article in Scientific American, last year I think, about how it is that dolphins can move so fast, while they don't have the muscle mass to do it (they use vortices, like from a boat's bow). There was also a bit about an artificial fish test in some university somewhere.