Some real world data on dual pump undervolting

[HanMies]

After consulting this forum extensively for my first and second build, I thought to post some data for those like myself who do not understand pump curves :). Maybe this will be helpful to someone. Sometime. Maybe. Basically I have played around with BigNG and MiniNG to see how pump voltage affects flow and temps. Also something regarding fan speed.

Unfortunately I have no way of measuring noise, so just subjective opinions. However, low noise was high on the agenda.

First the relevant hardware:
Asus Crosshair IV
Phenom II 1090T at 3.6
2x Asus 5870 V2 at stock clocks

Then the loop, in order:
EK Multioption Reservoir
MCP350 with XSPC V3 top
XSPC RX 120.3 radiator
XSPC RX 120.3 radiator
MCP350 with XSPC V3 top
Aquacomputer flowmeter
XSPC RX 120.3 radiator
2x EK Asus V2 full cover blocks with EK parallel bridge (temp probe in one inlet)
EK Supreme HF
EK full MB block

All rads with Scythe GT's (the 1150rpm kind, 6 in pull 3 in push) with shrouds, hooked to the BigNG, pumps to the MiniNG.

Flow
Let's see how this table works out:
% V rpm lpm dual lpm
70 8,4 2320 1,7 2,5
75 9,2 2500 1,9 2,8
80 9,5 2660 2,0 3,0
85 10,2 2820 2,3 3,4
90 10,6 3000 2,4 3,6
95 10,9 3160 2,6 3,8
100 11,9 3340 2,9 4,3

http://www.hostthenpost.org/uploads/a9ebc268dc9a8abc5dd24722db719d03.jpg (http://hostthenpost.org)

Just to clarify, "%" is the MiniNG setting, V is voltage reported by MiniNG, lpm is liters per minute with one pump and dual lpm is with both pumps at the same %. Please note that when using a single pump, the other pump is still in the loop but turned of.

So basically one MCP350 at 100% is roughly the same as two at 75%. Note how the 5% increases do not increase voltage (and consequently flow) evenly.

Temperature

For temp testing I ran Furmark and SuperPi at the same time. I admit that this is far from scientific as I was too impatient to give it much time to heat up or to cool down - however, I think the temperatures were fairly stable when taking the reading. The setups were:

1. Both pumps at 70% and the fans on a response curve (max reached was around 50%)
2. Both pumps at 100% and the fans on a response curve (same max)
3. Both pumps at 70%, fans at 100%
4. Both pumps at 100%, fans at 100%

Flow Water CPU GPU
1. 2,5 36 43 49
2. 4,5 36 43 48
3. 2,4 33 38 45
4. 4,5 33 38 42

http://www.hostthenpost.org/uploads/d22a9ab2a05a00dbb62adcc7b9c96e0a.jpg (http://hostthenpost.org)

Ambient was around 28 (all Celsius). What I found interesting is that increasing flow from 2,5 to 4,5 lpm (72% increase) drops GPU by 1 C, no change in CPU temperature. I would have expected much more!

A short comment on noise: I found the pumps at 80% to be the best. At 70% they are quieter, but there tone is more annoying. The GTs at 30% (about 400 rpm) are very quiet, and I don't notice them increasing to 50% with the curve. Hiking the up to 100% manually increases noise noticeably and the tone becomes slightly more annoying. I think one of the fans may have slight bearing damage.

I hope this stuff can help someone in choosing pumps/planning loops! Obviously any comments would also be appreciated. In the future I will probably try an Aquastream pump and try Nexus fans, as both are supposed to be extremely quiet.

[Martinm210]

Nice work!

Perhaps throw the results in some simple bar charts...seems like most peole can relate to that better than anything.

[BrokenArrow]

This is very good information. Thanks.

Did you have both MCP350s hooked to one miniNG? These are the 10W pumps, right instead of the 18W?

[HanMies]

Did you have both MCP350s hooked to one miniNG? These are the 10W pumps, right instead of the 18W?

Yes, both hooked to one MiniNG and 10W. Can't get the software to save settings, but the adjustment screws and jumpers on the thing work really well. 70% is below startup voltage, but the MiniNG gives a startboost of about 85% to start and then slows down (assuming that you have the cable that connects both the molex and rpm cable).

[HanMies]

Posted some charts.

What continues to baffle me is the minimal 1 degree GPU benefit of increasing flow from 2,4 lpm to 4,5 lpm (~75% increase), while I am nowhere near 1,5 GPM = 5,77 lpm, often quoted as the "point of no benefit".

[noob question]Is flow rate really relevant for a silence seeker at all?[/noob question]

[BlueAqua]

I think a lot of people over estimate the value of high flow rates. Higher fan flow rates are vastly more important.

I usually have a lower water flow rate while maintaining a medium fan flow rate, works out pretty well for my systems.

[Section8]

Nice work. The charts help out a lot from what you previously posted. It makes it much clearer. So now when I tell someone they don't have to have a million GPM flow rate to get good temps I have Data to back it up :) I also agree that having a good Radiator/Fan combination with matching characteristics is much more important. :) Nice work!! Thx.

[eclypse]

Makes me feel stupid buying 2 DDC 3.25 pumps for a single look with dual D5 pumps for another loop. One loop for 6 blocks and the other for a 480GTX and 2 360GTX rads.

Flow would be around 1.58GPM for the blocks and 2.04GPM for all the rads. I could of just used the dual DDC 3.25 pumps for the whole system and had a flow rate of around 1.60GPM.. but i have 3 D4 pumps laying around so why not i guess hehe.

[BrokenArrow]

I have seen temperature gains voided by the additional heat dump of the second pump. The real reason for using dual pumps in series is redundancy. Anything beyond that should be considered "gravy".

[eclypse]

I have seen temperature gains voided by the additional heat dump of the second pump. The real reason for using dual pumps in series is redundancy. Anything beyond that should be considered "gravy".

I doubt i'll have that prob with a HWLabs quad and 2 tripple fan rads.

[HanMies]

Let's keep in mind that the data applies to my specific scenario, where there is no significant overclocking and with more radiators then sense. Three 360 rads is probably a bit excessive for CPU, MB and 2xGPU.

I would hesitate before drawing conclusions for really high performance systems, as there the component temps could easily be 25C more. I think there is a test at dazmode.com where flow was found to have an insignificant temp effect in a CPU only loop, but by overclocking and adding blocks the flow became more relevant.

[Martinm210]

Yeah, it's nothing particularly new. Vapor has done extensive CPU flow rate senstivity work (http://skinneelabs.com/i7-blocks-2.html).

By the time you add in the heat dump, this is what you approximately get with a single triple radiator:
http://skinneelabs.com/assets/images/CPUBlocks/i7/Round2/Overall/Results22A13EP.png

Pumping power legend:

Very High Pumping Power: All three MCP355 pumps and the D5 are on at full speed--this has a very similar PQ curve to a pair of RD-30s at 20V.
High Pumping Power: Two MCP355s with EK V2 tops are on at full speed. The other two pumps are off.
Medium High Pumping Power: A single MCP355 with XSPC V3 top is on at full speed. The other three pumps are off.
Medium Pumping Power: The stock D5 is on at full speed and setting 5. The other three pumps are off.
Low Pumping Power: A single MCP355 with XSPC V3 top is on at minimum speed (~7.7V, ~2450RPM). The other three pumps are off.
Very Low Pumping Power: The stock D5 is on at minimum speed--setting 1. The other three pumps are off.

Hard to match up you're pumping power exactly, but the difference between medium to high pumping powers usually doesn't amount to much. We're talking fractions of a degree and many times nothing at all.


Most blocks today are very efficient even at low flow rates, although I expected the GPU block to be a bit more reactive...it didn't seem to change much either.

[MagisD]

Yeah, it's nothing particularly new. Vapor has done extensive CPU flow rate senstivity work (http://skinneelabs.com/i7-blocks-2.html).

By the time you add in the heat dump, this is what you approximately get with a single triple radiator:
http://skinneelabs.com/assets/images/CPUBlocks/i7/Round2/Overall/Results22A13EP.png

Pumping power legend:


Hard to match up you're pumping power exactly, but the difference between medium to high pumping powers usually doesn't amount to much. We're talking fractions of a degree and many times nothing at all.


Most blocks today are very efficient even at low flow rates, although I expected the GPU block to be a bit more reactive...it didn't seem to change much either.

OK unless I'm reading this wrong the Very high pumping temps are higher due to pump heat dump and less time in rads so the sweet spot for Loops would be at the med high to High range. Which is different for every single loop.

Anybody have a Number or way to quantify those as in Lpm flow rate per rad etc ?

[eclypse]

My original loop had only 1 Laing D4 pump for the CPU (EK Supreme HF) through a TC PA 120.3 and i can tell you that flow looked so slow that i bet it was barley a GMP if that. The other loop went through 2x DD 480GTX blocks in parallel and another TC Pa 120.3. That loop looked great with the flow.. guessing by looking at the water movement in the res. One its splashing and the other was basicly still.

Now the new loop is going to be D4s on a EK dual top for all the blocks including a EK Classified MB block and i'll get around 1.58GPM+ since the 3x DD 480GTXs will be in parallel and the dual DD DDC 3.25 top will be just going threw the restrictive HDWLabs 480GTX/2x 360GTX rads. Both loops sucking from the same rez. I'm guessing i'll get around 2.04+ GMP on that loop. I was told that blocks dont need huge flow but rads love high flow and i'll acheive that with the split loop. Any extra heat dump from the 4 pumps should be cooled off in the rads since there overkill for the system as it is.

All those numbers came from Martinm210 and Skinneeylab's Pumpand Rad Optimizer.. As i said it should be more as theres no option for a dual D5 top or DDC 3.25 pump.

[eclypse]

What kind of flow is very high? 2-3+ GPM?

[eclypse]

Looking back at the pump and rad optimizer, if i would of bought a damn TC PA 120.4 rad insted of the HW Labs 480GTX then i would of matched them up with 2 newer (15MM) TC PA 120.3 rads and had enough flow around 2GPM with one single loop and a DD DDC 3.25 dual pump top to of been happy.