Guide to Delta-T in Water Cooling

Add Your Comments

A water-cooled PC, like any other, can generate a significant amount of heat. Some parts produce massive amounts of heat due to the incredible number of calculations needed to give us what we crave – realistic games, distributed computing projects that might change the future of science and encoding and decoding audio and video, to name just a few. I’m going to focus on two parts of your PC: the CPU and the GPU, as these produce most of the heat. I’ll  also touch on a few other critical issues you need to understand.

Why Am I Doing This?

I used to teach electronics to young folks. I’ve been water cooling for a few years, so I think I can help get the concept of heat load across to any new person looking to expand their knowledge and build a really nice rig. Hopefully this will aid people in understanding the big picture and answering the eternal question on the Overclockers Forums – “What stuff do I need to cool my rig?”.

Most importantly, I must mention all the testers who post scientific tests and not hack reviews. Otherwise, this article wouldn’t exist. These are people who, for science and curiosity, spent hundreds of dollars of their own money and hours of their time building and testing rigs. They spend countless hours of their personal time to inform us, the uninformed public. I would really like to thank Martin and Skinnee – they are an amazing part of our water cooling world. There are a few others, but Martin and Skinnee lead the way.

The Important Stuff

Okay, you’ve got a processor. How much heat does it generate under load? The heat output depends on the voltage and the current in the CPU. Increasing voltage and clock speed really increases your heat output. If a CPU uses 100 Watts of power, about 80% to 90% will be lost as heat as the work is done in the CPU. You have to remove that heat.

Different amounts of heat can be created while doing different activities/tasks. For example:

  • Basic web browsing
  • Gaming
  • Benching while overclocked to obtain the best possible scores

Consider the following. All of them come into play when determining how to deal with said heat.

  • Your budget
  • Acceptable noise levels
  • What can fit in your case
  • Expected cooling capability: do you want merely acceptable temperatures or the best possible temperatures?

Determining Heat Output isn’t a Perfect Science.

You’ll find numbers for heat output all over the place: Googling terms like “heat-load i7 965”, “AMD xxx”, “Q6600 overclocked”, etc., reading 3-7 pages of Google results and following links within the Google results should get you an idea of what wattage you create in your PC furnace.

CPU Heat Output

For example, a Q6600 at 3.9 GHz produces approximately 150 W of heat and an i7 at 4.0 GHz approximately 250 W. Right now, as I type, my i7 965 stock is at 80 W. My peak load is 340 W, but for a very, very short time. So take the minimum and maximum with a grain of salt. These are max. testing loads, so real world is a bit lower if you’re a big gamer. My point is that it’s a lot of heat and you need to remove it. I use 250 W as my stress wattage for calculations on my chip.  Is it exactly what you’ll see?  Probably not but it’s what I have decided to use.

GPU Heat Output

GPUs are a big issue now. Some people are moving to water cooling, having never done it before, purely because their GPU screams like an undead banshee while gaming.

The following links are from another forum. From what others have said these folks did a TON of work and are trusted. They were able to look at just the GPU heat output (or in English here) of many cards, rather than full system power. The latter isn’t what we need here.

It would be best to calculate your CPU and GPU heat output before reading on.

How Basic Water Cooling Works

The CPU and GPU generate P watts of heat energy. The heat is transferred to the water via blocks on the CPU and GPU. The cooler the water, the more heat is transferred. As the water heats up, the heat is removed by the radiator. The more efficient the fan/radiator combination is, the cooler the water. The cooler the water, the cooler your parts. Simple.

Basic water cooling uses a pump, reservoir, a block (or blocks), a radiator (or radiators), and ambient air. Our water can be no cooler than the air temperature. A room at 12°C will keep the chips much cooler than a room at 30°C. Remember this!

Summers are much hotter and many back off on their overclocks in the summer. For this reason, there tends to be more discussion about the use of  chilled water, mini fridges, etc. during the summer. If you try to compare your cooling results to someone else remember ambient room temps and even the quality of the chip matters.

Delta T (DT) and Why it’s so Important to Understand it

DT is the foundation of your water cooling loop. The better your DT, the cooler your chips are. In water cooling, DT is simply the difference between the ambient air temperature and the water temperature on the outgoing side of the radiator. Room temperature vs. water temperature: that’s it. You can’t remove all the heat – no system is 100% efficient, nor can you go below ambient room temperature.

When you boot up a powered off, water-cooled PC, the water and your CPU are at room temperature. When you boot the PC up, the chip gets hot very fast. The water moves over the chip, it begins to remove heat, the heat goes to the radiator, and some of the heat is removed. Not all of it can be removed. You have to know a lot of thermodynamics theory deeply (more than me) to know exactly why. The water begins to warm up slowly, and in time it reaches a balance: an equilibrium. Heat is made and heat removed, the loop is stabilized and temperatures will not change.

If you change the room temperature, the load on the loop or your fan speed,  the loop needs to readjust. This is when we like to measure our cooling ability – usually 30 minutes at a stable load is long enough to begin to measure. If you increase your cooling capability, the water will get cooler.

Water temperatures in a stabilized loop, amazingly, are very similar anywhere in the loop. There is only a 2-3°C maximum difference between the radiator out temperature and the CPU out temperature; this has been verified by Skinnee. Remember, the water can’t remove all the heat, some is transferred to the air. Your radiator size, efficiency and fans play a big part in this. Look at it this way – it’s a system built on many parts and within the laws of physics. Every part affects the other.

Let’s Talk About What a “Good” DT is

A CPU loop needs a good DT. Under 10°C is just fine, getting closer to 5°C is very nice and important if you want big overclocks. Getting under 5°C is just overdoing it, unless you’re very extreme, need it for benching or just want a challenge. On an average CPU loop shoot for under 10°C and adjust your overclocks to be fine under the temperatures suggested by the CPU and GPU manufacturers.

A GPU loop used to be fine with a 15° to 20°C DT. If you’re a big GPU overclocker, then shoot for 15°C or lower. The Voltage Regulator Modules (VRM) on these new cards can be affected by temps. If you’ve got a GTX 280 like me, don’t worry about it, it’s the 5970 and the GTX480 and other really hot cards that can have this issue. A lower DT might be needed for a GPU loop if you’re a big overclocker – thats up to you as you design your setup.

So, How Do You Calculate All of This?

Use a bit of math, decipher some graphs – it’s easy! Just kidding: it took me more than a few attempts to figure it out, but you can do it easily with help. Thanks to Martin (retired) and Skinnee we have the data. Bless them.  The site that the chart comes from is:

http://www.skinneelabs.com/

Let’s look at an example, with my CPU standard of 250 W and this graph for a HWLabs SR1 360:

HWLabs SR1 360 Heat Load Chart

HWLabs SR1 360 Heat Load Chart - Courtesy skinneelabs.com

This is a quality radiator with low fins per inch, great for low RPM, low noise fans.

Go to the second chart, use the red line. It’s a good quality, popular, well performing fan at 1407 RPM, a good middle ground. At the bottom of the chart, find 250 watts and go up to the red line. A 5.5°C DT, very nice.

Remember my comment about a CPU likes low DT but a GPU doesn’t need it as much? Let’s toss a 250 W GPU into the same loop. You’re at 500 W now. Resevoir-Pump-Radiator-CPU-GPU type loop. Look what happens to the DT, it’s over 11°C now. Your CPU isn’t happy anymore, but the GPU is just fine. In this case you’re “under-radded” and need to split the loop or get your DT under control with more fans, better/more radiators, better fans, etc.

Remember This?

  • Your budget
  • Acceptable noise levels
  • What can you fit in your case?
  • Expected cooling capability

In many situations, splitting the loop is an easier option to keep temperatures under control. You can only put so much in one loop due to flow rate restrictions. Splitting the loop means having one fully separate loop for the CPU and one for the GPU(s). Two pumps, two radiator setups, two reservoirs, etc. It was uncommon a few years ago unless you had 2+ GPUs. Now the GPUs are incredibly hot and it’s becoming more common. It costs more but keeps the CPU temperature low in it’s own loop while having a higher DT loop just for the GPU.

Many of us do have massive GPU setups these days. If you want to play right with water cooling, you pay. You can easily go past your budget once you start. It’s amazing what a proper water cooling setup costs.

Lots start out with the wrong radiator type or too small a radiator. Some think they can live with the high speed fans to compensate for too small rads, but end up buying another radiator and trying to fit it in the case. Many times you just have to accept your beloved case isn’t going to work, or you’re going to have to hang radiators on the outside, or even make a separate radiator box. You can’t trick physics and think you don’t need all the radiators you actually do need.

In summary, do your homework, plan for your rig and the machine you might be running in a year. Perhaps you might want a second GPU in crossfire? Plan now.

Your DT matters, it’s worth the effort to understand.

Conumdrum

Thanks to all the input from many OCF members to get this cleaned up for Sticky use. I hope it guides many new folks to make the right choices.

Leave a Reply

Your email address will not be published. Required fields are marked *

Discussion
  1. You find references online to see what your wattage is. Then you use the supplied links at Skinneelabs.com to decide whats best for you. There is no exact science or ways. You decide what your wattage is total and what your Dt is good for you. Your noise level and your amount of space to fit the rads. That will determine the raddage you need.

    You use that info to decide what you want to do. There is no correct answer.

    To make it easy, no math or etc involved, buy two 120x3 Swiftech/XSPC etc etc brand of radiator. Six AP-15 fans and a fan controller with 30 watts per channel. Easy.

    And your case is too small unless you want to hang a rad off the back of the case.

    I wrote the sticky.
    I want to watercool 2x gtx 670, i7 3770k clocked to 4,3 ghz and a maximus v formula maybe also the ram corsair gt 16gb 2133 (just beceause it looks nice) in a nzxt switch case how many rads do i need?
    I have read this over and over and over. It's fun :D

    I ROFL when I saw enters GPU at 30c and leaves at 60c I'm like WTF system are you running. Glad that got cleared up :rofl:
    I stand corrected. your right. the temp doesn't go high enough to mean much. So flow restriction will play a greater role. series tubing will result in smaller tubing and hence better flow. correct.
    The water reaches an equilibrium, so the water temp differences between components aren't drastic like you're thinking. As I mentioned in my previous post, temp differences are only like 1-2C anywhere in the loop. The heat in watts for each component and the cooling capacity of the rad/fans combine to make the water get to a certain temperature.
    MySchizoBuddy
    heat transfer is best if there is a temp difference. for eg.

    water at 30C enters one GPU and leaves at 60C this 60C enters the CPU now this already hot water has less heat trapping capability. Thats thermodynamics. If water at 30C enters both GPU and CPU at the same time they both now have the same heat trapping capability. Thats thermodynamics.

    According to you a 30C water can remove the same amount of heat as 60C water thats wrong.


    Do you realize just how much heat it would take to raise the temperature of water 30C with as little contact as it makes with the block? It's enormous. You're looking at 1-2C difference MAX from the hottest to the coldest point in your loop, and at that point, you're gonna see more of a difference from flow than from any sort of gain you'd get from having 2C colder water.
    heat transfer is best if there is a temp difference. for eg.

    water at 30C enters one GPU and leaves at 60C this 60C enters the CPU now this already hot water has less heat trapping capability. Thats thermodynamics. If water at 30C enters both GPU and CPU at the same time they both now have the same heat trapping capability. Thats thermodynamics.

    According to you a 30C water can remove the same amount of heat as 60C water thats wrong.
    MySchizoBuddy
    what if you connect the GPU and CPU in parallel. So the cool water from the radiator goes to CPU and GPU in parallel and the output from the GPU and CPU go back in the inlet of the radiator. you will need a T junction to split and join the incoming and outgoing loop


    DT is 100% about your heatload and your generated heat. Your question has NO bearing on the fundamental physics of heat made vs dissipated in a radiator.

    No matter if series GPUs make 500 watts of heat or the same cards in parellel making 500 watts of heat.

    You STILL end up with the same wattage. And the same required cooling.

    Some folks get pretty silly and try to make magic, but physics always wins...........:clap:
    :welcome: to OCForums!

    It won't make much of a difference because the water temps throughout the loop is always within 1-2C of each other, no matter where you measure the temp. So, water temp after the rad and water temp after the GPU block will be almost the same, maybe 1-2C different.

    When splitting into parallel, you have to be careful with flow rate and block restriction to make sure you have enough flow and pressure for the blocks on each split.
    what if you connect the GPU and CPU in parallel. So the cool water from the radiator goes to CPU and GPU in parallel and the output from the GPU and CPU go back in the inlet of the radiator. you will need a T junction to split and join the incoming and outgoing loop
    But if we do not ask you stupid questions then how else will you spend your time? Not playing around on your own system I hope. That is for NERDS!