Corsair H110i GTX Performance Testing and Review

[rubix_1011]

I personally always prefer the theory over the application. my math degrees are in theoretical math for this reason. i can easily pick up a book and see what formula goes where myself. but to fully understand where the formula comes from and how it developed and be able to modify it on my own for a different use is where the true magic is for me.

so i appreciated and completely ate up the info in the review and the associated chart. from my previous posts it should be clear how quickly the data was to grasp and use for my own purposes. knowing the thermal load from the graph let me estimate various cpu's and various overclocks since i know these power draws. had it simply been a single cpu overclocked at temps reported, then i would be left wondering what other cpu's might do, this way i can make a decent guess myself for any cpu. i can appreciate the other opinions but i for one loved the way it was done and look forward to reading other such reviews. the more i learn about every little part of the puzzle, there more i will be able to master this aspect of the pc world as i have other areas. 😀

Very much appreciated, that was intended to be the goal of presenting the test data in this manner.

 

[JackNaylorPE]

BTW I looked at your radiator calculator .... I have some (rad model specific) posted on OCN that are based upon specific fan speeds, thicknesses and actual measured performance of specific models but I let user plug in TDP of components after overclocking. In other words, i didn't calculate the CPU OC as the generally accepted formula seems less accurate on modern CPUs / GPUs. You could use the same formula for GPUs but I think you would find the results well outside reported results.

1. I recognize the formula for CPU but I have found it doesn't stand up in practice using a Kil-o-watt meter. A overclocked 4770k / 4790k usually tops out at about 130 ... I got 177 using the estimator's "generally accepted formula".

2. The results for 140mm rads are exactly the same for Delta T of 5 and 10

3. The 15C numbers should be 2/3 of 10C numbers

4. The 10C numbers should be half the 5C numbers

5. I like that you included flow rate in there as many people wrongly assume flow rate is irrelevant above 0.75C... it's small but not irrelevant.

6. There's no way to account for fan speed and thickness, first being more relevant than the second, tho fan speed is more relevant in thicker rads. What speed and thickness is this based upon ?

7. I see you used 85% for your "fudge factor" in how much of the theoretical heat load must be handled by the rads. I found it to be closer to 60% in testing as the rad shrouds, tubing, water blocks, reservoir themselves also radiate heat. This number will be lower the more components you have.

8. My pump maxes out at 46 watts....dropdown goes to 36

9. Using my components , I came up with 649 watts and it suggests 5.65 x 140mm @ 1.0 gpm. I get about 8.4C Delta T w/

2 x 140 on 60mm rad @ 1200 rpm
3 x 140 on 45mm rad @ 1200 rpm

That's w/ filters removed and fans at full speed.

If you want to collaborate on fine tuning, it would be glad to provide the necessary data.

 

[rubix_1011]

I can make some adjustments, it's good that another set of eyes took a look at it. I put in a lot of work on it over the course of about a day and a half, but really haven't gone back through it to make any adjustments since then.

 

[JackNaylorPE]

I spent weeks on mine .... data collection and typing took the most time. I keep saying I am gonna add the newer rads that came out in the last 2 years but, since martin "retired", using data from 2 different test beds renders data comparisons unreliable.

 

[rubix_1011]

Yeah, I need to take another look at the sheet, so I may do that tomorrow. The majority of my data that is used in the spreadsheet comes from both Martin and Skinnee and then I extrapolated dimensions and results out, averaged them, and used the average as a global value in attempt to represent as broad of an example as possible. I think it was about 3am when I finalized the current technical version, but I think I made a few cosmetic changes a day or so later.

Where do you have your content hosted? I'd be curious to see what you have out there.

 

[4Ryan6]

Hows the progress going in getting a Water Cooling Section like we used to have?

I like the way you conducted your thermal load testing, I am curious as to how long it takes to reach a point of coolant temperature equilibrium with the aquarium heaters from the static ambient that reservoir will revert to when there is no heat load?

Maybe 10,15, 20, minutes?

So I assume you wait for a constant unfluctuating coolant temperature load before recording your results?

I have the opposite results with my TEC cooling setup, as I run a 10c coolant temperature and powering up from ambient it take 30 minutes to drop the coolant temperature down to 10c, so there is a certain amount of time for your coolant temperature to reach it's recordable test temperature?

 

[4Ryan6]

EK Predator and Swiftech H220-X will be done next as I get them.

Both of those run 6w pumps, I could not find any wattage rating for the Corsair pump, do you know what it is rated?

 

[rubix_1011]

I like the way you conducted your thermal load testing, I am curious as to how long it takes to reach a point of coolant temperature equilibrium with the aquarium heaters from the static ambient that reservoir will revert to when there is no heat load?

60 minutes from ambient to first working load. Then, I usually give it 20 minutes between load tests to allow for complete equilibrium to be reached. I then do tests for 20 minute runs. So you can see, it takes quite a while to do even one set of tests.

Both of those run 6w pumps, I could not find any wattage rating for the Corsair pump, do you know what it is rated?

I do not, but it would be pretty simple to figure out using a difference of wattage pulled from my power supply with and without the pump connected.

 

[JackNaylorPE]

Both of those run 6w pumps, I could not find any wattage rating for the Corsair pump, do you know what it is rated?

Martin tested it for the H100i ... it's in a thread on OCN as well as the 0.11 gpm pump flow. Note that in the test, it actually delivered less than half the rated wattage. Nop\ telling whether the H110 GTX uses the same pump, but I think it's a safe bet that wattage, flow and pressure **ratings** can't be trusted. It also describes the significant differences in pump quality between it and the other AIOs tested. Here's some quotes from the review

https://martinsliquidlab.wordpress.com/2013/03/13/corsair-hydro-series-h100i-aio-cpu-cooler/4/

The pump has similarities to the Alphacool DC-LT Ceramic, however the impeller blades are a bit different, the shaft is some sort of metal and the impeller bearing appears to be bronze and the back has four screws instead of two holes. Also the DC-LT is rated at 4.9W where this one measured less than half that.

So, the actual power consumption of just the pump and electronics is 12.10V x .17A = 2.2 Watts.

... then did three of what I call volume/time tests using the reservoir to draw from and a 1 Gallon Jug to fill. I do carefully make sure the inlet level and discharge levels remain constant as to not introduce static head errors. Then as the gallon jug is being filled, I am also adding water to the draw reservoir. I took the best of three tests and that was 9minutes 5seconds to fill 1 gallon or 0.11GPM.

While the flow rate is indeed too small to measure on my normal flow meter, the current on my cheap multimeter checks out ok and I was able to measure maximum pressure head which came it at 1.14PSI. Previously I read on Frosty Tech’s review here that the H100 specs were 112cm H2O (1.59PSI), but I was only able to measure an actual 1.14PSI.

While radiators normally are extremely low in restriction, that is not the case with the H100i. It must have very thin tubes to make it perform well at very low flow rates as the restriction is about 6X more than most radiators I’ve tested

Interesting fan speed / noise comments on page 10 too. It was mentioned in the article that the H100 and h100i used very similar but slightly different pump designs. It's likely that the H100 GTX has a slightly different design also but given the flow measurements, I wouldn't guess it to be very different.

 

[rubix_1011]

I measured the H110i GTX pump at about 0.25 GPM or so - it was just under 1 LPM when I tested 5 runs. I've never seen a pump run so slowly.

 

[JackNaylorPE]

I measured the H110i GTX pump at about 0.25 GPM or so - it was just under 1 LPM when I tested 5 runs. I've never seen a pump run so slowly.

Yes, I read that .... but in when putting some head on it, it went down to 0.18. The 18 inches in height that you measured to get 0.18 gpm is 0.69 psi

Martin measured the radiator restriction at about 0.34 psi and the block at about 2 psi on the H100i or about 3 times what you had when you measured 0.18 gpm..... so I expect that the 0.18 gpm is quite generous when

 

[rubix_1011]

BTW I looked at your radiator calculator .... I have some (rad model specific) posted on OCN that are based upon specific fan speeds, thicknesses and actual measured performance of specific models but I let user plug in TDP of components after overclocking. In other words, i didn't calculate the CPU OC as the generally accepted formula seems less accurate on modern CPUs / GPUs. You could use the same formula for GPUs but I think you would find the results well outside reported results.

1. I recognize the formula for CPU but I have found it doesn't stand up in practice using a Kil-o-watt meter. A overclocked 4770k / 4790k usually tops out at about 130 ... I got 177 using the estimator's "generally accepted formula".

2. The results for 140mm rads are exactly the same for Delta T of 5 and 10

3. The 15C numbers should be 2/3 of 10C numbers

4. The 10C numbers should be half the 5C numbers

5. I like that you included flow rate in there as many people wrongly assume flow rate is irrelevant above 0.75C... it's small but not irrelevant.

6. There's no way to account for fan speed and thickness, first being more relevant than the second, tho fan speed is more relevant in thicker rads. What speed and thickness is this based upon ?

7. I see you used 85% for your "fudge factor" in how much of the theoretical heat load must be handled by the rads. I found it to be closer to 60% in testing as the rad shrouds, tubing, water blocks, reservoir themselves also radiate heat. This number will be lower the more components you have.

8. My pump maxes out at 46 watts....dropdown goes to 36

9. Using my components , I came up with 649 watts and it suggests 5.65 x 140mm @ 1.0 gpm. I get about 8.4C Delta T w/

2 x 140 on 60mm rad @ 1200 rpm
3 x 140 on 45mm rad @ 1200 rpm

That's w/ filters removed and fans at full speed.

If you want to collaborate on fine tuning, it would be glad to provide the necessary data.

I'm more than happy to take feedback - and you're right, I did have a couple mistakes which I corrected in the now-available version.

1. I somewhat agree and disagree with you, here. Yes, I do agree that this estimates high, but it estimates high on an understanding that you're calculating a 100% watt drawn to watt in heat conversion. Not possible, but eh, that's why I try to offset for this in the final calculation with a hefty 85%. I'd rather err on the side of higher in the event someone is building a CPU and flagship GPU loop and need an accurate radiator estimate to fit inside a case.

2. Yeah, can't believe I missed that. Nice catch - should now be corrected.

3. The biggest one I was embarrassed of (but thankful you caught) was the 2/3 (66%) difference in 10C to 15C delta change...I had it at 1/2 (50%)...terrible, terrible basic math on my part.

4. 10C is 50% of the 5C - the delta chart I have assumes (and centers around) a 10C delta of all values, so 5C would have to be x2 of the 10C values, which I accounted for.

5. I wanted to provide the impact of flow on delta. In fact, the entire chart is meant to attempt to represent variations in variables - radiator area, flow rate and thermal load. Fan speed/air flow is the only 'constant' that the chart assumes.

6. 25mm / 1800 RPM

7. Right, somewhat addressed in #1. But yes, dissipation occurs in all areas of the loop, by all components, and as you add more power-drawing components, your average of power draw vs. output in watts also averages down. I had to make a number choice to land to maintain a level field for any possible load scenario, so I chose a little 'high' on 85%.

8. The scale now goes all the way to 11, er, I mean 50.

9. I would say the estimator vs. your calculations actually lend themselves to be very similar. I think my average radiator thickness calculation ended up being somewhere in the vicinity of 38-40mm thickness, so a slightly higher rad area recommendation vs. your current setup with thicker rads w/ slower fans seems fairly close - at least ball park, all things considered.

I also made a couple of formatting issues on the decimal outputs - I decided to opt for a single decimal vs. 2 places on the radiator recommendation (4.78 is far close enough to just be called 4.8). I also formatted the TDP fields to be whole numbers; decimals seemed pointless when you're dealing with hundreds of watts as well as detailing tenths and hundredths of watts...

Again, I appreciate you taking a look and helping point out those mistakes. Sometimes when you look at something so long, you tend to overlook the obvious. Also very nice to get input from someone else on adjustments or suggestions...thank you.

 

[JackNaylorPE]

Been there done that. I started with various formula for calculating OC wattage but users all had their own ideas.

Ya might wanna look at this one for comparison.... set it up (just MoBo and CPU) at stock, hit calculate and write down the number ... the OC the CPU and do again. If y use the formula on GPUs, the results are astronomical.

EDIT: Duh, forgot the link

 

[4Ryan6]

Both of those run 6w pumps, I could not find any wattage rating for the Corsair pump, do you know what it is rated?

Martin tested it for the H100i ... it's in a thread on OCN as well as the 0.11 gpm pump flow. Note that in the test, it actually delivered less than half the rated wattage. Nop\ telling whether the H110 GTX uses the same pump, but I think it's a safe bet that wattage, flow and pressure **ratings** can't be trusted. It also describes the significant differences in pump quality between it and the other AIOs tested. Here's some quotes from the review

https://martinsliquidlab.wordpress.com/2013/03/13/corsair-hydro-series-h100i-aio-cpu-cooler/4/

The pump has similarities to the Alphacool DC-LT Ceramic, however the impeller blades are a bit different, the shaft is some sort of metal and the impeller bearing appears to be bronze and the back has four screws instead of two holes. Also the DC-LT is rated at 4.9W where this one measured less than half that.

So, the actual power consumption of just the pump and electronics is 12.10V x .17A = 2.2 Watts.

... then did three of what I call volume/time tests using the reservoir to draw from and a 1 Gallon Jug to fill. I do carefully make sure the inlet level and discharge levels remain constant as to not introduce static head errors. Then as the gallon jug is being filled, I am also adding water to the draw reservoir. I took the best of three tests and that was 9minutes 5seconds to fill 1 gallon or 0.11GPM.

While the flow rate is indeed too small to measure on my normal flow meter, the current on my cheap multimeter checks out ok and I was able to measure maximum pressure head which came it at 1.14PSI. Previously I read on Frosty Tech’s review here that the H100 specs were 112cm H2O (1.59PSI), but I was only able to measure an actual 1.14PSI.

While radiators normally are extremely low in restriction, that is not the case with the H100i. It must have very thin tubes to make it perform well at very low flow rates as the restriction is about 6X more than most radiators I’ve tested

Interesting fan speed / noise comments on page 10 too. It was mentioned in the article that the H100 and h100i used very similar but slightly different pump designs. It's likely that the H100 GTX has a slightly different design also but given the flow measurements, I wouldn't guess it to be very different.

Thanks for that info Jack, much appreciated!

 

[4Ryan6]

60 minutes from ambient to first working load. Then, I usually give it 20 minutes between load tests to allow for complete equilibrium to be reached. I then do tests for 20 minute runs. So you can see, it takes quite a while to do even one set of tests.

60 minutes, that's longer than I thought it would take to reach equilibrium but you are running aquarium heaters so they're probably designed to heat slowly.

I have to admit I am not familiar with aquarium heaters, so yes I can see it would take quite some time to run the tests.

 

[rubix_1011]

Actually, the aquarium heaters are quite good and work very quickly. I just didn't want to leave anything to chance when first starting up my testing, so I allowed them to run for quite a while in order to ensure working temps. They do very well at heating the water for testing and it's very simple to control their output with the dial on the AC Variable transformer...just dial it up and down; tiny adjustments need to be made in order to get precise loads.

It's a lot like the simple HERMS system I made for our brewing setup - it's a stainless 1/2 keg with a 2000w hot water heater element and a temperature controller connected. However, the PID on the HERMS tank is temp controlled to keep mashing wort at 155 F and the sides are insulated. The aquarium heaters have been over-ridden by soldering a jump on the PCB to always be on. But, both are essentially reservoirs for heat load and conductivity.

 

[nukemaster]

Great write up and very interesting way to deal with repeatable load.

 

[rubix_1011]

Thank you. It was a lot of fun and quite an eye opening experience since I've never worked with a closed loop cooler and so many people use them, yet I've never been able to fully recommend them.

 

[nukemaster]

Always that fear of leaks or pump failure?

I have 2 in this system(cpu and video card. Saves lots of space for me), but it is not primary system as such since it is just for games(and Windows 10 play).

I will admit I keep more of an eye on these than normal heatsink systems since a pump failure will overheat under load very fast while a large heatsink like a NH D14 will run fanless upto a certain point. For me it was about space more than anything.

 

[rubix_1011]

I'm just not a fan of how weak the pumps are, but given they are only designed for a CPU to be cooled, the sizes of these larger radiators alone does help the poor flow. Using high powered fans also makes a big difference, but as the chart shows, they are very loud...and if you slow them down to slower speeds, the cooling potential takes a hit.

But as far as pump failures or leaks, doubtful, but keep an eye on airlocks as they are somewhat common on these units - air bubbles trapped in the cooler causes the pump to stall. Easiest way to fix this is to rotate the radiator with the pump running, tapping it as well.

Also really surprised at the very low volume of liquid contained in the entire unit ....not much there at all. Most radiators will hold a decent amount of coolant, but this one really didn't. 150 or so milliliters total (little over half a cup)...that isn't much at all.

 

[JackNaylorPE]

Ya don't save space, ya just relocate it :0

Any CLC that gives the DH-14 a run for its money will have a 2 x 120mm rad or greater.....and that's bigger than a DH-14.

 

[nukemaster]

You are bang on with that. I have 1 fan outside the case so it saves space inside the case

Since I was not even close to maxing what a D14 can dissipate, I saved space for my system with a "smaller"(it replaces a case fan and the other fan and a shroud are outside[sort of] of the case) H80i.

For video cards it makes a surprising difference(much lower temperatures, but you still need to cool the VRM are) with even a slim 120mm rad.

 

[rubix_1011]

Please address whether the hoses on the Predator can be removed at the "other ends" , aside from the QD fitting.

Got the unit last night. I can confirm that both the CPU block and radiator/pump connections are via normal compression fittings.

240 does not have the disconnect.

 

[4Ryan6]

Thank you. It was a lot of fun and quite an eye opening experience since I've never worked with a closed loop cooler and so many people use them, yet I've never been able to fully recommend them.

That is my hopes from all this testing, having a solid fact base from a reliable source that we actually trust, for recommending these coolers.

 

[Petabyte]

Only 150 ml of coolant wow. Thanks for dissecting that cooler for the greater good. We wouldn't have known the flow rate other wise