- Jan 12, 2007
- 16,689
- 658
- 60,190
Corsair H110i GTX Watercooling Performance Testing
I had the opportunity to put the Corsair H110i GTX closed loop liquid cooler through its paces on my watercooling test bench. This has allowed me a couple of very distinct advantages over CPU and case load testing using benchmark tools and tests, which, don’t get me wrong – are very beneficial from a real-world perspective. But I wanted to dig deeper on this cooler (like I do on all watercooling components) and actually find out specific heat loads and flow testing for a comprehensive set of quantitative results that can be compared on an apples-to-apples basis to other watercooling tests and reviews.
Overview
The Corsair H110i GTX is an all-in-one liquid cooler, which means that the system is factory sealed and should never need maintenance (i.e, coolant changes or tubing replacement). The pump is situated atop the CPU block, which is comprised of milled and polished copper and adorned with a pre-applied thermal paste. The CPU block/pump unit is covered by a plastic Corsair cover and has back-illuminated LEDs to light the 3-sail Corsair logo.
The tubing is sleeved with black, braided nylon sleeving that runs the length of the tubing from the pump fittings to the radiator fittings. The radiator is a 2x140mm fan format, and this is the only sized set of fan mounts able to be mounted. Likewise, this also means that the radiator will only mount to a case with a 2x140mm set of fan mounts, as well, so make sure your case has the ability to support this cooler. There is not an option of mounting 120mm fans or mounting the radiator to 120mm case mounts.
The radiator is aluminum and sports a fin/fold density of around 20 per inch. This means that it is considerably more restrictive than normal watercooling radiators that utilize a more common 9-12 FPI density. However, this allows the Corsair unit to potentially dissipate more heat due to the addition of these fins ‘wicking’ heat away from the radiator tubes that run the length of the radiator and allow for more surface area contact with air to dissipate the heat from the coolant inside. The radiator is rather thin by comparison (30mm) to most traditional watercooling radiators, as most are 35mm or greater. The actual radiator core itself is just around 22mm thick as the radiator bezel makes up the remainder of the overall width, and is meant to allow fans to be mounted as well as to protect the core inside.
The H110i GTX pump is powered by a SATA power connector and also has a dual 3-pin fan tail for controlling the two 140mm SP140L fans (2300 RPM max) by connecting the PWM header to your motherboard’s CPU_FAN header. This allows the unit to monitor the CPU load temperatures and vary the fan speeds as needed. For my testing, I connected the fans to a fan controller to manually control the speed and cooling potential at each setting. A mini-USB connector located on the pump allows the unit to be connected to a USB header on your computer to allow Corsair’s Link software the ability to control and monitor your H110i GTX from your Windows desktop. Of course, there is the typical offering of mounting hardware and options for both Intel and AMD processors.
Overall, the H110i GTX is a handsome looking cooler and the radiator size itself is larger than you probably expect until you have it in your hands. However, the unit as a whole is very lightweight, which tells me that the radiator is made of aluminum (obviously), has nominal LxWxH volume, and due to this smaller volume, it also has a reduced volume of coolant it can hold.
A Look Inside
At this point, the H110i GTX I am testing is still covered by warranty. What I am about to do (opening the unit) will permanently void this coverage, so be advised before attempting this on your own cooler. You will not be able to return/exchange the unit for any reason if the unit fails.
There are 2 rubber caps covering the ends of the tubing and the nylon hose braid. With this rubber cap removed, I was able to simply slide the tubing off of the aluminum barbs on the radiator. I then carefully emptied the coolant of the unit into a beaker to see what overall volume it held. This is where I was surprised – the entire cooler only held around 150ml of a clear, (what I assume to be) glycol coolant.
The tubing is black, rubberized, thick-walled tubing that is approximately 3/8” ID (inside diameter) and 5/8” OD (outside diameter) or, 10mm ID by 16mm OD. I like the braided sleeving, but the tubing itself is very rigid and has limited flexibility.
At this point, I modified the setup for my test bench. I connected the pump unit to my custom built reservoir and my King 7530 3.5 GPM (gallons per minute) flow meter to see what kind of flow performance the pump actually had. This has been one of the long-standing debates around these coolers – what the actual flow rate of the coolant is, and due to voiding of the warranty, not many people have actually wanted to find out.
The flow for this cooler is very low and didn’t even register against the float in my flow meter, which tells me the pump also operates at very low pressure. Given that this cooler is never meant to operate in a capacity more than the pump and radiator alone, this isn’t at all surprising to me. I removed the unit from the flow meter setup and decided to go with an old-fashioned flow rate test, falling back to a digital stopwatch and a 1 liter beaker. I timed five separate, one-minute intervals to ensure I was getting consistent readings.
What I found was somewhat (but not completely) surprising. The pump consistently produced a flow rate result of 0.25 GPM (gallons per minute) or 0.95 LPM (liters per minute). For comparison, most of us in the watercooling community try to shoot for a flow rate goal of 1.0 GPM (3.785 LPM) in their custom loops. Another curious fact I noticed was the drop in flow rate when I raised the tubing end from being on-level to the pump to 18 inches above; flow rates dropped by almost 0.25 LPM – down to around 0.70 LPM (0.185 GPM).
Thermal Test Setup and Methods
This CPU cooler is designed to only cool processors, so let’s move onto the thermal load testing.
The lab setup for this set of individualized unit testing consists of an ATX power supply (A) powering the Corsair pump, my independent fan controller and the CrystalFontz CFA-633 (B) data collection board. The CFA-633 is connected to my laptop via USB and logs out the temperature sensor data every second. It also has the capability to monitor and report fan RPM, but for this test, I wanted static speed control of the Corsair SP140L fans, so I opted to use my Scythe Kaze Master fan controller (C). I am utilizing seven Dallas DS18B20 one-wire temperature sensors with a +/-0.5C accuracy range per second in waterproof stainless and rubber sleeves. Four of these sensors are utilized to monitor ambient air temperature, which will all be averaged together for greater accuracy.
Three sensors are used within the loop - one in the reservoir chamber and the remaining two sensors are mounted inside T-fittings (D) to allow the loop water to flow around the sensors in the direct line of the flow without disrupting flow rates or adding any additional restriction. These are also averaged for accuracy.
Inside the reservoir (E), I have two, 300 watt aquarium heaters that are connected to a single power strip. This is powered by an A/C variable transformer (F) and the dial on top allows me to alter the current supplied to the outlet, which is monitored with a Kill-A-Watt meter (G) that displays the live power draw on the outlet. By turning the dial on the A/C variable transformer, I can power the heaters to supply anywhere from 1 to 630 watts of heat to the reservoir, and therefore directly into the loop to simulate exact heat load scenarios.
The cooler itself is then connected to the custom reservoir in a complete loop with some additional fittings and 3/8 inch ID tubing. The loop is tested using distilled water.
I decided to test what I felt to be fairly common points of thermal load testing: 95w, 130w, 150w, 175w, 200w, 225w, 250w and 300w.
Understanding TDP
Thermal design power, or TDP, is the manufactured design energy draw and heat output of a component in a PC – whether that is at stock speed or overclocked.
For more on TDP, please see the section in the Watercooling Lab Equipment page on understanding TDP.
For each of these load tests, the system was tested with fan settings of 1200 RPM, 1800 RPM and 2300 RPM (full speed). The reasons behind the heat load wattages I am testing is due to the typical CPU TDPs of both stock and overclocked processors that might typically be used. Also, when you start to establish the cooling performance curve, it doesn’t matter what CPU you use since I’m testing thermal load in watts. If a TDP falls between a testing point, the curve still maintains the expected output between two points.
Here we see three distinct groupings of testing results – each grouping is the push and pull of the fan RPM chosen; 1200 RPM, 1800 RPM and full speed at 2300 RPM. In each grouping, notice that the ‘pull’ fans actually keep the loop just a bit cooler in most scenarios by an average of 0.5°C – 1.0°C. Corsair indicates in its documentation that the unit should be setup as an air intake, but it doesn’t specify that the unit necessarily should be in pull configuration. However, given that the unit is typically installed inside the case with the radiator mounted to the case fan mount locations, a pull setup is about the only way you can accomplish this. So, by default, Corsair coaxes you into this installation and you’d have to actually find different screws to assist in alternative mounting.
Considering that you’d typically not see a CPU overclock over 250 watts (let alone 300 watts), I simply wanted to continue the graph out to 300w as a way to show how the unit fared in the event it was used for graphics card mounting and cooling. It does offer some interesting data in terms of visual representation.
Understanding Temperature Delta (or DT)
In the world of cooling (especially watercooling), the term ‘Delta’ is used for determining the performance of the cooling solution in question. This is the temperature of the water in the loop as compared to the temperature of the ambient room air. This is important not to confuse these temperatures with your CPU reported temps in CoreTemp, SpeedFan, RealTemp, etc., as these readings are reported from the CPU die thermal sensors at any single second, but do not represent the actual coolant temps. The temperature of the loop water is the basis that we use for watercooling performance evaluation. For more information on understanding temperature delta (DT), please visit the section on temperature delta in the Watercooling Lab Equipment page.
It should also be noted that it is impossible to have a cooling delta that is equal to or less than zero with normal air or liquid cooling as you are using ambient air to cool the loop coolant, and the coolant itself can never be equal to, or lower than ambient (due to those pesky laws of physics). As you would expect, the lower the delta, the better the cooling performance.
Pure water is always a better cooling medium than coolants and additives when it comes to watercooling loops, but coolants are sometimes required to prevent corrosion between mixed metals such as copper and aluminum being present in the same liquid loop (such as the Corsair H110i GTX and most other closed loop coolers). However, coolants aren’t going to make a substantial difference in cooling properties, it is just notable that there is a technical difference, however small it may be. Still, once you get to the point of heat exchange at the radiator, you’re now performing radiator-to-ambient air exchange as you would in a normal air cooler. This is why water and liquid cooling is still technically air cooling, although the methods of transporting the heat energy and where it is exchanged with the ambient air is different. This is why you cannot cool lower than ambient room temperatures, as mentioned earlier.
Fan Noise Levels at Testing Levels
Fan noise is often a matter of personal perception, but we can measure the sound level output of the fans in a quantitative way: decibels. Measurement of sound using the decibel is a logarithmic factor, so as the value of the measured level increases, it increases as a steady multiplier. This means that every increase of 10dB means a doubling in perceived noise/sound level.
Using my digital decibel meter, I captured the noise levels at 0 RPM, 1200 RPM, 1800 RPM and 2300 RPM at a distance of one foot (30cm):
Fans off shows a room ambient sound level at 32.1 dB. At 1200 RPM, things are mildly higher at 38.2 dB. But look at 1800 RPM , 55.4 dB and 2300 RPM, 64.2 dB – around 4-8 times as loud as ambient room noise levels. 30 dB is comparable to a quiet room or library while 60 dB is comparable to a normal speaking conversation. 70 dB is comparable to noise levels inside your car while driving at highway speeds.
Conclusion and Final Thoughts
The Corsair H110i GTX is a great looking cooler that offers an impressive radiator footprint and considerable cooling potential for almost every CPU cooling application, whether it’s at stock speeds or the heavily overclocked. It requires a bit more planning and a slightly higher difficulty level than installing an aftermarket heatsink with a backplate simply because you must also manage the routing of tubing and the mounting of the radiator. Outside of these items, it is a relatively simple DIY cooler project for most novice users to enthusiasts alike. Overall, I was impressed with the cooling ability of the H110i GTX radiator when utilized with good airflow, although this masks the bigger issue of the low-flowing pump as you immediately see cooling results slipping (higher delta) as you begin to dial fans back to quieter sound levels and lower speeds. Most people don’t realize that having two (2) 140mm fans running at 2300+ RPM are actually quite loud at 64+ dB, but this is where you are able to reach the maximum performance of the cooler. Lower speeds and sound levels directly relate to warmer temperatures as we saw earlier in the chart.
For the set it and forget it user, the H110i GTX is a decent cooler if you’re wanting to make a jump into the small, liquid cooler scene. Of course, there are other high-end air coolers that will still run you less money with similar performance, but this is about liquid, isn’t it? However, if you’re looking to get into water or liquid cooling with the potential to expand at a later date, there are other options relatively similar in design, but a bit more in cost depending on solution chosen. If you are of the first group, the H110i GTX is a solid choice and a handsome cooler, if you have the space to mount it. If you’re part of the second group, you likely will want to do more research to fit your future plans and long-term budget.
I had the opportunity to put the Corsair H110i GTX closed loop liquid cooler through its paces on my watercooling test bench. This has allowed me a couple of very distinct advantages over CPU and case load testing using benchmark tools and tests, which, don’t get me wrong – are very beneficial from a real-world perspective. But I wanted to dig deeper on this cooler (like I do on all watercooling components) and actually find out specific heat loads and flow testing for a comprehensive set of quantitative results that can be compared on an apples-to-apples basis to other watercooling tests and reviews.
Overview
The Corsair H110i GTX is an all-in-one liquid cooler, which means that the system is factory sealed and should never need maintenance (i.e, coolant changes or tubing replacement). The pump is situated atop the CPU block, which is comprised of milled and polished copper and adorned with a pre-applied thermal paste. The CPU block/pump unit is covered by a plastic Corsair cover and has back-illuminated LEDs to light the 3-sail Corsair logo.
The tubing is sleeved with black, braided nylon sleeving that runs the length of the tubing from the pump fittings to the radiator fittings. The radiator is a 2x140mm fan format, and this is the only sized set of fan mounts able to be mounted. Likewise, this also means that the radiator will only mount to a case with a 2x140mm set of fan mounts, as well, so make sure your case has the ability to support this cooler. There is not an option of mounting 120mm fans or mounting the radiator to 120mm case mounts.
The radiator is aluminum and sports a fin/fold density of around 20 per inch. This means that it is considerably more restrictive than normal watercooling radiators that utilize a more common 9-12 FPI density. However, this allows the Corsair unit to potentially dissipate more heat due to the addition of these fins ‘wicking’ heat away from the radiator tubes that run the length of the radiator and allow for more surface area contact with air to dissipate the heat from the coolant inside. The radiator is rather thin by comparison (30mm) to most traditional watercooling radiators, as most are 35mm or greater. The actual radiator core itself is just around 22mm thick as the radiator bezel makes up the remainder of the overall width, and is meant to allow fans to be mounted as well as to protect the core inside.
The H110i GTX pump is powered by a SATA power connector and also has a dual 3-pin fan tail for controlling the two 140mm SP140L fans (2300 RPM max) by connecting the PWM header to your motherboard’s CPU_FAN header. This allows the unit to monitor the CPU load temperatures and vary the fan speeds as needed. For my testing, I connected the fans to a fan controller to manually control the speed and cooling potential at each setting. A mini-USB connector located on the pump allows the unit to be connected to a USB header on your computer to allow Corsair’s Link software the ability to control and monitor your H110i GTX from your Windows desktop. Of course, there is the typical offering of mounting hardware and options for both Intel and AMD processors.
Overall, the H110i GTX is a handsome looking cooler and the radiator size itself is larger than you probably expect until you have it in your hands. However, the unit as a whole is very lightweight, which tells me that the radiator is made of aluminum (obviously), has nominal LxWxH volume, and due to this smaller volume, it also has a reduced volume of coolant it can hold.
A Look Inside
At this point, the H110i GTX I am testing is still covered by warranty. What I am about to do (opening the unit) will permanently void this coverage, so be advised before attempting this on your own cooler. You will not be able to return/exchange the unit for any reason if the unit fails.
There are 2 rubber caps covering the ends of the tubing and the nylon hose braid. With this rubber cap removed, I was able to simply slide the tubing off of the aluminum barbs on the radiator. I then carefully emptied the coolant of the unit into a beaker to see what overall volume it held. This is where I was surprised – the entire cooler only held around 150ml of a clear, (what I assume to be) glycol coolant.
The tubing is black, rubberized, thick-walled tubing that is approximately 3/8” ID (inside diameter) and 5/8” OD (outside diameter) or, 10mm ID by 16mm OD. I like the braided sleeving, but the tubing itself is very rigid and has limited flexibility.
At this point, I modified the setup for my test bench. I connected the pump unit to my custom built reservoir and my King 7530 3.5 GPM (gallons per minute) flow meter to see what kind of flow performance the pump actually had. This has been one of the long-standing debates around these coolers – what the actual flow rate of the coolant is, and due to voiding of the warranty, not many people have actually wanted to find out.
The flow for this cooler is very low and didn’t even register against the float in my flow meter, which tells me the pump also operates at very low pressure. Given that this cooler is never meant to operate in a capacity more than the pump and radiator alone, this isn’t at all surprising to me. I removed the unit from the flow meter setup and decided to go with an old-fashioned flow rate test, falling back to a digital stopwatch and a 1 liter beaker. I timed five separate, one-minute intervals to ensure I was getting consistent readings.
What I found was somewhat (but not completely) surprising. The pump consistently produced a flow rate result of 0.25 GPM (gallons per minute) or 0.95 LPM (liters per minute). For comparison, most of us in the watercooling community try to shoot for a flow rate goal of 1.0 GPM (3.785 LPM) in their custom loops. Another curious fact I noticed was the drop in flow rate when I raised the tubing end from being on-level to the pump to 18 inches above; flow rates dropped by almost 0.25 LPM – down to around 0.70 LPM (0.185 GPM).
Thermal Test Setup and Methods
This CPU cooler is designed to only cool processors, so let’s move onto the thermal load testing.
The lab setup for this set of individualized unit testing consists of an ATX power supply (A) powering the Corsair pump, my independent fan controller and the CrystalFontz CFA-633 (B) data collection board. The CFA-633 is connected to my laptop via USB and logs out the temperature sensor data every second. It also has the capability to monitor and report fan RPM, but for this test, I wanted static speed control of the Corsair SP140L fans, so I opted to use my Scythe Kaze Master fan controller (C). I am utilizing seven Dallas DS18B20 one-wire temperature sensors with a +/-0.5C accuracy range per second in waterproof stainless and rubber sleeves. Four of these sensors are utilized to monitor ambient air temperature, which will all be averaged together for greater accuracy.
Three sensors are used within the loop - one in the reservoir chamber and the remaining two sensors are mounted inside T-fittings (D) to allow the loop water to flow around the sensors in the direct line of the flow without disrupting flow rates or adding any additional restriction. These are also averaged for accuracy.
Inside the reservoir (E), I have two, 300 watt aquarium heaters that are connected to a single power strip. This is powered by an A/C variable transformer (F) and the dial on top allows me to alter the current supplied to the outlet, which is monitored with a Kill-A-Watt meter (G) that displays the live power draw on the outlet. By turning the dial on the A/C variable transformer, I can power the heaters to supply anywhere from 1 to 630 watts of heat to the reservoir, and therefore directly into the loop to simulate exact heat load scenarios.
The cooler itself is then connected to the custom reservoir in a complete loop with some additional fittings and 3/8 inch ID tubing. The loop is tested using distilled water.
I decided to test what I felt to be fairly common points of thermal load testing: 95w, 130w, 150w, 175w, 200w, 225w, 250w and 300w.
Understanding TDP
Thermal design power, or TDP, is the manufactured design energy draw and heat output of a component in a PC – whether that is at stock speed or overclocked.
For more on TDP, please see the section in the Watercooling Lab Equipment page on understanding TDP.
For each of these load tests, the system was tested with fan settings of 1200 RPM, 1800 RPM and 2300 RPM (full speed). The reasons behind the heat load wattages I am testing is due to the typical CPU TDPs of both stock and overclocked processors that might typically be used. Also, when you start to establish the cooling performance curve, it doesn’t matter what CPU you use since I’m testing thermal load in watts. If a TDP falls between a testing point, the curve still maintains the expected output between two points.
Here we see three distinct groupings of testing results – each grouping is the push and pull of the fan RPM chosen; 1200 RPM, 1800 RPM and full speed at 2300 RPM. In each grouping, notice that the ‘pull’ fans actually keep the loop just a bit cooler in most scenarios by an average of 0.5°C – 1.0°C. Corsair indicates in its documentation that the unit should be setup as an air intake, but it doesn’t specify that the unit necessarily should be in pull configuration. However, given that the unit is typically installed inside the case with the radiator mounted to the case fan mount locations, a pull setup is about the only way you can accomplish this. So, by default, Corsair coaxes you into this installation and you’d have to actually find different screws to assist in alternative mounting.
Considering that you’d typically not see a CPU overclock over 250 watts (let alone 300 watts), I simply wanted to continue the graph out to 300w as a way to show how the unit fared in the event it was used for graphics card mounting and cooling. It does offer some interesting data in terms of visual representation.
Understanding Temperature Delta (or DT)
In the world of cooling (especially watercooling), the term ‘Delta’ is used for determining the performance of the cooling solution in question. This is the temperature of the water in the loop as compared to the temperature of the ambient room air. This is important not to confuse these temperatures with your CPU reported temps in CoreTemp, SpeedFan, RealTemp, etc., as these readings are reported from the CPU die thermal sensors at any single second, but do not represent the actual coolant temps. The temperature of the loop water is the basis that we use for watercooling performance evaluation. For more information on understanding temperature delta (DT), please visit the section on temperature delta in the Watercooling Lab Equipment page.
It should also be noted that it is impossible to have a cooling delta that is equal to or less than zero with normal air or liquid cooling as you are using ambient air to cool the loop coolant, and the coolant itself can never be equal to, or lower than ambient (due to those pesky laws of physics). As you would expect, the lower the delta, the better the cooling performance.
Pure water is always a better cooling medium than coolants and additives when it comes to watercooling loops, but coolants are sometimes required to prevent corrosion between mixed metals such as copper and aluminum being present in the same liquid loop (such as the Corsair H110i GTX and most other closed loop coolers). However, coolants aren’t going to make a substantial difference in cooling properties, it is just notable that there is a technical difference, however small it may be. Still, once you get to the point of heat exchange at the radiator, you’re now performing radiator-to-ambient air exchange as you would in a normal air cooler. This is why water and liquid cooling is still technically air cooling, although the methods of transporting the heat energy and where it is exchanged with the ambient air is different. This is why you cannot cool lower than ambient room temperatures, as mentioned earlier.
Fan Noise Levels at Testing Levels
Fan noise is often a matter of personal perception, but we can measure the sound level output of the fans in a quantitative way: decibels. Measurement of sound using the decibel is a logarithmic factor, so as the value of the measured level increases, it increases as a steady multiplier. This means that every increase of 10dB means a doubling in perceived noise/sound level.
Using my digital decibel meter, I captured the noise levels at 0 RPM, 1200 RPM, 1800 RPM and 2300 RPM at a distance of one foot (30cm):
Fans off shows a room ambient sound level at 32.1 dB. At 1200 RPM, things are mildly higher at 38.2 dB. But look at 1800 RPM , 55.4 dB and 2300 RPM, 64.2 dB – around 4-8 times as loud as ambient room noise levels. 30 dB is comparable to a quiet room or library while 60 dB is comparable to a normal speaking conversation. 70 dB is comparable to noise levels inside your car while driving at highway speeds.
Conclusion and Final Thoughts
The Corsair H110i GTX is a great looking cooler that offers an impressive radiator footprint and considerable cooling potential for almost every CPU cooling application, whether it’s at stock speeds or the heavily overclocked. It requires a bit more planning and a slightly higher difficulty level than installing an aftermarket heatsink with a backplate simply because you must also manage the routing of tubing and the mounting of the radiator. Outside of these items, it is a relatively simple DIY cooler project for most novice users to enthusiasts alike. Overall, I was impressed with the cooling ability of the H110i GTX radiator when utilized with good airflow, although this masks the bigger issue of the low-flowing pump as you immediately see cooling results slipping (higher delta) as you begin to dial fans back to quieter sound levels and lower speeds. Most people don’t realize that having two (2) 140mm fans running at 2300+ RPM are actually quite loud at 64+ dB, but this is where you are able to reach the maximum performance of the cooler. Lower speeds and sound levels directly relate to warmer temperatures as we saw earlier in the chart.
For the set it and forget it user, the H110i GTX is a decent cooler if you’re wanting to make a jump into the small, liquid cooler scene. Of course, there are other high-end air coolers that will still run you less money with similar performance, but this is about liquid, isn’t it? However, if you’re looking to get into water or liquid cooling with the potential to expand at a later date, there are other options relatively similar in design, but a bit more in cost depending on solution chosen. If you are of the first group, the H110i GTX is a solid choice and a handsome cooler, if you have the space to mount it. If you’re part of the second group, you likely will want to do more research to fit your future plans and long-term budget.
Facebook
Twitter
Instagram