• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

new RBX test results

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
A great discussion! I thought I’d add to it and hopefully clarify some points:

The testing regime I’m undertaking is to enable an apples-to-apples comparison among waterblocks. The concept is to control inlet temp, waterflow, heat load and mounting pressure, so that the only variable is the waterblock’s performance at these parameters.

If I were to introduce another variable by letting waterflow vary among the blocks, then I think the comparison is then compromised – I can see a chorus of voices saying “But block x is at 1 gpm and block y is at 0.5 gpm – unfair!”

The point was raised that this may make unfair comparisons between high and low pressure drop blocks. I think users must make a determination if, in using commonly available waterpumps, it is better to go for a high pressure waterblock at a lower C/W or one with low pressure at a higher C/W and bank on increased performance due to higher waterflow through it.

Bear with me as I present some data:

I interpolated data from Bill A’s summary charts as follows:

Data from Bill A’s Tests

Delta C/W

<1 gpm >1 gpm
Tsu 4.5 2.5
WW 3 C 2 C
Slit 3 1.5
SW 1.5 0.5

WW = Whitewater, Tsu = Tsunami, Slit = BeCooling Slit Edge, SW = Swiftech 5002.

The deltas represent the difference between data points at the low and high ends of the curves.

A VERY clear message – the performance risk with waterblocks is under-performance at the low end, and more so with high pressure waterblocks. In absolute terms, the “high flow” advantage may be about 2C, while the “low flow” risk is more like 2-5 C.


***NOTE: I excised comments about radiators and flow per Cathar's comment below.***


As to displaying data, I again interpolated data from Bill A’s tests:

C/W @ 1 gpm vs Head Loss
WW 0.18 @ 1.05 mH2O
SW 0.21 @ 0.48
Slit 0.22 @ 0.3
Tsu 0.24 @ 1.85

If you imagine a graph of the data, you will see a graph similar to the approach I’m using currently. There is a general conclusion that Whitewater is the block to beat, but a simple rank of C/Ws at 1 gpm could be misleading without noting pressure drop.

OK – let’s keep pressure drop constant and let flow rates “float”:

C/W @ 0.5 mH20
WW 0.190 @ 1.0+ gpm flowrate
SW 0.210 @ 1.5+
Slit 0.215 @ 1.9
Tsu 0.262 @ 0.7

Same relative ranking, although positions could conceivably change. Note the obvious correlation to pressure drops.

We both felt that a good representation of performance would on the basis of C/W @ 1gpm, noting pressure drop, as it is an important piece of data that should be included (for reasons which now should be obvious). Users must take note of the “cost” of performance due to pressure drops, specifically as to waterpump selection.

Some concern has been expressed about using a different waterpump and its impact on results. I will argue that there is no difference due to the pump used. The reason is that 1 gpm through the test rig will yield the same waterblock psid regardless of how 1 gpm is generated – it could be from a pump x, y, or from a 100 gallon tank of water suspended 15 feet over the test rig using a gravity feed. I changed to the more powerful pump so that I could get to 1 gpm for some very high pressure drop waterblocks.

For each waterblock, psid @ 1 gpm is independent of the pump used. If this assumption is incorrect, don’t hesitate to point out the why and how.

It would then seem that I’m arguing for presenting more data, at different flow rates, than I’m currently showing.

I think I might be able to do this without undertaking a week’s worth of testing – right now, each block takes about two days to test. I am not getting paid directly for this work, and even if I did, I can’t afford a week per block.

This is a bit of a ramble and quite long, but I want to present data the best possible way within constraints I have to live with, so your opinions are of great value to me. I am on the road, so bear with my response time. Let the discussion fly!

Joe

PS: I hope my interpolations are OK!
 
good points joec
ur interpolations look fine, lol
did u use the graph for data? i couldn't find a data table :-/

i know this is a bit off topic but i was wondering why don't u try real world situations?
like taking the top 4 - 5 most used pumps and testing them on a block.
kind of a hassle.. but it'll be more fair and you'll be able to find the best solution for each block
 
JoeC, with regard to the more water-flow and radiators comment, I thought that you were aware that BillA had found a problem in the calculations that went into the graphs for the radiator heat dissipation roundup that you linked to.

Bill posted corrected graphs at his web-site:

http://www.thermal-management-testing.com/corrected heat dissipation.htm

These graphs do indeed show that higher water-flow is actually beneficial for many radiators, and more importantly, for the heater-core style radiators that many people use today.

I had often wondered why the OC.com review had remained out of date with respect to the corrected information from Bill. Perhaps the corrected results were not communicated to you after all?
 
Thanks for the heads-up on the corrections - that's one that slipped through the cracks and I'm sure I missed that somehow; now that I know. I'll correct it.

Re real world: It's always a question of time and effort - I have to noodle through how to best apply tests such as this to "usable" data.

And no tables - I read data off the graphs.
 
wrt single point testing:

below are the curves for the 3 sizes of tubing used with wbs
the pressure is shown in mv (to 0.001) as that is how I record it
of note is the curve for 1/4" ID tubing, such as for the Innovatek XX

connection%20correction%20curves.gif


this is the European approach to WCing, but one can see from these curves that the dp (for just the test connectors) at a flow rate of ~2.0lpm in 1/4" would correspond to a flow rate of ~6.0lpm in 3/8", and ~9.5lpm in 1/2"
- clearly the mfgrs of such wbs have no intention of pushing 1gpm (3.8lpm) through the system -> so why test it that way ?

based on my testing, and that of pHaestus; it would seem that a 'typical flow rate' could be characterized for WCing systems based on the tubing size

eyeballing the 2 groups (and note pHaestus had a paddle wheel flow meter in his loop):
1/2" would be 7 to 8.4 (WA) or 4.8 to 6.6 (DP), say 7.6lpm (2gpm) - specifically selected low flow restriction components
3/8" would be 4.3 to 5.2 (WA) or 2.6 to 3.9 (DP), say 3.8lpm (1gpm) - includes a low restriction GPU wb
for 1/4" I would suggest a MAXIMUM of 1.9lpm (0.5gpm)
-> the testing flow rate to be based on the size of the wb's connections

if the goal is apples to apples, what is the point of (single point) testing at a flow rate not likely to be encountered by the user of that particular component ?

food for thought anyway
comments anyone ?

be cool
 
wow... what a wealth of information by all the knowledgeable members.

I think its fair to use different pumps as long as the flow is restricted to give the same amount of flow rate to all the blocks which JoeC did. However, as somebody else mentioned, would you have to take pump heat into consideration or is that negilible.

As for testing higher than 1.0GPM, let's say that a high restrictive block has lower C/W than a low restrictive block at that flow... Wouldnt testing at higher flowrate just increase the gap between the low restrictive and high restrictive block? like shown in the WW and 5002 graph.

To recap I think testing at 1.0 GPM is failry reasonable as long as that same flowrate is fed to all WBs. If the flowrate was higher than the performace gap would just increase anyways. And since blocks are rated by people in relation to other blocks' performance, the exact value of C/W won't really matter THAT much, what matters is the where the block stands with its current C/W relative to other blocks...
 
As for testing higher than 1.0GPM, let's say that a high restrictive block has lower C/W than a low restrictive block at that flow... Wouldnt testing at higher flowrate just increase the gap between the low restrictive and high restrictive block? like shown in the WW and 5002 graph.

Be careful. You are correct that an equally increased flow would make the difference even larger, but when you remove the artificial flow restrictor, the actual flow increase will be less in the comparatively restrictive, lower C/W system than the less restrictive, higher C/W system. And, that difference is flavored by the parameters of the pump used. So, the reality is that the gap could grow, shrink, or even swap sides. A high pressure, low flow pump will tend to favor the more restrictive block when the cuffs are removed, particularly if the artificial restriction had to hold back quite a bit to keep it down to 1GPM. A low pressure, high flow pump (I believe most of these pumps fall into that category) will tend to favor the less restrictive block; shrinking the gap, and possibly even reversing it.
 
Hmm interesting point Omaticrail...

I do understand that in reality, pump parameters come into play however lets say that the user is using a Eheim 1040 and he has both WW and Swifty. Would Swifty perform bettter than WW with that pump? I realize that the pump would favor Swifty better however even in this situation, I think that WW would perform better than the 5002. I do realize the gap difference the two will decrease significantly, but WW will still perform better tho.(the WW and 5002 graph)...

What I was trying to say is the C/W are useful at making comparison with other blocks. If the gap between block X and block Y is some what significant then even when using a low performing pump, the gap would still be there, just less. I dont think it will swap and change places...
 
I hope I'm not making a huge error of thought, but...


Ok, agreed the PSID for a block does not change if you change the pump.

But won't the c/w be affected by changing the pump?

Surely if you use a pump with alot of horsepower compared to one which is weak, even if you make the flowrate the same, the more powerful one will be better able to sustain that flowrate through a restrictive block than a weak pump?

So the upgrade to the 1060 (aka 1260) just for the RBX block is unfair, and I still dunno why Joe did it!



No offence Joe, but why didya change the pump? :confused:
 
ILikeMy240sx said:
I do understand that in reality, pump parameters come into play however lets say that the user is using a Eheim 1040 and he has both WW and Swifty. Would Swifty perform bettter than WW with that pump? I realize that the pump would favor Swifty better however even in this situation, I think that WW would perform better than the 5002. I do realize the gap difference the two will decrease significantly, but WW will still perform better tho.(the WW and 5002 graph)...

Again, that question is best answered by BillA's own test data. Bill plotted block performance vs pumping pressure (the size of the pump). The following graph pretty much allows one to look at a point on a curve, and then look directly above and below to get a rough idea of performance with the same pump. There are some small variations due to the way that centrifugal pumps work, but fairly insignificant.

I believe that the following graph is about the singularly most useful piece data that Bill presents, meaning that if you could only choose a single graph to take from Bill's tests, then the C/W vs pressure drop is the one you would choose, rather than C/W vs flow, because it gives the best indication of block performance give the same sorts of pumps. I really believe that this is the direction that reviews should be taking.

summar5.gif
 
Cathar said:

I believe that the following graph is about the singularly most useful piece data that Bill presents...[snip]...because it gives the best indication of block performance give the same sorts of pumps. I really believe that this is the direction that reviews should be taking.

Yeah but idiots like me dont' get it! :p

I get the c/w versus flowrate and it makes alot of sense to me. In testing you just restrict the flow with a tap or whatever and then measure the temp and calculate the c/w.


But with that graph, how is the pressure drop changed for each measurement? Is a different pump used? And what is the flowrate for those measurements?
 
BillA said:
.......
comments anyone ?

be cool
Not sure.
However from Simulations the the choice of "1/2" would be 7 to 8.4 (WA) or 4.8 to 6.6 (DP), say 7.6lpm (2gpm)" is not an unreasonable Flow-rate choice.
Pump3.jpg

from here
Intersections [ red (WW) with purple(Eheim1250) at ~6.5lpm, and brown(Swiftech5000) with blue(Eheim1048) at ~7lpm] would suggest 6-7 lpm.
Several of the WW/Pump intersections have been roughly confirmed by Cathar's observations.May be of note that pHaestus's 1/2"System corresponds closely with the red (WW) simulation.

May have my stupid head on, but am confused by "7 to 8.4 (WA) or 4.8 to 6.6 (DP)".What are WA and DP? Numbers do not seem fit any units that I can think of.
 
Last edited:
Les
the flow rate #s from the tests in these links:
Swiftech tests aka WA
pHaestus tests aka DP
(our initials)

will_maltby
if you look at my graphs you will see 'dots' on the curves
from left to right each dot corresponds to a flow rate of 0.3, 0.5, 1.0, 1.5, 2.0, and 3.0 gpm
some curves will not have the higher flow rates because the pump could not achieve it

-> wb head loss "C/W"s over 1.0 mH2O or so don't mean much as the typical WCing pump does not have the capability of pumping much at those pressures (given that the rest of the WCing system will have additional flow resistance as well)

be cool
 
Les,

Please check me on this.

Been thinking on the problem of single value rankings for blocks. I definitely like BillA's more comprehensive data better, but if the data must be reduced to a single value...

I think there is a way to give a single C/W that is a much better indication of the block's performance in a system composed of an Eheim 1048, a Chevette HC, and about 6 feet of tubing. (Or any other real system.)

The following graphs show curves relevant to such systems for a White Water block and a Swiftech MCW-5000 block. (These are somewhat near the flow resistance extremes for high performance blocks available today.)

fair1048.gif


The light blue line shows the value dP*Q/5 for the Eheim-1048. (The division by five is just to scale the value for graphing purposes.)

By drawing a line up, from the intersection of the pump curve and the 'loop' curve, to the dP*Q/5 curve, and then left to the vertical axis, it can be seen that dP*Q/5 is roughly:

1.025 for the MCW5000
1.08 for White Water

There is about a 5% difference in the value of dP*Q for the two blocks in the system although there is about 33% difference in the flowrates. Any block with a flow resistance between an MCW-5000 and about 28% higher than a White Water is going to be within about a 6% range of dP*Q.

So, how does one make use of this?

First you need a device which provides a flow resistance equivalent to a Chevette heatercore and some tubing. You could actually use a Chevette heatercore and tubing, or you could take a 2" length of 1/2" diameter brass rod, and drill the correct size hole down the length. (Probably best to chamfer the ends to minimize turbulence generated at the inlet and outlet.)

With this device in series with the block you adjust the flow so that the flowrate multiplied by the dP across the series combination of the waterblock and the resistive device is equal to 5.25 lpm*mH2O. [ ((1.025 +1.08)/2)*5) ] You then measure C/W at whatever the flowrate happens to be.

The flowrate can vary substantially depending on the block tested, and yet the single C/W measured will be a much more meaningful number for comparison purposes.

Edit:

I should note that this works out particularly well with an Eheim 1048. With an Eheim 1250 there is a difference (for the two blocks discussed) of 42% for flowrate and 21% for dP*Q. I still think the technique described would result in better 'single point' comparison numbers though.

Second edit:

Doing the above, except with an Eheim 1250, would give White Water about a 4% gain in flowrate over what it 'should' have and the MCW-5000 about a 3% penalty in flowrate. Not really enough to mean anything IMO.

fair1250.gif


I'm not explaining this one other than to say it illustrates the point of my second edit.
 
Last edited:
Since87 said:
.....Been thinking on the problem of single value rankings for blocks. I definitely like BillA's more comprehensive data better, but if the data must be reduced to a single value...

Will give some thought after coffee.
However,of top of the head, it does remind me of the concept of using a single term something-like "PaC/W"(Hydraulic Resistance"x"Thermal Resistance) to describe a wb's performance http://forum.oc-forums.com/vb/showthread.php?s=&threadid=89986

Bill
Ta for explanation.
I do note that the "7 to 8.4 (WA)"* does fit my simulated curves(MCP600 now included) except for the enigmatic Eheim1250 - notwithstanding the different radiators.

* http://www.swiftnets.com/Technical/RealWorld-Pump-testing.asp
 
Last edited:
Sean
Find no flaw in manipulations.
Yes, I think measuring C/W at a constant dP*Q would give a practical index of performance with centrifugal pumps.However,I think, it is "wb and centrifugal pump combo" testing.
Additionaly ,I think, our understanding of the basic physics/engineering has a long way to go before this kind of manipulation can be envisaged as a "centrifugal pump"/wb testing norm.
 
since87
I like 'difficult' things, and have no quibble with your manipulations
-> but how many readers do you think followed that through ?
or should I ask: who is your intended audience ?

a relationship I have longed to reduce is that between °s of 'cooling', and cost, and noise
this could be applied to each part of any cooling system, as well as to the complete system

interested ?

be cool
 
BillA said:
since87
I like 'difficult' things, and have no quibble with your manipulations
-> but how many readers do you think followed that through ?
or should I ask: who is your intended audience ?

I expect you, Les, pHaestus, Cathar and myv65 (and a few 'mostly lurkers'), can follow it with a bit of effort. (A lot of that effort due to my somewhat terse and disorganized explanation.)

I don't know JoeC well enough to guess how well he might follow it.

I do think that it is worthwhile, for anyone wanting to present 'single point' numbers as a meaningful basis of comparison, to put the effort into understanding it.

In actual testing implementation, it is not a big added difficulty. It means that you have to dial in the flow to the point where dP*Q is at the target value rather than just Q. A bit more troublesome on initial test setup. (I am assuming there is zero 'drift' to deal with in the course of a testing run, but I can't imagine it is enough to have a meaningful impact on the result.) Test reports should indicate that the testing was done at the flowrate expected when the block is used with pump X, heatercore Y, tubing Z, etc. A link to a clearer explanation for the testing rationale could be provided.

The end result is that:

You have a test procedure that is slightly more complex on initial setup.

You have a document explaining the flowrate selection for the particular block tested.

You have a single C/W number for each block that is a reasonable (or at least explained) basis for comparison. Assuming you choose a good midrange dP*Q target point, the relative rankings of the blocks is unlikely to change over a fairly broad range of centrifugal pumps, rads, etc.

I guess I was mainly trying to get this idea across to 'the experts' in hopes they would agree that this sort of testing made sense. If 'the experts' agree this is a reasonable way to produce creditable single point numbers, JoeC and other testers who only want to generate a single data point could use this methodology to make that point meaningful. (And receive less criticism.)

Don't get me wrong. I MUCH prefer the kind of test results you have produced and I don't really want to encourage testing at a lower level, but if a tester insists on single point results, I think this is the way to do it.

BillA said:

a relationship I have longed to reduce is that between °s of 'cooling', and cost, and noise
this could be applied to each part of any cooling system, as well as to the complete system

interested ?

I would think the (going nowhere) simulator would be fundamental for this. I'd also think that reliability would be a key component of such an analysis, and that's very difficult to quantify. I think without reliability thrown in, the analysis would point to Maxijets (or some other cheap pump) as the pump to go with.

Sounds like more work than I'm up for without a paycheck.
 
Back