hahahaha
750W Thermaltake Modular PSU
DFI UT X58-T3eH8
Core i7 920 @ 20 X 200 1.325V
CORSAIR XMS3 DHX 4GB (2 x 2GB) DDR3 1600
768 MB EVGA 8800GTX
1 X 36GB WD Raptor
2 X 150GB WD RAPTORS
1 X SpinPoint P Series SP2504C 250GB
1 X Maxtor 6L300S0 300GB
16 X NEC DVD Burner
7 120mm Yate Loon LED Intake Fan
4 120MM Yate Loon Exhaust Fan
28" HANNSPREE Monitor
Watercooling Loop:
1 X PA120.3
1 X PA120.2
2 X Laing DDC's w/EK-DDC Dual Turbo Top
7 X Yate Loon Blue LED Intake Fans
4 X Yate Loon Blue LED Exhaust Fans
1 X Swiftech GTZ
1 X GPU EK Fullcover Waterblock
1 X XSPC Dual Bay Reservoir 5.25" with Bubble Window
You're actually quite correct, but there are caveats. The rate of thermal dissipation is proportional to the difference between the water and the air. If the water 20C warmer than the air, assuming all else is equal (air-flow, water-flow, air-inlet temp, etc) then the radiator will dissipate heat at twice the rate of if the water was only 10C warmer than the air. All that alters, theoretically, is the scale of the final result.
There is a caveat to that though. The higher the water temperature, the more "runny" water is. At 85C, the flow resistance of water through something like a flat radiator tube is around 35% (!!) of what it is at 25C. This only further brings into question the discrepancies of the liquid pressure drops stated. It also may alter, enough, the flow characteristics of what's going on in each radiator according to their internal design.
Also, the very high temperature of the water will heat up the air enough to actually influence/alter the air-flow resistances by a significant amount (~10%). Once again, this may alter the overall air-flow characteristics.
I don't doubt that the relative results at the same sorts of air-flows, but given smaller air-water deltas, would change too much from what's given. We should be aware though that there are enough effects at play here that would certainly have the potential push the relative results around in a significant fashion.
What the end effect would be, no one could say without testing. It could go either way, or it could have no net influence at all.
Suffice it to say, the air-flows, the pressures required to push those air-flows, and the air-water deltas, really are so significantly outside of the normal range of PC operation, that even general given assumptions about the relative stability of air-water delta relationships and their (linear) impact on radiator performance may no longer be true at such extremes.
[Edit: Been running some simulations with the flow calculator. It seems that the difference between 25C and 85C water is enough at 5LPM flow rates to conceptually alter whether or not the water is flowing in a laminar or turbulent flow fashion, depending on the tubing shape. It is actually quite a real possibility that one design may benefit in a significant fashion from dramatically higher water temperatures while another does not.]
Last edited by Cathar; 06-17-2007 at 05:18 AM.
Thank you, broadening my horizon once again.. Considering air and water viscosity and the flattening performance gain profiles of radiators with increased water and air flow, high temperature tests seems to favor restrictive designs more.
First off...I would like to say as an "aspiring Liquid-cooler" I found most of this discussion to be both thought-provoking as well as educational. Notwithstanding the lessons in thermal capacitance and fluid dynamics there was a fair amount of information parsed out throughout the dialogue. (Though the flaming of even the most ignorant statements still leaves a bitter taste in my mouth...though that could be the gin).
The active participation of mfg reps and cathars candid and extremely well researched input is not only welcomed, but comforting in the sense that their are knowledgeable assets merely a few keystrokes away.
The one thing that I have found lacking, however, is the absence of any commitment to "pc suitable" benchmarks and/or trials. I feel that HWlabs, Marci, Cathar are missing an opportunity to eradicate erroneous claims with empirical data. Perhaps something is in the works, I will most certainly be watching this thread with anticipation.
In the end, I would just like to thank all active participants in this lively discussion as it has proved both stimulating and educational (if not entertaining)
____________________________________
"Le Rubix n00b" Opty 175 ccbbe 0617 FPMW||2*1gb gskill HZ pc4000||DFI UT LP SLI-DR EXPERT (HR05 cooled)||evga 8800gts 640(flashed to KO edition)||X-fi xtremecrapper(gamer)||Samsung 183L sata DVR||2*320gb WD Cav RAID 0; WD 80gb IDE||TR ULTRA 120 w/fm121 push.cfg
"Le Iron LLama" Venice 3500+ (chipped ) 2.8ghz 11*255 1.42v||Asus A8N-Sli deluxe (premium bios 1302)||2x512mb Value crap 200.7mhz 3.3.3.6 1t @ 2.85v||2x7600gt 635/1.5||Ultra Xfinity 500w||WD:500gb sata300||2xLG 18xSuperdrive||Lapped Zalman 9500led (modded w/FM92silverstone)
"Taking Noobage one day at a time"
All this talk, and so little information...
(Hi Cathar and Marci, long time!)
I think everyone understands that the results are skewed. The problem, as I see it, is that the average Joe could infer that, if the performance at one set of parameters shows one product as being superior, then it must retain its superiority under "slightly" different parameters.
Which we know is false, but only a rare few understand why.
I've done a fair amount of work in water block design (theory) and have been following radiator design, because a lot of the principles are very much the same (just in a different order).
So here's my attempt at an explanation, and hopefully this can be complemented by our other "rare few" members:
First, what we start with:
Testing air flow: 330 cfm
Testing coolant flow: 5 lpm, 10 lpm
Testing air temp: 25 C
Testing water temp: 85 C
Then, what would be ideal:
Desired air flow: 80 cfm
Desired coolant flow: 5 lpm
Desired air temp: 25 C
Desired water temp: 30 C
In short, a quarter of the airflow, and a much lower water temperature.
So first, we break down the radiator function into two components, then focus on the elements that make a difference, for the purposes of identifying the testing flaw.
1) Heat transfer from the coolant, to the radiator.
2) Heat transfer from the radiator, into the airflow.
In 1 and 2, I'll define the variables as:
a) flow rate
b) flow geometry
c) flow turbulence
d) temperature
e) surface area
f) surface area effectiveness
I believe that the core of the flaw in the tests reside in item f, so I'll expand: while it's always desirable in a heatsink to have large and plentiful fins, it does not mean that infinitely long fins add any significant cooling, as the variable I've dubbed "surface area effectiveness" depends on d: temperature, or more specifically, how the temperature differs, between the mass of the fin, and the air flow. In other words, you're not going to transfer a lot of heat from a mass (fin) to the air, when the temperature is just about the same.
To make this clear, I'll exagerate: if you were running a heatsink on a CPU, it would dissipate just about as much heat, as it would if the fins were insanely long, like 1 meter/3 feet (fan aside). Why? Because *most* of the heat will still be dissipated at the bottom of the fins, regardless of how long they actually are, unless of course, the original heatsink is poorly designed, with fins that are too short.
The fins on a heatsink, correspond to the fins in a radiator. In a compact design, like the Black Ice and Thermochill units, the "fin length" is just right for the intended purpose. The Koolance is actually larger than it needs to be, for water cooling (but that's ok, because longer fins won't have much impact, other than ending up with a larger radiator than needed).
Now we put the same water cooling rads through a testing sequence, where the coolant temperature is too high, and what do we have? A huge difference. Why? As I mentionned above, a heatsink is considered poorly designed when the fins are too short. When you test a water cooling radiator under automotive conditions, you end up with the same "poorly designed" heatsink: the Thermochill and Black Ice unit's fins (and fin effectiveness) are too short, and are overwhelmed, as any poor designed heatsink would be.
The "correct" conclusion from Koolance's testing is:
-The Black Ice and Thermochill radiators make poor automotive heatercores.
And that's why Koolance's contracted testing is irrelevant.
I wouldn't even go on to say that Koolance's testing was accurate.
There is simply very little sense when you consider higher air and waterside pressure drops are yielding over 50% more thermal capacity with a heat exhanger of the same size. No heat transfer engineer would stake their reputation here let alone have such results freely released into public domain.
I guess there might be a reason for the notation regarding the limitations of the usage of the said certification.
Koolance radiators are supplied to them by a manufacturer whom they do not control. They also have a track record of mistranslating results and misrepresenting them for the sheer purpose of depracating their competitors.
In theory, the shorter the fins the more efficient the fin, as the temp delta from the tube to the center of the fin, i.e. the conduction of heat is more efficient given the relatively shorter distance the heat has to travel. Having more material in this respect tends to only make a costlier and less efficient unit. The only problem with having shorter fins is that you end up having more tubes that reduce the overall frontal heat transfer surface area and cause higher air pressure drop. So one would need to introduce a more aerodynamic tube geometry.
At best, what we can see here is a clear obfuscation of facts. Koolance's rebuttals are nitpicking at worst and dismissive at best. Previously they claimed that oxidation on brass/copper radiators retarded performance, now they are completely ignoring that assertion but instead maliciously insinuating naturally occurring oxidation is a manufacturing anomaly.
The tests they conducted are in no way shape or form representative of the application they intend to address and their presentation of inapplicable data, pending verification, are just as misleading.
Last edited by hwlabs; 06-17-2007 at 01:58 PM.
It's really frustrating, I know.
Some of the statements advanced by Koolance are completely wrong (read false), such as the glue (which is more obvious), then the oxidation of copper (not so obvious until properly refuted). In fact, they're so wrong... that it leads me to think that Koolance is not just clueless, but intentionally trying to deceive people, and that's *not* "K"ool. Where's the effort of honesty?
I most appreciated HWLabs reply, in which I learned the practical thickness of copper versus aluminium, when designing a rad; this is the kind of information that shows that you guys *really* know what you're talking about.
I believe that my explanation above (while simplistic) does cover why the results are off: the Koolance can handle a larger heat load, that's what it was originally designed to do, while the other rads tested came up short, because they were overwhelmed by the testing conditions, which had nothing to do with water cooling.
I've been getting ready to do some rad testing myself, but hit a wall with the flow nozzles, used to measure air flow with accuracy, as they cost 300 to 600 USD. Otherwise anyone can put together a test rig, with a couple of 55 gallon drums; it's really not hard. The sensors do have to be configured correctly, and it does look like (at first hand) that Katech is doing an OK job there (but somehow messed up the hydraulic measurements - liquid side): I noted the proper placement of the temperature probes, but also that they appeared to be RTD elements, which have a slow response time; ok under steady temperature testing, but won't be able to track temperature fluctuations in any kind of usefull way. So Katech would be logging the temperature readings, and performing an average. I would have opted for small thermocouples, ones with a low mass, so that temperature fluctuations could be monitored.
There's also no effort to control the inlet air temperature, which would be expected for heatercore testing: the inlet air temperature fluctuations would have a negligeable effect on the deltaT (i.e. coolant to air temperature difference). I.e. a couple of degrees on a difference of 60 degrees is a 3% error margin (negligeable). In water cooling, 2 degrees on a deltaT of 5 degrees, is a 40% error margin (unacceptable, useless). [note that I'm not stating that Katech testing is 3% accurate: their accuracy is a lot lower than that, when the whole testing rig is considered].
In water cooling, the inlet air temp has to be controlled. Since Katech does not have air temperature controlled, they do not have the capability of testing rads for water cooling.
[for the benefit of everyone]
The process of controlling an air temperature is actually a lot more complicated than it seems. When I looked into this, and thanks to much information from Bill Adams, I originally thought that it would be possible to use an ambient air supply of ~20 deg C, and heat it up to 25 deg C, in a steady way, but I turned out to be wrong; no combination of fast-acting temperature sensor, and heating element can actually achieve this.
Instead, the air has to be first cooled, to create a large enough delta T to be registered by a temperature probe, then re-heated to the desired temperature. The cooling portion would ideally cool the air, by a factor of three, over the device being tested. In other words, if I'm testing a radiator that's going to dissipate 200 Watts, I have to apply 600 Watts of cooling power (done through phase change, aka "refrigerator"), then re-heated (proportionally) by, in this particular example, 400 Watts, in a tightly controlled manner (a good temperature controller, or computer).
One then simply has to put a radiator, with the usual combo of fans inside the chamber, and route the coolant hoses out, where the coolant temperature is controlled, in exactly the same way (cooled, then control-heated).
Bill's done it with a modified incubator. I've drafted plans to do it, building my own environmental chamber. I lack the resources to complete it at this time, but I still gnaw at it, when I can. The power usage is just astonishing, and remains a problem that I have yet to get around. What I did resolve, is making sure that the heat that my proposed rig generates, doesn't flood the room: it's going out the window .
Last edited by BigBen2k; 06-17-2007 at 02:47 PM.
Ben, I'm tending to lean to being in agreement with hwlabs on this one.
I'm in two minds about dismissing results without knowing the full details of the test procedure. As I highlighted with the water temperature and its effect on viscosity, the liquid flow resistance through the radiators at 5LPM, with 85C water, should be around 1/3 of what it is at ~25-30C. Instead we have liquid flow resistance values that are >6x that of measured values from other tests, and that's not even factoring in the decreased liquid resistance from the temperatures. We're talking liquid flow resistances of 15-20x of that which we'd expect.
Now we get to ventilation (air-flow) resistance. The tests would indicate that GTStealth and Thermochill radiators are of similar air-flow resistance. Now, without making any judgement as to the validity of the different design approaches clearly taken between HWLabs/TC, if anyone has ever seen the two radiators side-by-side, it's pretty obvious that the PA rads have focused on the lowest possible ventilation resistance and attempting to scavenge what heat it possible from an increased air-flow, while the GTStealth approach is to scavenge the maximum heat possible from a decreased air-flow. Truly, I am not trying to make a statement here on which is better, just that it's obvious that's the respective approaches taken. One does less with more air, the other does more with less air.
Running through an air-flow resistance calculator, which has proven to be fairly accurate for me in history thus far, if not from an absolutely correct standpoint, then at least from a relative standpoint. Calculating all the parameters for each radiator design, I arrive at the following:
GTS240 @ 320cfm => 335Pa air-flow resistance
PA120.2 @ 320cfm => 84Pa air-flow resistance
Once again, I'm not trying to state which is better. There are pluses/minuses to each approach. What I am trying to point out is the obvious discrepancies between the obvious structural differences, the measured (I know for a fact that the PA120.2 is one of the least air-flow restrictive rads on the market by a long measure), the theoretical, and then ultimately we have Koolance's results, which simply don't make sense.
Ben, I don't even want to begin to speculate on the design reasons for radiators on the basis of Koolance's test results. They are simply too incongruous to even begin to have faith that they truly represent what is going on, much less attempt to draw conclusions on the results.
Last edited by Cathar; 06-17-2007 at 03:02 PM.
Right; there are many design variations possible in radiators, each optimized for one variable or another. You know what radiators are about. [I didn't even consider water viscosity, because I never had to, until now].
Koolance is wrong on so many levels, it's not even funny. On one side, there's the marketing PR, and on the other there's the flawed technical info. I'm not a people-person, so I'm not going to venture an opinion on why Koolance is doing this, but I will gladly take on the technical innacuracies.
I just want to make sure that people can understand why the results posted cannot be interpreted as being applicable, under different conditions. In simpler terms, I want it to be known that the test results posted are no indication of how any of the radiators would perform for water cooling.
Of course it's obvious when one can demonstrate the difference under the proper test conditions, but only a few people have the ability to do that.
85 deg C would be the typical coolant temperature for my car. If my PC had an 85 deg C coolant temperature, I'd expect my CPU to suffer terminal heat damage, or throttle to a crawl.
If Koolance can use different testing conditions, then lets see some tests with ridiculous figures: 125 deg C coolant temperature (oil would have to be used), and 600 cfm airflow: that should make Koolance "The Supreme King of the Hill"...
It would be interesting to speculate on the automotive application their rad would match.
It seems small for most engine rads, although possibly not motorcycle. Could it be an ancillary cooler, oil or transmission?
It's a heatercore: the internal radiator of a car, designed to keep the cabin warm, in winter.
hey ben, can I have my swissflow?
your rig? I know a long time ago you mentioned that you were going to attempt to calibrate the swissflows with your mag flow meter. Did that ever happen?
I was going to compare it with my turbine flow meter; I don't have a mag flow meter. In fact, I've had a hard time even finding a mag flow meter that works at low flow rates. You have PM.
Ben, just to be clear, the more that I, and it appears HWLabs as well, look into this, the more we're both feeling that the results might not even be valid, even for the testing conditions given.
Then we have Marci pointing out the legal disclaimers attached to the results, which basically in a short-term summary says that the results cannot be trusted to be accurate for legal purposes, and perhaps even more telling, should not be used for advertising purposes. Heck, you've got to be asking questions as to the validity of results when the tester is basically saying that they don't want to have their name or reputation publically associated with the results.
Normally when independent people test stuff, they're quite happy to state that they stand by their results and believe them to be accurate within the best of their ability. Quite unusual for an independent tester to put a disclaimer in that effectively invalidates the veracity of the results.
Just a funny obvious thought that we missed out in all this brouhaha...
WIth a 350 cfm fan, why would you even need watercooling?
Core i7 920 D0 B-batch (4.1) (Kinda Stable?) | DFI X58 T3eH8 (Fed up with its' issues, may get a new board soon) | Patriot 1600 (9-9-9-24) (for now) | XFX HD 4890 (971/1065) (for now) |
80GB X25-m G2 | WD 640GB | PCP&C 750 | Dell 2408 LCD | NEC 1970GX LCD | Win7 Pro | CoolerMaster ATCS 840 {Modded to reverse-ATX, WC'ing internal}
CPU Loop: MCP655 > HK 3.0 LT > ST 320 (3x Scythe G's) > ST Res >Pump
GPU Loop: MCP655 > MCW-60 > PA160 (1x YL D12SH) > ST Res > BIP 220 (2x YL D12SH) >Pump
Testing done in Korea, by The "Korean Automotive Technology Institute". Honestly, I don't even know who they are. For all I know, this is a school, and a bunch of students did this work, dirt cheap.
http://www.katech.re.kr/eng/index.asp
They have a pretty website. Note how there is no reference to radiator testing, in their listed testing capabilities; it's all engine related, but they couldn't have this testing facility for nothing.
http://en.wikipedia.org/wiki/Korea_U..._and_Education
shows a cooperative agreement with the Korea University of Technology and Education .
Another collaboration, with DFR:
http://www.dfrsolutions.com/newslett...Newsletter.pdf
The disclaimer is unusual.
Granted that the viscosity effect wasn't considered. Granted that the pressure measurements could be off. I'm not convinced that Katech even knew what they were testing. Either way, dissipating 6 to 9 kW isn't anything like dissipating 200 Watts, so we're already passed the point of determining wether or not the data has any value (it doesn't).
I don't have any experience dealing with Asian companies, but I would expect it to be hard to get them to do something, unless you pay them a hefty fee, and I seriously doubt that Koolance forked over a lot of money for this. It's woefully inadequate information, but if it was free/cheap, and Koolance could get it, I wouldn't be surprised to see them making an effort to make it public. I'm leaning towards Koolance coming across this data, as it was volunteered by their supplier: note the name of the "applicant" (Park Jae-Sung) is Korean, not American. Unfortunately, this is the kind of background information that we never find out about, but would be oh so revealing. Why Koolance would go through the effort of defending this information is puzzling.
Just wanted to through in some links. This discussion has been going on for years:
The aluminum bias:
http://www.stockcarproducts.com/rad2.htm (mainly geared towards large radiators)
The copper bias:
http://www.usradiator.com/testing.htm
http://www.copper.org/applications/a...clability.html
And not the only ones who have been arguing over this question:
http://www.eng-tips.com/viewthread.cfm?qid=61247&page=7 ( from some years back)
http://forums.nicoclub.com/zerothread/247182 ( present argument, even brought koolance into it)
.....
All mighty confusing, but I think I will go with the real world tests. I'll be sticking with the copper until I get some actual (installed in a system) tests done using copper rad vs aluminum rad of the same size.
Dateranoth
EVGA X58 mobo
6gb RAM 1600mhz cas9
Core i7 w/ D-Tek FuZion @3.9Ghz (43C idle) (65C load) (27C ambient)
EVGA 680GTX SC w/ MCW-60 and stock Heat Spreader (With some Modification) (32C idle) (53C load) (27C ambient)
Water Cooling Loop:
1/2" Tygon Tubing
Laing DDC w/ Petra'sTech DDCT-01s Top >Swiftech MCR-320 w/ 120mm Yate Loon D12SH-12 x3>D-TEK FuZion>MCW-60>Swiftech MCRES-MICRO
My Worklog
Shop Petras
If anyone wants to buy the Koolance radiator separately, for testing purposes, it can be done so here:
http://www.koolance.com/shop/product...roducts_id=241
Interesting. Plugged in the Koolance specs into the air-flow resistance calculator, and actually arrive at a value consistent (within 3%) of what is stated for their radiator. For the PA/GTS, the calculated values and the reported values disagree by a huge amount.
OK. Got them now. They're lying, unless they've managed to change the laws of physics.
Their results are impossible.
Air @ 25C has a density of 1.169kg/m³
Air has a specific heat capacity of 1012J/kgC
Air therefore has a thermal capacity of 1.169 x 1012 = 1183 J/m³C
Air-flow through their radiator is 5m/s, with an orifice area of 0.13m x 0.24m
That's 5 x 0.13 x 0.24 = 0.156m³/s
Therefore, the thermal resistance of the air-flow is:
0.156 m³/s x 1183 J/m³C = 184.55 J/sC
A Watt is a Joule/sec
Therefore the thermal resistance of the air is 185.55 W/C, or inverting, a C/W of 0.0054186
i.e. for each watt of heat energy dissipated into the air, the air will warm up by 0.0054186C
Koolance claim a heat dissipation of 9.62kW, or 9620W
9620 x 0.0054186 = 52.13C
i.e. to dissipate that amount of heat, the air MUST have risen by 52.13C above the inlet air temperature.
So the exhaust air MUST be 24.46 + 52.13 = 76.59C
This is basic thermal physics here.
Here's the crunch part. Their results claim that the water discharge temperature is 56.20C
i.e. In a total breakdown of the laws of thermodynamics, the water discharge temperature has somehow managed to exit the radiator at 20C less than the air discharge temperature.
THAT is impossible.
Last edited by Cathar; 06-17-2007 at 06:41 PM.
Bookmarks