The consensus of 90 degree elbows on older w/c hardware

D

DarkenReaper57

2[H]4U
Joined
Oct 2, 2003
Messages
2,173
There has been some heated debate over the impact of tubing size given an optimal pump+radiator+block setup, mostly in the thread started by Cathar. If memory serves me right, 90 degree elbows came up in that discussion as well, and their overall impact seemed to be negligible - something about 0.1C net increase for every elbow.

I'd like to take this time to thank Cathar for taking his own time and putting together all of the work he has done on the matter. I know he has received quite a bit of heckling, but he means well and simply wants to expose the myths.

That said, the research done was using more modern equipment, and am not sure whether or not it applies to my particular setup. My rig is in my signature, but here are my w/c components:

Pump: AquaXtreme 50Z
Tubing Size: 7/16" ID
Blocks: D-tek WhiteWater CPU block, Older modded D-tek GPU block
Reservior: Criticool Waterplant 6" with 1/2" ID barbs
Quick Disconnects (2) - Made by Colder
Heat Exchangers: (1) 77' Bonneville Heater core in series with (1) 86' Chevette Heater core
External: Yes

As you can see, I am using fairly old hardware, but the system works quite well - temperatures are in my signature. I put a lot of effort into building this sucker, so I honestly don't want to get a Thermochill 120.3 or anything like that, since it would require a TON of work to fit it in. It also would cost a lot of money.

Here is a link to my worklog (which will be updated very soon!) http://www.hardforum.com/showthread.php?t=1112694&page=3 that has images of the external system. I am having a tube routing issue above the pump area and would like to use anywhere from 2 to 4 90 degree elbows. How much would these affect my temperatures? I'd imagine that even if my flow was cut a little, temperatures would barely change - heck, the change would probably be well within the error margin of the measuring equipment, right?

Any positive input would be great. Don't go and say get 3/8" ID tubing as that would require way too much work (taking it all apart, sweating new barbs on the heater cores, etc). The system performs very well as it stands imho, so I don't want my CPU temperatures to jump up 2-3C overall.
 
Good job on that watercooling box! - also nice to see the old WW still beeing put to good use.

I would say go right ahead with the 90 angles if it helps - won't make a noticeable difference in my experience. What you can do to minimize the effect though, is to use oversized elbows, with an inner diameter the same, or even a little larger, as your tubing ID. So either use metric 12mm elbows or 1/2" if you can fit the hose on them - then there will be basically zero difference in flow or temps.

;)
 
DarkenReaper57 said:
There has been some heated debate over the impact of tubing size given an optimal pump+radiator+block setup, mostly in the thread started by Cathar. If memory serves me right, 90 degree elbows came up in that discussion as well, and their overall impact seemed to be negligible - something about 0.1C net increase for every elbow.

I'd like to take this time to thank Cathar for taking his own time and putting together all of the work he has done on the matter. I know he has received quite a bit of heckling, but he means well and simply wants to expose the myths.

That said, the research done was using more modern equipment, and am not sure whether or not it applies to my particular setup. My rig is in my signature, but here are my w/c components:

Pump: AquaXtreme 50Z
Tubing Size: 7/16" ID
Blocks: D-tek WhiteWater CPU block, Older modded D-tek GPU block
Reservior: Criticool Waterplant 6" with 1/2" ID barbs
Quick Disconnects (2) - Made by Colder
Heat Exchangers: (1) 77' Bonneville Heater core in series with (1) 86' Chevette Heater core
External: Yes

As you can see, I am using fairly old hardware, but the system works quite well - temperatures are in my signature. I put a lot of effort into building this sucker, so I honestly don't want to get a Thermochill 120.3 or anything like that, since it would require a TON of work to fit it in. It also would cost a lot of money.

Here is a link to my worklog (which will be updated very soon!) http://www.hardforum.com/showthread.php?t=1112694&page=3 that has images of the external system. I am having a tube routing issue above the pump area and would like to use anywhere from 2 to 4 90 degree elbows. How much would these affect my temperatures? I'd imagine that even if my flow was cut a little, temperatures would barely change - heck, the change would probably be well within the error margin of the measuring equipment, right?

Any positive input would be great. Don't go and say get 3/8" ID tubing as that would require way too much work (taking it all apart, sweating new barbs on the heater cores, etc). The system performs very well as it stands imho, so I don't want my CPU temperatures to jump up 2-3C overall.
Click to expand...

As long as the elbows aren't near the pump itself, the theoretical impact should be around .1 to .2C for each. However, on one of my show builds, I was forced to use two elbows, one of which was close to the outlet of the pump and my temps rose by 1.2C.

A few other people conducted real world tests and each elbow added the same pressure drop of about 4 to 6 feet of tubing depending on the test.
 
Ranker said:
However, on one of my show builds, I was forced to use two elbows, one of which was close to the outlet of the pump and my temps rose by 1.2C.

A few other people conducted real world tests and each elbow added the same pressure drop of about 4 to 6 feet of tubing depending on the test.
Click to expand...

How about some pics to see what you did on your show builds? Links on the real world tests please. :)

I think the pressure drop must be somewhat related to high pressure pumps because a lot of people using low pressure pumps don't see these kinds of results from using 90 or 45 degree elbows. My Feeding Frenzy and Feeding Frenzy Interlude projects have lots of 90's and they cool just fine. Others I know who have used a lot of 90's report similar results when using similar pumps as I do. :)
 
Top Nurse said:
How about some pics to see what you did on your show builds? Links on the real world tests please. :)

I think the pressure drop must be somewhat related to high pressure pumps because a lot of people using low pressure pumps don't see these kinds of results from using 90 or 45 degree elbows. My Feeding Frenzy and Feeding Frenzy Interlude projects have lots of 90's and they cool just fine. Others I know who have used a lot of 90's report similar results when using similar pumps as I do. :)
Click to expand...

Well, if you have a higher pressure pump with a given setup and use a 90 degree elbow, there will be a greater pressure drop across that elbow than a lower pressure pump and that same given setup with a 90 degree elbow. This is basic fluid mechanics.

The question at hand is the impact on overall cpu temperature when using 90 degree elbows in the setup I described. Of course the effect of 90 degree elbows will be fairly minimal in your Feeding Frenzy setup (which is beautiful btw!) - you are using hardware that is more immune to such pressure drops and is designed around a lower-flow methodology. Not to say that is bad, just different. I honestly don't want to start a huge large bore/small bore/low-flow/high-flow argument here though :p.

That said, if I were to use one of the pumps you have in your setups with mine (the eheim 1046?) it would give me worse performance - my blocks are designed for a pump with high head pressure. When I upgraded from an Eheim 1250 to the current AquaXtreme 50Z, I saw a net overall 2-3C decrease in temperatures. No, I don't have pictures or raw data to back it up as the switch was conducted over a year ago, but there is no reason for me to lie in a thread when I am trying to find honest accurate answers.

So, what is the effect of 90 degree elbows on this particular setup? I don't think I'd see much of anything as long as the pump isn't starved and the elbows are kept to 4 or less. If they are incorporated I'll take some measurements and such.
 
Geez,

If you all are going to spout off about pressures and pressure drops, get it right. Pumps DO NOT create pressure. Restrictions, such as elbows, poorly sized pump to tubing size, and blocks create system pressure. The pump has a flow/pressure curve. Think it of as flow/restriction curve.

Now, if you all were saying high FLOW pumps, then all would be cool. The higher the FLOW, the higher the system pressure drop, if everything else remains constanst. That is simple flow dynamics. And, please, do not ask me to blow the dust off of my Fluid Dynamics book to prove simple concepts. If you do not understand them, take a few ME classes.

Elbows in a WC system will add pressure drop. I have never done any calculations in a WC system, and I do not really care to. My "educated" guess and based on my work "experience", there is not going to be a major differance, from tubing size and # of elbows, to the flow rates involved in WC'g. As long as one keeps such restrictions to a reasonable minimum, and you have a decent pressure head reserve on the pump.

I do have to agree with Ranker. You want to keep restrictions in the suction (pump inlet) line to a minimum. This would be HIGHLY unlikely in WC'd setups, but cavitation can literally destroy a pump very quickly. That is why you always see, "do not run pump dry".
 
Now see this is where I have the problem about all this pump starvation business. I think it is more involved than you say and is related to the pump AND the tubing used. My Aquastream pump doesn't have this problem and I use PUR tubing. The coolant coming from the Aquatube goes through a pair of 90 degree fittings right into the pump intake.






FFI_263.jpg







FFI_052.jpg
 
I use 4 90 degree elbows in my loop AND turn the pump down to "2" (MCP655) without any harm...
 
JC634 said:
...Pumps DO NOT create pressure.
Click to expand...

Correct me if I am wrong, but a pump creates a pressure differential across its inlet and outlet. This is the mechanism that allows for flow of water and a flow/pressure curve to exist. A given flow for a particular setup depends on the overall system restriction. This given flow can be found on a pump's flow/pressure curve. Is my reasoning flawed here?

I do agree with everything else you said, though :D .
 
DarkenReaper57 said:
Correct me if I am wrong, but a pump creates a pressure differential across its inlet and outlet. This is the mechanism that allows for flow of water and a flow/pressure curve to exist. A given flow for a particular setup depends on the overall system restriction. This given flow can be found on a pump's flow/pressure curve. Is my reasoning flawed here?
Click to expand...

A pump adds kinetic energy to the fluid. Pressure is the result of resistance to the motion of the fluid. Since pressure is proportional to flow and resistance, and easily measured, it is convenient for expressing the performance characteristics of fluid systems.
 
You want to keep restrictions in the suction (pump inlet) line to a minimum. This would be HIGHLY unlikely in WC'd setups, but cavitation can literally destroy a pump very quickly.
Click to expand...

^^^ Regarding elbows in a WC computer pump inlet. It is simply GMP (Good Manufacturing Practice) to keep sharp bends to a minimum in a fluid line, particularly the suction line.

And Bones gives a reasonable explanation on the relationship between flow, pressure, and energy.:cool:
 
Well the people I have read about who seem to have a problem with this pump starvation problem all seem to be using very flexible tubing where the tube collapses. If you use PUR tubing you aren't going to have the tubing collapsing.
 
migueld said:
TN why dont you get a DDC?
Click to expand...

Because it wouldn't make much difference with what I already got. However, seeing as I have a couple of Aqua Computer Poweradjust's on order I intend to get one just to see what it is capable of when properly controlled and interfaced with the Aquasuite software. I don't know how many are spoken for from Sharka's shipment, but I do know that of the 20 ordered I only took 2 of them.




poweradjust_5.jpg
 
Top Nurse said:
Because it wouldn't make much difference with what I already got. However, seeing as I have a couple of Aqua Computer Poweradjust's on order I intend to get one just to see what it is capable of when properly controlled and interfaced with the Aquasuite software. I don't know how many are spoken for from Sharka's shipment, but I do know that of the 20 ordered I only took 2 of them.




poweradjust_5.jpg
Click to expand...

I see. Well the main reason I asked is because your tubing routing seemed to give you trouble with your eheim pump and since the DDC is much smaller it might solve your problem. That and the niceness of a DDC with an alphacool top with the different options for mounting.

Regarding AC's Poweradjust do you mean to control the RPM and use the interface for emergency shut down or to undervolt the DDC or both? I haven't had good results on my bigNG for undervolting, it gets too hot. I wonder how AC's controller measures up.
 
migueld said:
I see. Well the main reason I asked is because your tubing routing seemed to give you trouble with your eheim pump and since the DDC is much smaller it might solve your problem. That and the niceness of a DDC with an alphacool top with the different options for mounting.

Regarding AC's Poweradjust do you mean to control the RPM and use the interface for emergency shut down or to undervolt the DDC or both? I haven't had good results on my bigNG for undervolting, it gets too hot. I wonder how AC's controller measures up.
Click to expand...

It does all of the above and more! Here's the link to a review if your interested.



USB controlled Poweradjust by Aqua Computer is particularly made for the regulation of Laing DDC pumps. The controller uses a programmable DC voltage (no PWM):

  • The Laing pumps are integrated with the following features:
  • Adjustable start burst to eliminate non-starting pump problems
  • Measurement of the maximum speed
  • Adjustable flow program
  • Can program firm output voltage or RPM (default)
  • Characteristic of regulation for number of revolutions default details configurable
  • Passing on of the RPM of the pump at the Motherboard or production of an artificial RPM signal, which can be deactivated in the case of errors
  • Number of revolutions monitoring with renewed Burst with stop
  • Programmable jumper, for which different power outputs can be pre-defined
  • USB communication to the PC
  • Easy to use Aquasuite software to configure Poweradjust USB
  • Aquasuite integrated with the Poweradjust USB allows a connection for a flow sensor
 
TN, thanks for the info.

Regarding elbows, my $0.02 is this: don't just put them in there for the heck of it, but if it will allow a better location for res, rad, etc. then it's okay.

My rig, Headrush, has a couple of 1/2" 90's that allow me to reduce the length of tubing and have a great location for the radiator. And you know what? The temps are 3C cooler than a crappy rad location with no 90's.

Hope that helps.

IMG_7545Medium.jpg


IMG_7548Medium.jpg


IMG_7550Medium.jpg
 
So judging by headala's response and setup with a 90 degree elbow at the outlet of his pump, it shouldn't harm me much to use a 90 degree elbow on the outlet on my Aquaxtreme 50Z pump. I can see where a 90 might be harmful at the inlet of a higher-flow setup due to starvation, although this probably has tons of variables associated with it - pump type, tubing setup, etc.

I might be able to get away without using the 90's, but it requires some fairly tight turns in the tubing. It isn't kinked where it connects to my barbs, but it isn't completely *round* either. I wonder whether or not it would be better to use 90 degree elbows in this situation.
 
Go for it. I doubt you would see any "real" world differances, especially if you have to make a tight tubing bend ILO.

Also, looks does enter into the equation...
 
JC634 said:
Go for it. I doubt you would see any "real" world differances, especially if you have to make a tight tubing bend ILO.

Also, looks does enter into the equation...
Click to expand...

Ranker said:
Here's some of the tests/work done by regular contributors over on XS:

By Marci of Thermochill: http://www.xtremesystems.org/forums/showpost.php?p=1782017&postcount=5

Tests done by hand:
http://www.overclock.net/water-cooling/198795-one-1-2-elbow-4-1-a.html

Theoretical results by Cathar:
http://www.xtremesystems.org/forums/showpost.php?p=2263143&postcount=49

and follow-up post: http://www.xtremesystems.org/forums/showpost.php?p=2265828&postcount=63
Click to expand...

icon_risota.gif


My post in that topic never really got answered. I think Cathar doesn't take kindly to people questioning his hypothesis and most certainly his disciples take an even less kindly look. So anyway lets add up the bent fittings I have in my loop, okay?

90's = 5 (one of these is in an area that is supposedly a big no-no at the pump inlet)

45's = 7

We know from Cathar's hypothesis that 90's equal approximately a .1C temp increase per each fitting, right? So in just 90's alone we would see a .5C increase in temps, right? Now he didn't comment about the 45's, but other people have figured it had a significant effect over a straight fitting. So let's take a conservative figure and say a 45 has a 0.05C temp increase per fitting, okay? So my 7 45's should equal about a .35C increase. Now the total increase would be approximately .85C for all 12 of those fittings, right?

My current loop with an AMD 3500+ OC to about 2.7Ghz at 1.525 volts puts out a bit of heat right? My NB also puts out even more heat and when we add in a VR and GPU cooler we get even more heat and so on as we add more and more coolers, right?

I currently have about a .9C to 1.1C delta between my ambient and out going (from radiator) temps. I think these temps are pretty solid as I have calibrated the sensors in an ice bath plus I measured the same out going temp with a glass thermometer and a medical grade digital thermometer (nice to have access to medical grade technology, heh?)

Since other people say it is a physical impossibility to get less than your ambient temp we are somewhat at an impasse here because with the added temps we are almost at the theoretical limit, correct? Assuming I ditch all the bent fittings and go strictly straight fittings I would be pretty close to the same output water temp as I had going in, right?

Now all the above threads that provided testing were done in relatively high flow state. Now from what my brother says (he's a degree'd fluid dynamics engineer who designs and builds factories where there are zillions of feet of piping going every which way) the higher the flow the more pronounced degree to which you have an effect on flow with 90 degree fittings.

So my take on all this 90 degree business is two fold. First of all I think there is something missing in the hypothesis and the software modeling that Cathar has developed. Second of all I don't think that adding in the necessary bent fittings to your loop is going to make a hill of beans difference in the Real World as it certainly wouldn't make a bit of difference in my loop. :D

Note: Question Authority :D
 
Gee,

Why are you ROFL at me, or was that directed only at Ranker?

1st topic:

Regardless, very true that the higher the flow thru ANY pipe/tube/elbow will increase pressure drop/decrease flow. Laminer flow is good in transport lines, turbulent flow is not. Simple physics.

All I was saying by "real" world is that the flow thru a WC loop is most likely in the laminer range thru the TRANSPORT lines (tubing, etc). There will be small, but measureable differances when introducing elbows, smaller tubing, etc. I really have no desire to do the calcs, since common sense says flows are pretty minimal thru these lines.

Laminer flow is good for transport lines, bad when you get to the blocks. Laminer flow does not allow much heat transfer from one surface to another. Turbulent flow does. Which is why nozzles are often used to create that turbulent effect. In addtion to greater heat transfer area, the "nobs" on blocks also produce a "turbulent" effect, further enhancing heat transfer.


2nd topic:

As far as AC is concerned, everyone knows that you promote AC over anything else, regardless of cost, effectiveness, etc. Interpretation: "fangirl".:p Good for you. I am a Crystalfontz kind of guy, which IMHO, provides a very cost-effective alternative to AC's Aquaero solution. Do you see me dissing the Aquaero over the CF solution at every turn? No. To each their own. I simply value the dollar being in my pocket vs. AC's over-priced (IMO) products. You see the value of AC, I do not. Again, to each their own. I do have to admit that the AC software is quite flashly compared to the CF software. Point in favor of AC. Both, however, are equally effective at providing a fan control and monitoring solution.

In the end, from what I have seen of your machines, you do very nice work. You appear very meticulus (sp?) in your setups and ask some very probing questions, at times. Unfortunately, it seems quite transparent that everything you say/question is done in an effort to "prove" that AC is best, no matter what.

Please make comments to provide answers and ask questions for the sake of knowledge, without adding the AC slant at almost every turn. They would be received in a much more favorable light to those who are not AC disciples.:D

I apologize if I have unfairly offended you. Just remember that for every 10 people in 1 room, you will have at least 20 different adament opinions on a particular subject.;)
 
I, for one, think that TN has a very valid point. According to the data from Cathar, et al., her loop defies physics. I am curious as to what the explanation for that is. I'd like to hear it.

As far as AC goes, I have nothing for it or against it. Sure, TN has a strong tendencey (to put it lightly) towards AC, but she said nothing about it here. So now this thread that was a good discussion has now degraded to the same 'ole stuff.

So, please, stick to the point, not what someone might or might not think or what they have posted in the past. She posted data relevant to the current discussion. She didn't pimp AC (or would that be "madam AC").
 
Self testing means nothing, especially if we lack the proper resources or techniques such as those created/used by Cathar, BillA, etc. Our equipment can't even resolve things down to the .1C level accurately. Self testing usually only produces the results we're looking or hoping to find. I wouldn't even trust my own results as even I have a prejudice of what I'd expect. People can spin information for their own purposes. The temptation is especially there when we conduct tests ourselves. It's pretty obvious that I have a high-flow/all copper bias while TN vehemently advocates purely for the opposite. If I was a third party, I'd take everything TN and I say with a grain of salt if we were to produce "results" or conduct "tests". Cathar has no bias as he's offered his work and has revised it as new technology changes the landscape, sometimes going in the opposite direction of what was previously best. Ex: There was a time when 1/2" tubing resulted in a significant improvement in temps and was considered the standard. However, as shown by his recent tests, recent advances in pumping power (most specifically Laing's DDC pumps) have minimized the impact of tubing size to the point where 3/8" offers a negligible impact on performance.

Laws of physics are laws for a reason. How would Cathar's data indicate that TN's loop defies the laws of physics. The accuracy of the sensors available to the typical watercooling enthusiast can't even resolve at the levels we're looking at. He specifically states that as a ROUGH RULE OF THUMB, so and so is equal to so and so for those of us who lack the time or the patience to work out the equation for our personal build.

I don't really care who it is and even I may have been guilty of it in the past, but to say "well I got so and so results so the laws of physics must be wrong" just means you're an idiot with an agenda. That applies to everyone be it me, TN, Erasmus, rick, arcygenical, BillParrish, or whomever.

So the data and physics are there. Interpret the data however you wish, but for fucks sake, lets all just stop spinning it and just let the user determine how it affects them. Present the info and move on.
 
Top Nurse said:
I currently have about a .9C to 1.1C delta between my ambient and out going (from radiator) temps. I think these temps are pretty solid as I have calibrated the sensors in an ice bath plus I measured the same out going temp with a glass thermometer and a medical grade digital thermometer (nice to have access to medical grade technology, heh?)
Click to expand...

Perhaps I am missing something here, but what does the temperature delta of a radiator's inlet vs. outlet have to do with CPU temperature due to waterblock performance? Higher flow is better overall for heat flux, but that pertains more to a waterblock rather than the radiator or other heat dissipation device.

Besides, you are giving a delta without an absolute temperature to give a reference. That delta might be fairly good, but if the radiator liquid is ranging between 35 and 36C in a 20C environment (I highly doubt it, but you see my point), then the small delta is moot for what we are trying to achieve.

TN, what pump did you use for this testing? The Eheim or the DDC?
 
JC634 said:
Gee,

Why are you ROFL at me, or was that directed only at Ranker?
Click to expand...

Not you. ;)

1st topic:

Regardless, very true that the higher the flow thru ANY pipe/tube/elbow will increase pressure drop/decrease flow. Laminer flow is good in transport lines, turbulent flow is not. Simple physics.

All I was saying by "real" world is that the flow thru a WC loop is most likely in the laminer range thru the TRANSPORT lines (tubing, etc). There will be small, but measureable differances when introducing elbows, smaller tubing, etc. I really have no desire to do the calcs, since common sense says flows are pretty minimal thru these lines.

Laminer flow is good for transport lines, bad when you get to the blocks. Laminer flow does not allow much heat transfer from one surface to another. Turbulent flow does. Which is why nozzles are often used to create that turbulent effect. In addtion to greater heat transfer area, the "nobs" on blocks also produce a "turbulent" effect, further enhancing heat transfer.
Click to expand...

Sounds reasonable, but as you put it all this is minimal. So if you need the fittings for your layout use them. If you don't need them, don't use them. :)


2nd topic:

As far as AC is concerned, everyone knows that you promote AC over anything else, regardless of cost, effectiveness, etc. Interpretation: "fangirl".:p Good for you. I am a Crystalfontz kind of guy, which IMHO, provides a very cost-effective alternative to AC's Aquaero solution. Do you see me dissing the Aquaero over the CF solution at every turn? No. To each their own. I simply value the dollar being in my pocket vs. AC's over-priced (IMO) products. You see the value of AC, I do not. Again, to each their own. I do have to admit that the AC software is quite flashly compared to the CF software. Point in favor of AC. Both, however, are equally effective at providing a fan control and monitoring solution.

In the end, from what I have seen of your machines, you do very nice work. You appear very meticulous (sp?) in your setups and ask some very probing questions, at times. Unfortunately, it seems quite transparent that everything you say/question is done in an effort to "prove" that AC is best, no matter what.

Please make comments to provide answers and ask questions for the sake of knowledge, without adding the AC slant at almost every turn. They would be received in a much more favorable light to those who are not AC disciples.:D

I apologize if I have unfairly offended you. Just remember that for every 10 people in 1 room, you will have at least 20 different adamant opinions on a particular subject.;)
Click to expand...

So what does this have to do with the topic at hand? Other than a few side issues that have come up I don't think AC has anything to do with this topic.

It's hard not to mention Aqua Computer as they make more water cooling products than just about any water cooling company out there. AC just isn't a water block company, they make their own pump controllers for Eheim pumps, they make Laing DDC pump controllers, they make water cooling control circuitry along with a wide variety of accessories that no one else makes and tie it all together with their own proprietary software, they make high end stand alone cooling solutions, oh and they also make their own water blocks. Quit trying to say that the white rabbit is a black rabbit. ;)
 
Ranker said:
Self testing means nothing, especially if we lack the proper resources or techniques such as those created/used by Cathar, BillA, etc. Our equipment can't even resolve things down to the .1C level accurately. Self testing usually only produces the results we're looking or hoping to find. I wouldn't even trust my own results as even I have a prejudice of what I'd expect. People can spin information for their own purposes. The temptation is especially there when we conduct tests ourselves. It's pretty obvious that I have a high-flow/all copper bias while TN vehemently advocates purely for the opposite. If I was a third party, I'd take everything TN and I say with a grain of salt if we were to produce "results" or conduct "tests". Cathar has no bias as he's offered his work and has revised it as new technology changes the landscape, sometimes going in the opposite direction of what was previously best. Ex: There was a time when 1/2" tubing resulted in a significant improvement in temps and was considered the standard. However, as shown by his recent tests, recent advances in pumping power (most specifically Laing's DDC pumps) have minimized the impact of tubing size to the point where 3/8" offers a negligible impact on performance.

Laws of physics are laws for a reason. How would Cathar's data indicate that TN's loop defies the laws of physics. The accuracy of the sensors available to the typical watercooling enthusiast can't even resolve at the levels we're looking at. He specifically states that as a ROUGH RULE OF THUMB, so and so is equal to so and so for those of us who lack the time or the patience to work out the equation for our personal build.

I don't really care who it is and even I may have been guilty of it in the past, but to say "well I got so and so results so the laws of physics must be wrong" just means you're an idiot with an agenda. That applies to everyone be it me, TN, Erasmus, rick, arcygenical, BillParrish, or whomever.

So the data and physics are there. Interpret the data however you wish, but for fucks sake, lets all just stop spinning it and just let the user determine how it affects them. Present the info and move on.
Click to expand...

Zzzzzzzzzzzzzzzzzzz

1. Another Ranker diatribe.
2. When you don't like the results, attack the messenger
3. When you have something valid to discuss here please feel free to jump on in.
 
grinchy said:
TN what are you using to get your temps?:)

and which 3500 are you running?
Click to expand...

I am using an Aqua Computer Aquaero (similar to mCubed design). The sensors (ambient, water into rad, water out of rad) were calibrated using an ice bath. I also checked my results with a glass immersion thermometer and a digital thermometer from the hospital where I work for the water out of the radiator.

I am running a socket 939 AMD Athlon 3500+ OEM Winchester core.
 
DarkenReaper57 said:
Perhaps I am missing something here, but what does the temperature delta of a radiator's inlet vs. outlet have to do with CPU temperature due to waterblock performance? Higher flow is better overall for heat flux, but that pertains more to a waterblock rather than the radiator or other heat dissipation device.

Besides, you are giving a delta without an absolute temperature to give a reference. That delta might be fairly good, but if the radiator liquid is ranging between 35 and 36C in a 20C environment (I highly doubt it, but you see my point), then the small delta is moot for what we are trying to achieve.

TN, what pump did you use for this testing? The Eheim or the DDC?
Click to expand...

This has nothing to do with the waterblock, but is simply a discussion of water out vs ambient temperature and the temps that would be created from eliminating 90 and 45 degree fittings. The sensors were calibrated in an ice bath (and verified with medical grade thermometers in the case of the output temp) so we are discussing the absolute temps as well are we not?

I'm using an Aqua Computer OEM pump called an Aquastream made by Eheim and controlled by Aqua Computer control electronics.
 
When I said absolute temperature, I was referring to a *normal* temperature with respect to the freezing point of water (i.e. 23C). Technically, absolute temperature is the temperature measured with respect to absolute zero (commonly referred to as the Kelvin scale). So, I was asking for the temperature range that you actually measured, not the delta. Sorry for the confusion on my part.

I'm curious to see the temperature difference from inlet to outlet of higher-flow systems using the Thermochill 120.3 and a powerful pump, such as the DDC. The same goes for heater cores paird with a DDC. If I had more accurate equipment at home, I would do it on my setup, but unfortunately I am limited to a single compunurse display that has a +/- 1.5C accuracy (as said on box). Who knows what temperature it is calibrated to.

Something people need to realize is calibration doesn't always achieve accuracy. If the electronics support it, it would be better to calibrate to room temperature, or 25C, rather than 0C. This is because error begins to increase from the calibration point, usually linearly.

I've done a lot of work with thermocouple electronic measurement equipment design at a lab and basically you have an error window that grows as you deviate from the calibration point. So, if something lists a +/- 1.5C accuracy, this could be the average over a given range (i.e. 0-100C), or perhaps at room temperature. If you don't believe me, take a look at the bottom of page 6 the AD595 datasheet, a common solid-state cold junction compensator http://www.analog.com/UploadedFiles/Data_Sheets/AD594_595.pdf. This circuitry creates an "electronic ice-bath" and does so more accurately. I'm fairly certain that thermistors and RTD's suffer from this increasing error window as well due to the limitations of supporting circuitry (op-amps, etc).

Take a look at that calibration error curve on the datasheet. If the IC were designed to be calibrated at 0C and the error lines had the same slope, there would be a +/-1.2C error at 25C, but nearly zero error at 0C. See what I am getting at here? TN, I'd take a careful look at the datasheet of that medical grade thermometer and see how accurate it is as the temperatures we are measuring. Perhaps it really does have +/- 0.1C (or whatever it is) accuracy overall. If so, then that is great. Then again, products lie about their specs, as do datasheets from time to time. It can really leave us up in the air.

Fortunately we are measuring a delta from inlet to outlet. If the same measuring device is used, the error tends to be very low.
 
Top Nurse said:
I am using an Aqua Computer Aquaero (similar to mCubed design). The sensors (ambient, water into rad, water out of rad) were calibrated using an ice bath. I also checked my results with a glass immersion thermometer and a digital thermometer from the hospital where I work for the water out of the radiator.

I am running a socket 939 AMD Athlon 3500+ OEM Winchester core.
Click to expand...

I remember my ole 3500 winchester....;)

I did not watercool my winchester I only used the XP-90 HSF on it...and I remember going by the bios reading I could Idle 2C over ambient....

but you are using something different to measure temps....

have you thought about upgrading to an intel setup?

I have noticed that making the move to intel my temps are ALOT higher than they were using AMD...:confused:
 
grinchy said:
I remember my ole 3500 winchester....;)

I did not watercool my winchester I only used the XP-90 HSF on it...and I remember going by the bios reading I could Idle 2C over ambient....

but you are using something different to measure temps....

have you thought about upgrading to an intel setup?

I have noticed that making the move to intel my temps are ALOT higher than they were using AMD...:confused:
Click to expand...

I'm glad you can remember it as if you couldn't we might suspect Alzheimer's being it was only 2 years ago they were selling them. ;)

Actually at idle my delta's are far greater because I have the fans barely turning over or off. However, when I checked all this I manually set the fans at 100% input power.

No C2D's for me! My office computer does everything I want it to do so why would I want to put in a POS chip that's has an IHS that is bent and creates a shitload more of heat? If anything this might get an Opteron 185 next month.
 
DarkenReaper57 said:
When I said absolute temperature, I was referring to a *normal* temperature with respect to the freezing point of water (i.e. 23C). Technically, absolute temperature is the temperature measured with respect to absolute zero (commonly referred to as the Kelvin scale). So, I was asking for the temperature range that you actually measured, not the delta. Sorry for the confusion on my part.

I'm curious to see the temperature difference from inlet to outlet of higher-flow systems using the Thermochill 120.3 and a powerful pump, such as the DDC. The same goes for heater cores paird with a DDC. If I had more accurate equipment at home, I would do it on my setup, but unfortunately I am limited to a single compunurse display that has a +/- 1.5C accuracy (as said on box). Who knows what temperature it is calibrated to.

Something people need to realize is calibration doesn't always achieve accuracy. If the electronics support it, it would be better to calibrate to room temperature, or 25C, rather than 0C. This is because error begins to increase from the calibration point, usually linearly.

I've done a lot of work with thermocouple electronic measurement equipment design at a lab and basically you have an error window that grows as you deviate from the calibration point. So, if something lists a +/- 1.5C accuracy, this could be the average over a given range (i.e. 0-100C), or perhaps at room temperature. If you don't believe me, take a look at the bottom of page 6 the AD595 datasheet, a common solid-state cold junction compensator http://www.analog.com/UploadedFiles/...7AD594_5_c.pdf. This circuitry creates an "electronic ice-bath" and does so more accurately. I'm fairly certain that thermistors and RTD's suffer from this increasing error window as well due to the limitations of supporting circuitry (op-amps, etc).

Take a look at that calibration error curve on the datasheet. If the IC were designed to be calibrated at 0C and the error lines had the same slope, there would be a +/-1.2C error at 25C, but nearly zero error at 0C. See what I am getting at here? TN, I'd take a careful look at the datasheet of that medical grade thermometer and see how accurate it is as the temperatures we are measuring. Perhaps it really does have +/- 0.1C (or whatever it is) accuracy overall. If so, then that is great. Then again, products lie about their specs, as do datasheets from time to time. It can really leave us up in the air.

Fortunately we are measuring a delta from inlet to outlet. If the same measuring device is used, the error tends to be very low.
Click to expand...

Actually the Aquaero measures in Kelvin, but outputs in Centigrade. The Aquaero was providing all the temps. I just decided to spot check the results on the output of the radiator because it was easy to do.

It's my understanding that the NTC thermistors have a linear output so that error on one end is linear to the other end of the temperature scale?

The link provided was broken.
 
Top Nurse said:
Actually the Aquaero measures in Kelvin, but outputs in Centigrade. The Aquaero was providing all the temps. I just decided to spot check the results on the output of the radiator because it was easy to do.

It's my understanding that the NTC thermistors have a linear output so that error on one end is linear to the other end of the temperature scale?

The link provided was broken.
Click to expand...

Wouldn't matter, as the error would produce an area of uncertainty on the graph. If it's +/-.5'C at 0'C and at 25'C, it could read 24.5'C for 25, and 0.5'C for 0. The same uncertainty, different adjusted values.
[Edit; I'll add, that if you're implying the thermistor would read +x'C at every temperature, you're also implying that the thermistor actually has no uncertainty, which you won't get anyone to swallow :D]
You should use the more precise thermistor to measure the temperatures you're interested in, not reference temperatures.
 
NTC thermistors generally aren't the best devices to do accurate scientific readings with either. Earlier, I talked about the linear error window of a typical IC that calibrates a thermocouple. Well, there is more to it than that, as the thermocouple itself has a completely non-linear (probably around 10th or 20th order) polynomial. This polynomial is approximated as it is, and then linearized. Basically, the end result is you have a window of inaccuracies not only due to the measuring device, but the instrumentation that calculates the measurement.

Now, enough talk about thermocouples and back to NTC thermistors. These devices are nonlinear as well due their nature of operation. You can linearize the thermistor and use a first-order equation to approximate it, but this is inaccurate. Usually for more important work, a 3rd or higher order equation is used. Nonetheless, there is still an error window since you don't know how the electronics are approximating the thermistor. Even if a 3rd order approximation is used, there are inherent inaccuracies due to op-amp technology and such (remember, semiconductors are non-linear devices as well, even if they are treated as such to simplify problems!).

TN, as you can see, thermistors are nonlinear, even if they are often treated as linear. In addition, I think you are thinking of a constant error. If there is a linear error window that increases as you move to either side of a calibration point (see p. 6 of that datasheet for an example of what I am talking about) then your error is not always the same on both sides of the spectrum as you suggest.

RTD's would probably be best suited for the accuracy of work we are doing here, especially when comparing the temperatures of inlet vs. outlet. Then again, if the specs of that medical grade thermometer are, say +/- 0.1C or less at 25C, then it is more than sufficient.
 
You must log in or register to reply here.
Back
Top