Sign in with
Sign up | Sign in
Your question

Help me understand water delta

Last response: in Overclocking
Share
July 7, 2012 10:50:55 AM

Hi

I'm new to watercooling and am trying to get as good an understanding as I can. I've read a post connected to the Beginner's stickyabout water delta. The author Conundrum states that low water delta is good. This makes sense to me as it indicates the loop is getting rid of the heat effectively so the CPU, GPU's etc will be well cooled.

Then he says that a delta of 10-15 degrees (or maybe even higher) is fine for GPU loops but a lower figure of 5-10 degrees is better for CPU loops. There is no explanation as to why this difference is desirable. I would have thought the lower the delta the better for both classes of components. I'm planning a single loop so the water delta will be common to both. As a gamer with an sli system my main reason for wanting to switch to water is to calm down my noisy, hot GPU's. The Sandy Bridge CPU is less of a problem.

Can anyone explain the technical reason why water delta should favour CPU loops over GPU'd?

Thanks in advance

Mag

More about : understand water delta

a b à CPUs
a b K Overclocking
July 7, 2012 12:22:47 PM

Delta is simply the difference in the ambient (room) and water temperatures. So if the room where you PC is 25C, and the water in your loop is 30C, then you have a loop with a 5C delta.

A lower delta effectively means that your loop is cooling better. Due to thermodynamics, the lowest the water temperature can naturally be is your room temperature (all things in the room will be at the same temperature). So the closer your coolant temperature is to the room temperature, the more heat your loop is getting rid of.

Where the delta is important is in determining what fans/radiators to use. Most of the charts have heat removal numbers for a certain delta. You read the charts in the following way: in order to achieve a certain delta (lets call it D), you need model X radiator with fans at Y RPM in order to remove Z watts of heat. As long as that number Z meets or exceeds the total amount of heat generated by your components (conservatively estimated by the sum of the TDPs of the components), the loop will achieve the delta D.
a b à CPUs
a b K Overclocking
July 7, 2012 12:39:28 PM

Practical Example: Let's say we want to cool a Sandy Bridge CPU (stock) and a GTX 670, and we're thinking about using an RX360 to keep the noise down.

1. First we calculate the TDP for the components:
Sandy Bridge CPU - 125W (off the top of my head - could be wrong)
GTX 670 - 170W (know this one ;) )

Total TDP = 295W (we'll say 300W for simplicity)

2. Next let's look at the rad chart for the RX360. The axes are Delta vs. Heat Load (removed in this case), so for some heat load removed on the X axis, there's a corresponding loop delta that is achieved on the Y axis.



3. Note that as the slope of the lines increases, the lower the RPM and cooling ability. So at 300W, the fastest fans (black line) manage a 3C delta (which is fantastic), while the slowest fans (purple line) only manage a 12C delta (not good). However, the black line represents high RPM (and thus high noise) fans, and the goal was to have a quieter loop. How else could we achieve a lower delta and meet the cooling and noise requirements?

4. Using more than one radiator
In this case, you would halve the heat load that you use on this chart - the same amount of heat (300W) is being dissipated by more than one radiator, so each radiator must remove a fraction of the heat in order to meet the 300W requirement. If you used two of these rads, they would need to remove 1/2 the heat, or 150W; 3 rads, 1/3 (100W); etc.

If two rads are used (likely scenario), then the delta range would be read at 150W, which would be 2-6C. Therefore we could use two RX360s with the slowest fans (purple) and expect to achieve around a 6C delta and extremely low noise (600RPM fans). This is a great candidate for a near silent loop.


In the end, the delta isn't overly important, since it's more of a result that it is a cause of something. If you plan your loop properly, it generally won't cost you any more money to have your loop achieve a 5C delta than a 10C delta - it just comes down to proper research ahead of time.
Related resources
July 7, 2012 3:32:56 PM

The reason for the difference in the recommended delta between GPU loops and CPU loops is simple: a single CPU usually produces less heat than a GPU, because with a GPU you are typically talking about full cover coolers which are not only cooling the GPU but also VRMs and memory. Plus CPUs are usually more sensitive to high temperatures, with a TJ Max between 95 and 110°C, while GPUs can go as high as 125°C before taking damage.
The other difference is that when a GPU is overheating you often see warning signs such as gliches, errors, driver recovery errors... When the CPU overheats the PC just shuts down, or you get a bluescreen.
July 8, 2012 8:56:43 AM

boiler1990 said:
Practical Example: Let's say we want to cool a Sandy Bridge CPU (stock) and a GTX 670, and we're thinking about using an RX360 to keep the noise down.

1. First we calculate the TDP for the components:
Sandy Bridge CPU - 125W (off the top of my head - could be wrong)
GTX 670 - 170W (know this one ;) )

Total TDP = 295W (we'll say 300W for simplicity)

2. Next let's look at the rad chart for the RX360. The axes are Delta vs. Heat Load (removed in this case), so for some heat load removed on the X axis, there's a corresponding loop delta that is achieved on the Y axis.

http://i1123.photobucket.com/albums/l554/bravokiloromeo/234b2bc7.jpg

3. Note that as the slope of the lines increases, the lower the RPM and cooling ability. So at 300W, the fastest fans (black line) manage a 3C delta (which is fantastic), while the slowest fans (purple line) only manage a 12C delta (not good). However, the black line represents high RPM (and thus high noise) fans, and the goal was to have a quieter loop. How else could we achieve a lower delta and meet the cooling and noise requirements?

4. Using more than one radiator
In this case, you would halve the heat load that you use on this chart - the same amount of heat (300W) is being dissipated by more than one radiator, so each radiator must remove a fraction of the heat in order to meet the 300W requirement. If you used two of these rads, they would need to remove 1/2 the heat, or 150W; 3 rads, 1/3 (100W); etc.

If two rads are used (likely scenario), then the delta range would be read at 150W, which would be 2-6C. Therefore we could use two RX360s with the slowest fans (purple) and expect to achieve around a 6C delta and extremely low noise (600RPM fans). This is a great candidate for a near silent loop.


In the end, the delta isn't overly important, since it's more of a result that it is a cause of something. If you plan your loop properly, it generally won't cost you any more money to have your loop achieve a 5C delta than a 10C delta - it just comes down to proper research ahead of time.



Thanks for your reply. This helps me greatly in planning my loop. However I’m still bemused regarding the targets. The generally held belief seems to be a water delta of 10 degrees is good. Is there any rationale behind this? I assume it is a figure where the cooling of your components will be good without having to resort to very large radiators and/or very fast and noisy fans. Has anyone produced any figures to back this up?

Another generally held belief I have come across is that a loop flow rate around 1 US gallon per minute is fine. After looking into this I found an explanation by a chap called Charliehorse (I think) that explained that modern blocks are normally happy with this as it is sufficient to turn the flow from laminar (smooth) to turbulent which allows the heat to be transferred to the water much better. Test results from people like Skinnee show that flow above 1 gpm does not improve cooling much. Is there any comparable logic for the 10 degree delta figure?

Regarding my original question why the CPU delta should be lower than the GPU delta the next reply I received seems to answer this. I think the original writer didn’t explain it very well. All he seems to be saying is that a CPU loop will have a lower GPU than a comparable sli GPU loop using the same cooling components as it generates less heat. I would have thought this was obvious myself. However if planning two separate loops I would have thought you aimed for 10 degrees for both (if this magic 10 degree delta is indeed correct) and simply used smaller radiators in the CPU loop and larger ones in the GPU loop.

Thanks again for your help. By asking what probably seems to you to be daft questions I’m learning a lot

July 8, 2012 8:58:48 AM

JustAnotherNoob said:
The reason for the difference in the recommended delta between GPU loops and CPU loops is simple: a single CPU usually produces less heat than a GPU, because with a GPU you are typically talking about full cover coolers which are not only cooling the GPU but also VRMs and memory. Plus CPUs are usually more sensitive to high temperatures, with a TJ Max between 95 and 110°C, while GPUs can go as high as 125°C before taking damage.
The other difference is that when a GPU is overheating you often see warning signs such as gliches, errors, driver recovery errors... When the CPU overheats the PC just shuts down, or you get a bluescreen.



Thanks for your reply. I think this clears it up. What he is saying is that with the same cooling hardware the CPU loop should have a lower delta than the GPU loop as it does not have to dissipate as much heat. However I would have thought that if a delta of 10 degrees is the normal target why do you not use this for both loops and use different size radiators accordingly if running a twin loop system?

I’m planning on putting both CPU and GPU’s on a single loop so there can only be one delta. I’m planning my components using a delta of 10 degrees (using the help from the poster above) as this seems to be the accepted target. I refer to me answer to the above reply – why is 10 degrees accepted as the best target delta? Is there any underlying physics explaining this value?

Again you have been a great help and I feel my understanding of watercooling principles is nearly there and I’m going to avoid many mistakes.

a c 103 à CPUs
a c 190 K Overclocking
July 8, 2012 9:03:47 AM

If you don't ask, you will never know :) 
a 10'c delta is regarded as a 'usual' target to aim for because if you aren't getting that then the loop isn't really performing any better than air cooling could, you want to see some results for all that time/money spent right?
I wouldn't worry about it overly right now, but if you grab an inline temp sensor you can measure the temps of the water, and an external thermometer or sensor will measure ambient (room) temp, making it easy to see what you have achieved, and show the benefits/losses of any alterations you make, like altering fanspeeds
Moto
a b à CPUs
a b K Overclocking
July 8, 2012 1:26:26 PM

Moto's reason is spot on.

Keep in mind that delta is not a cause in a loop - it is an effect. Like I said, delta is an indicator of how well the loop is cooling.
a b à CPUs
a c 331 K Overclocking
July 8, 2012 5:37:10 PM

^ Well said.

Flows of 0.75 GPM + are generally considered good for watercooling and above 1.0 GPM is ideal, if possible. However, depending on your pump and loop in question, these flows might also be restrictive and head pressure of a pump can improve flow in a loop where a pump with actual higher flow rate might not perform as well.
a c 103 à CPUs
a c 190 K Overclocking
July 8, 2012 6:53:43 PM

Must buy a flow meter sometime hehe I'd be interested to see what I actually have with all my crap in the loop lol, 750lph reduced to??
:) 
Moto
a b à CPUs
a c 331 K Overclocking
July 8, 2012 7:01:12 PM

750 lph = 12.5 lpm
a c 103 à CPUs
a c 190 K Overclocking
July 8, 2012 7:11:34 PM

Meh, typo gave me Lpm instead of Lph, edited :) 
Moto
a b à CPUs
a c 331 K Overclocking
July 8, 2012 7:12:53 PM

:) 
July 9, 2012 10:24:52 AM

Motopsychojdn said:
If you don't ask, you will never know :) 
a 10'c delta is regarded as a 'usual' target to aim for because if you aren't getting that then the loop isn't really performing any better than air cooling could, you want to see some results for all that time/money spent right?
I wouldn't worry about it overly right now, but if you grab an inline temp sensor you can measure the temps of the water, and an external thermometer or sensor will measure ambient (room) temp, making it easy to see what you have achieved, and show the benefits/losses of any alterations you make, like altering fanspeeds
Moto

Thanks for your reply. That makes sense. So 10 degrees is the benchmark that everything is cooling well. I presume 9/8/7 etc degrees are better but would be overkill.

Regards

Mag
July 9, 2012 10:27:26 AM

boiler1990 said:
Moto's reason is spot on.

Keep in mind that delta is not a cause in a loop - it is an effect. Like I said, delta is an indicator of how well the loop is cooling.

Thanks again. If I get down to 10 degrees everything is fine then. No need to go lower as its excessive.

Regards

Mag
July 9, 2012 10:33:28 AM

rubix_1011 said:
^ Well said.

Flows of 0.75 GPM + are generally considered good for watercooling and above 1.0 GPM is ideal, if possible. However, depending on your pump and loop in question, these flows might also be restrictive and head pressure of a pump can improve flow in a loop where a pump with actual higher flow rate might not perform as well.

Thanks for your reply. That was the first misconception I had. I thought Laing D5 was the best simply because it had official flow of 25 litres per minute. I now understand we are not trying to fill a swimming pool so this is largely irrelevant and the head pressure is far more important to overcome restriction. I've discovered from research that at 1 gallon per minute flow the stock Laing on highest setting has about the same head pressure as the DDC 3.1 (Swiftech 650) 9w version with a good aftermarket top which quite surprised me.

Thanks again

Mag
a b à CPUs
a b K Overclocking
July 9, 2012 4:43:20 PM

If your loops isnt very restrictive (maybe one or two rads and 2 unrestrictive blocks) you'll *technically* get better cooling from the higher flow rates. Will you notice it? Unlikely.

Head pressure is really important once your loop gets longer and you add more restrictive components (certain blocks, especially RAM blocks).
a c 103 à CPUs
a c 190 K Overclocking
July 9, 2012 5:51:14 PM

**So 10 degrees is the benchmark that everything is cooling well. I presume 9/8/7 etc degrees are better but would be overkill.
**
Theres a saying here, there is is no overkill in W/c, just stuff I don't have in my loop yet :) 
10'c Delta is what to roughly aim for yes, but don't confuse that with 10'c Cpu temps or 10'c room temp,
for example I just turned my Pc on and my room temp was 23.63,
my water was showing as 22.85, five minutes later after throwing some heat in from the Cpu at 32'c idle (3.6GHz) the water temp is up to 23.69'c,
this is an example of thermal eqilibrium, after an hour of surfing etc, my water temp will be roughly 2-3'c above the ambient room temp, after a hard gaming session it may get to 8'c above ambient,less if I turn my fans on/up
now I have what could be considered an overkill loop to attain that result,
a 'normal' loop will take much less time to equalise than a larger over-radded one but at the end of the day, however long it takes, your loop will find the point at which its absorbing as much heat as it can, and dissipating as much as it can and it will settle, that right there will give you the truest measurement of your Delta, short term numbers may look good, 'Hey everyone, my water is below ambient temps yay me!' but then once the heat starts kicking in, that same guy will find his water temps creeping up until T.E. occurs
As I said earlier its not something to fret over too much right now, later on when you are up and running you may want to assess performance and see how to improve it maybe, and that is when Delta-T really becomes a number you want to know
Moto
July 12, 2012 11:47:43 AM

boiler1990 said:
If your loops isnt very restrictive (maybe one or two rads and 2 unrestrictive blocks) you'll *technically* get better cooling from the higher flow rates. Will you notice it? Unlikely.

Head pressure is really important once your loop gets longer and you add more restrictive components (certain blocks, especially RAM blocks).

Thanks for that. The spreadsheet I've used at Martin's Liquid Lab predicts I'll get well over 1 gpm with the parts I've ordered which gives me a bit of leeway for error.

Regards

Mag
July 12, 2012 11:49:44 AM

Motopsychojdn said:
**So 10 degrees is the benchmark that everything is cooling well. I presume 9/8/7 etc degrees are better but would be overkill.
**
Theres a saying here, there is is no overkill in W/c, just stuff I don't have in my loop yet :) 
10'c Delta is what to roughly aim for yes, but don't confuse that with 10'c Cpu temps or 10'c room temp,
for example I just turned my Pc on and my room temp was 23.63,
my water was showing as 22.85, five minutes later after throwing some heat in from the Cpu at 32'c idle (3.6GHz) the water temp is up to 23.69'c,
this is an example of thermal eqilibrium, after an hour of surfing etc, my water temp will be roughly 2-3'c above the ambient room temp, after a hard gaming session it may get to 8'c above ambient,less if I turn my fans on/up
now I have what could be considered an overkill loop to attain that result,
a 'normal' loop will take much less time to equalise than a larger over-radded one but at the end of the day, however long it takes, your loop will find the point at which its absorbing as much heat as it can, and dissipating as much as it can and it will settle, that right there will give you the truest measurement of your Delta, short term numbers may look good, 'Hey everyone, my water is below ambient temps yay me!' but then once the heat starts kicking in, that same guy will find his water temps creeping up until T.E. occurs
As I said earlier its not something to fret over too much right now, later on when you are up and running you may want to assess performance and see how to improve it maybe, and that is when Delta-T really becomes a number you want to know
Moto

Thanks for your help - this is all starting to make sense now.

Regards

Mag
July 12, 2012 4:24:36 PM

There are multiple reasons why a 10° delta is the usual target for a watercooling setup.
If your delta was significantly higher your components would run hotter, possibly reducing your overclock. If you wanted a significantly lower delta the investment becomes disproportionately larger (same as with graphics cards, you get 10% more performance for twice the price).
Finally most of us use our systems at room temperature, which will usually be between 20-35°C. That puts the maximum water temperature during summer somewhere around 45-50°C. That is still very comfortably within specs for the pumps and (more importantly) the tubing. If the tubing gets too hot, it can soften and slip off the barbs.
And finally, every kind of plastic (again, including pumps and especially tubes) contains softeners which are needed to retain some flexibility. Otherwise most plastics would be to brittle and break very easily. These softeners slowly dissipate out of the plastics (i.e old rubberbands break and crumble), and this is accelerated by heat.
July 13, 2012 7:38:38 AM

JustAnotherNoob said:
There are multiple reasons why a 10° delta is the usual target for a watercooling setup.
If your delta was significantly higher your components would run hotter, possibly reducing your overclock. If you wanted a significantly lower delta the investment becomes disproportionately larger (same as with graphics cards, you get 10% more performance for twice the price).
Finally most of us use our systems at room temperature, which will usually be between 20-35°C. That puts the maximum water temperature during summer somewhere around 45-50°C. That is still very comfortably within specs for the pumps and (more importantly) the tubing. If the tubing gets too hot, it can soften and slip off the barbs.
And finally, every kind of plastic (again, including pumps and especially tubes) contains softeners which are needed to retain some flexibility. Otherwise most plastics would be to brittle and break very easily. These softeners slowly dissipate out of the plastics (i.e old rubberbands break and crumble), and this is accelerated by heat.

Thanks for your reply. This is quite interesting - I never considered how warm water could affect things like the grip of tubing. As this is my first build I've been experimenting putting tubing on barbs. Using XSPC 1/2 inch ID tubing and Bitspower fatboy barbs I cannot get the tubing off the barbs again without cutting. However I can see this would be different if it is softened. I will be putting hose clamps on my barbs anyway so I hope I don't suffer from any slippage.

Regards

Mag
!