does water cooling of most parts in a pc have an effect on overall power usage?

[Emperor]

alot of the hardware in the pc generates quite a large amount of heat and some resorted to water-cooling to deal with the heat that would have been dissipated into the air. the room is then cooled by air conditioning.

my question is, does the use of water-cooling in a way lower the overall power consumption in the house?

[zanzabar]

you will add in about 18-25W for the pump, and about 1-6W per fan. you will use less power on with cool chips, but unless there is something like the 480 again i doubt you will see lower power consumption with liquid cooling.

[Proximon]

If I understand your question correctly, no.

No matter how you cool your PC, it is still generating the same amount of heat. Because a larger water system absorbs some of the heat at first, there might be a delay before that heat hits the air, but it will eventually arrive and your AC will have to work the same amount.

[Manicdan]

my pc uses less than 70w in idle. thats with a pump, 4 fans, 2500k and 680

a pump only adds about 5w when its running at a low speed, and fans might use well under 1w each when lowered as well.

but parts do consume less power if the temp is lower. for the 480 in example every 1.5C lower was 1w less power. swapping to a water loop shaved off like 40w in load. so an effecient water loop can probably be a wash in idle, but save a few watts under load.

[Levish]

If I understand your question correctly, no.

No matter how you cool your PC, it is still generating the same amount of heat. Because a larger water system absorbs some of the heat at first, there might be a delay before that heat hits the air, but it will eventually arrive and your AC will have to work the same amount.

could have sworn I saw some benchmarks that demonstrated the opposite, something to the tune of 20% power savings for a hot gpu.

[Manicdan]

could have sworn I saw some benchmarks that demonstrated the opposite, something to the tune of 20% power savings for a hot gpu.

that should answer the question about power draw and temps (for the most extreme case weve seen recently)

[Proximon]

The question has to do with the amount of heat being generated, not the relative GPU temps. Without some sort of link and hopefully some context that chart doesn't tell me much. Is the temp going up due to the ambient room temp rising? Is it due to more power being supplied to the GPU?

Ah well here is the link
http://www.techpowerup.com/reviews/Z...dition/27.html

However, does this hold true for a water cooled card? The statement there is that the power savings are linear.... are they still linear at 50C?

If the electronics consume less power, does that then mean that the total heat production is also lower? After all, less resistance from lower temps would then result in lower temps...

Still, I think the question had more to do with keeping the room cool and the work the AC had to do.

[Manicdan]

The question has to do with the amount of heat being generated, not the relative GPU temps.

they are directly affected. if temps went down then i either increased the cooling capacity or lowered the heat output.
in that chart i showed you they increased fan speeds to simulate what better cooling can do for temps, and as a result power consumption came down greatly. im sure that someone around here can give a serious lesson on how this all works. i really dont know why, i just know that it happens.

[earthwormjim]

The question has to do with the amount of heat being generated, not the relative GPU temps. Without some sort of link and hopefully some context that chart doesn't tell me much. Is the temp going up due to the ambient room temp rising? Is it due to more power being supplied to the GPU?

Ah well here is the link
http://www.techpowerup.com/reviews/Z...dition/27.html

However, does this hold true for a water cooled card? The statement there is that the power savings are linear.... are they still linear at 50C?

If the electronics consume less power, does that then mean that the total heat production is also lower? After all, less resistance from lower temps would then result in lower temps...

Still, I think the question had more to do with keeping the room cool and the work the AC had to do.

It's a transistor, its power consumptions is also its heat generation.

It is a fact that transistors WILL use more power switching at a constant frequency when they are operated at a higher temperature. Leakage current is directly dependent on temperature, it's actually an exponential relationship.

As far as linearity, who knows. I really do wish they had tested the card to lower temperatures or took the time to water cool it. That card in particular had a ton of leakage current, so it probably would continue to see gains all the way up to water cooling temperatures.

[babalouj]

It's a transistor, its power consumptions is also its heat generation.

It is a fact that transistors WILL use more power switching at a constant frequency when they are operated at a higher temperature. Leakage current is directly dependent on temperature, it's actually an exponential relationship.

Exactly this! I came here to say this also. Its a well known fact in the semiconductor industry.

I believe that if you are modest with your watercooling setup and optimize fan quantity and pumping power that you will be able to reduce the overall system power consumption from the wall on an overclocked pc. This is just an educated guess though and I have no proof to back it up. For a completely stock system, I am guessing nothing will beat undervolting the cpu and gpu to achieve lower consumption.

[Proximon]

So what this is all leading to is that the PC will actually produce less total heat if it is kept cooler? So the AC will work less?

If this is true there would have to be some datacenter charts somewhere....

Here is one study
http://www.ecoinfo.cnrs.fr/IMG/pdf/M...Efficiency.pdf
Some of it is just a bit too technical for me but the conclusion seems to be that it's a wash. But then it seems to indicate that improved direct cooling of the CPU might result in lower total power cost.

This Intel promo is suggesting that cooling is best left to the servers I think... the idea being to NOT cool the whole room... but the total concept involves a lot more than that.
http://www.intel.com/content/dam/doc...tion-brief.pdf

Finally, this paper is easiest to understand and a bit more relevant. They are of course measuring the total datacenter power usage, and results don't completely correlate to a room with a gaming PC in it... but I think it works out to be the same in theory.
http://www.apcmedia.com/salestools/J...TR88_R1_EN.pdf

The testing
revealed a complex interaction between server power
and total data center energy consumption in which
energy savings is realized within a temperature sweet
spot. This temperature sweet spot varies by equipment,
containment solution, and other factors.
> Executive summary
white

Since you are going to operate the AC at a temp that is comfortable for you, it seems to me that it would have to be tested. I'm guessing, based on all this, that it's more efficient to NOT use an AC at all. Because they aren't really efficient and good water cooling is efficient. There are more efficient methods of cooling rooms though, which I have seen in some places where it makes sense. Evaporative cooling works very well in dry climates and works even a bit in damp climates like the Gulf Coast.

[gmat]

So what this is all leading to is that the PC will actually produce less total heat if it is kept cooler? So the AC will work less?

No not at all. The heat produced is just extracted from the PC in a more efficient manner, but if you consider the system as a whole it amounts to the same.

Now if we consider the small savings in power consumption of cooler chips, i think those are balanced out by pump power anyway, in the end the overall system power is about the same (of course, it's more efficient in removing heat from the electronics... but that's not the point).
In a datacenter, where you can mutualize pumps and liquid, you mitigate the pumping power losses and have an overall more efficient cooling (also because you dont have to "fight" vs hot alleys and extract very hot air from there...). But it's a very different situation from a "simple" gamer room.

[Manicdan]

an AC system is not the same as a radiator system
its much more efficient to just move heat than it is to invert it.

a water system has a pump, fan, radiator. an ac system has to use a compressor in addition. to get more cooling with a water loop, you may not need to increase pump power by anything, fans are probably linear (if 1 fan is good for 100w, then 5 fans for 500w can be pretty close to expected), and radiators dont add that much energy consumption as you get more. but for an ac you need a bigger compressor to get more cooling. so in a quick summary, to go from .2 C/W to .1 with water, you might need 20% more power, but to get the same increase with an AC system you probably need 100% more power.


for the data center you see they found a sweet spot. at a certain point the power savings is not worth the extra energy needed to get the temps that low. those guys are talking about chips that run 100% of the time and thousands of them in a big room. so they need a whole different kind of system to keep temps down than us typical users. going from air to water could shave off 20c-50c depending on the parts and overclock, while we only add an extra 10w to the system for a pump and a few fans. so its easy to see how quickly we can get back those 10w thanks to how much we shaved off the temps. the data center however to get an extra few C, has to reduce the whole rooms temp by the same amount, or get beefier heatsinks with a better C/W (just cause the room is 60F dosnt mean the cpus are as well) i think it would be interesting to see how much power consumption is saved by spending more on better heatsinks, and how long it would take to pay back that cost.

EDIT, tonight i might turn off my fans and run a stress test and see how much my power draw goes up as i let the water warm up just to show you guys how real this can/cant be.

[prava]

alot of the hardware in the pc generates quite a large amount of heat and some resorted to water-cooling to deal with the heat that would have been dissipated into the air. the room is then cooled by air conditioning.

my question is, does the use of water-cooling in a way lower the overall power consumption in the house?

What nobody said, and this might interest you, is that by running water-cooling you might be able to put the rads of your computer elsewhere and thus take all that heat away directly. That is something you can not do with air cooling, and if you have the chance to put the rad somewhere else (like a basement, for instance) it might be a good idea to look into it.

[Manicdan]

heres an actual example for you guys.

i ran OCCTs built in cpu stress test that was great for not causing a spiky load. when it first started there was only a 2W swing in load

the chart is of the northbridge temps which is water cooled and shows a very pretty curve as temps go up. the first part up to 8 minutes is with fans running and im just letting the water warm up. after i turned off the fans you can clearly see the linear growth in temps until it eventually caused an error (which im super excited about since i expected a crash)

the 229 watts is from the first 10-30 seconds as i waited for it to max out the cores. 234w seemed to be where it was capping out and is when i decided it was time for some extreme temps. so a 5W change is found in just a normal growth as water increases to its peak temp. going up to more extreme temps, even though just a little higher, power draw increased by 12 more. (all numbers were from a kill-a-watt meter from the wall)

[zanzabar]

but we need a diffrence from air or stock v liquid, it should be about the same for most people.

[earthwormjim]

but we need a diffrence from air or stock v liquid, it should be about the same for most people.

No we really do not. What we need are temperature measurements of the devices, and the power consumed by the computer as temperature varies. Manicdan provided that data.

You can easily simulate air cooling temperatures by simply turning down your radiator fans or turning them off. You can also just measure the power consumption of the pump with a multi-meter and subtract that when simulating the air cooling temps.



Manicdan do you have your GPU watercooled? I'd love to see some test data on a GPU, they're typically high leakage parts.

[Manicdan]

Manicdan do you have your GPU watercooled? I'd love to see some test data on a GPU, they're typically high leakage parts.

sure do. but its a 680 which has like the worst overvoltage ever (none) so it may not draw as much as the cpu did, but it will be stable over a longer range of temps and thus more datapoints.

[Proximon]

This has been a great education. So ideally Emperor would water cool with his rads outside the room if possible, to keep heat in the room to a minimum and avoid the AC all together, which would minimize power consumption. I have taken this same question from people in hot places in the past, and not had a good answer.

[Manicdan]

as requested:



here i recorded, in order of the charts: the gpus power consumption that it thinks it uses, the clock rate, temps, gpu load to show it never changed, and the voltage. the recording was set to 2s intervals, so your looking at 25 minutes of stressing through OCCT gpu test.

the top chart alone shows that what we were testing is real. the gpu itself knows its drawing more power as the temps were increasing. clock rates there so we can see the drop at 70c. the temp line shows slight hints around 25% in where i turned the fans to half speed. at around 50% in theres a clear jump where i unplugged 2 of them, so for the second half its 1 fan at 1000rpms cooling EVERYTHING, and its still colder than the stock heatsink could really do without going deaf. the bottom one is there to show a voltage drop at 70c


and the results:



same as before with reading from the wall using a kill-a-watt meter. when first turned on it was 309w at 45c. the 312w and 50c is the peak where all fans running at full speed kept it. onto the red icons is where i keep the fans, nearly silent, but power draw is higher, even though im saving a little power from not running them so hard. between the red and green icons its clear that power draw from the fans is really noticed, as the temps went up a little more, the power draw was nearly identical. the rest of the icons, green and the purple, were all with just one fan at 1000rpms and i had plenty of time to grab almost every C increase. once we hit 70c the gpu has a built in voltage and clock rate drop.

so overall if i ran the fans at full speed i can expect 312w and noise
if i ran them at half speed its really quiet and 318w
if i used inadequate cooling the power draw hits 333w and the gpu becomes unhappy, but even then it dosnt drop power draw by much, even though were getting a lower voltage and clocks.

i put the trend line in there to show what we can expect at the best and worst. so if i get a whole bunch of rads, we might only shave off 10w, but probably to do that we have to use fans too powerful to keep it on the curve. (the trend does actually factor in power from the fans, but only for my few examples.)

EDIT: the site shrunk the images a little, right click and choose view image to see full size clarity.

[gmat]

Yeah the actual solution is to displace the radiators entirely to another room or outside or as some do, underground - geothermal cooling, this is by far the best solution. Try & find the posts of people who dug holes under their house and buried tubing and water tanks deep enough, it's very efficient.

[MadHacker]

Yeah the actual solution is to displace the radiators entirely to another room or outside or as some do, underground - geothermal cooling, this is by far the best solution. Try & find the posts of people who dug holes under their house and buried tubing and water tanks deep enough, it's very efficient.

I have my rads in the other room as well, with the hose going throught the wall
it keeps it a bit cooler in the summer, and more silent, i run the fans at full speed but don't hear them
now if only i could watercool my monitors then it would reduce the heat that much more.