Corsair H110i GT installed...temps went UP?!?

[letmepicyou]

My current build is as follows:

CPU: Intel i7 4790k
CPU COOLER: Corsair H110i GT
MOBO: Gigabyte GA-Z79X-UD5H-BK
RAM: 16 GB DDR3-2133 (EVGA)
HD : 1 TB SSD (Samsung EVO)
VIDEO: XFX R9 290X DD
OPTICAL: LG Blu-Ray burner
CASE: AZZA Hurrican 2000

I recently (last few weeks) replaced my Thermalright Silver Arrow (a darn good cpu cooler in it's own right) with the Corsair H110i GT AIO water cooling setup. Now, I know that water coolers are never going to cool below ambient. My issue is, shouldn't the H110i be able to be at least AS cool as my Silver Arrow? After I installed the H110i, my max temps at 100% load went from roughly 73 to 78-ish. I didn't expect massive drops in temperature, but I didn't expect it to jump UP, either. I have the Corsair software set to run the pump at full (Performance mode I think they call it) and the fans are both 100%. No tubing kinks, or anything like that. It's installed at the top of my Hurrican full tower case, below the two 230mm fans at the top. All of the fans are working together. Is this result normal? Should I pull the cpu block off and clean the stock paste and run some of the good stuff I have, and possibly polish the block surface, or is it possible the Silver Arrow was BETTER than the H110i?

Ideas much appreciated.

 

[Ma_Deuce]

I would give your normal thermal paste a shot. I don't ever use what comes stock on the coolers. IMO it's usually too much.

 

[DigDog]

oh derp ..

1. you messed up the install and the paste is not touching properly.
(probably not; and the stuff that comes preinstalled is not that bad)
2. you have a faulty unit

i can't think of any other reasons.

 

[BonzaiDuck]

I look at a lot of reviews when I want to buy a new cooler.

My personal experience includes the TRUE, an older Noctua tower, a Megahalems tower, an NH_D14 and an EVGA ACX cooler -- in addition to a 212-EVO.

I just looked at a review bar-chart someone had posted in another site's forum, showing that the Silver Arrow came close to outperforming my NH-D14 by about 10C with a test-bed thermal wattage and a SB-E 3960 processor. IIRC, those processors had a stock TDP of ~130W. Last year, I concluded that my "modded" D14 came close or matched an H110. But the stock installed D14 could fall short of it by 10C.

So I'd expect something closer to a match between the Silver Arrow and the H110 (if the bar-chart can be trusted). A 5C performance advantage of the TR cooler COULD be attributable to suboptimal airflow or a problem with mounting the waterblock. Then, someone else could comment on the possibility that a pump is not performing as it should. (You could try running the fans and pump at full-bore without motherboard thermal control, to see what the best airflow and water-flow would get you.

And it's always possible that one performance comparison review (only referenced in a web-forum) could deviate from what some reputable testing sources prove.

UPDATE: JUst found a comparison review between the TR SA and the D14, showing them neck-and-neck for clocks up to 4.2 with an i7-950, but the TR cooler seriously underperforming the D14 at 4.4.

So you need to do some troubleshooting, I think.

 

[ClockHound]

UNLESS he has a Silver Arrow IB-E - that will best a D14 or H100. Prolly a H110 as well.

However, the block base could be out of spec - had a H100 that was hilariously concave - RMA'd, as it was worse than a 212 at mild clocks.

 

[MoInSTL]

If you decide to pull the block, pay attention to how tight or loose it was. When you reinstall it, make sure you use a criss-cross pattern like how tire wheels are mounted. Don't over tighten. Make sure any residual thermal paste is completely removed before reapplying. Don't use a paper towel. Use paper coffee filters or a micro fiber cloth. Does your motherboard require a back plate? If so, was it notched correctly?

Edit: Link http://www.corsair.com/en-us/blog/2015/march/how_to_h110i_gt Might be reassuring. Also, is your BIOS set to 100% for the fans? Are the 230mm fans regular case fans?

 

[dlerious]

oh derp ..

1. you messed up the install and the paste is not touching properly.
(probably not; and the stuff that comes preinstalled is not that bad)
2. you have a faulty unit

i can't think of any other reasons.

3. The 230mm case fans at the top are affecting performance?

 

[letmepicyou]

Sorry it took so long to get back to this thread, life happens has it will I guess. Like to thank everyone who replied for their valuable input. Even as long as I've been working on computers, sometimes even I get a head scratcher.

So I finally got around to removing the water block from my i7. Cleaned off all the stock thermal gunk (for lack of a better term). What I expected was a nice shiny surface, and boy was I wrong. I work in a body shop, so I kind of have an idea about varying types of sandpaper grit. The surface looked like it had been DA'ed with some 600 grit at best. The pattern in the surface looked exactly like a DA pattern, and the sand scratches were deep. No wonder they lay such a thick layer of TIM down.

So anyway, I took the Never Dull to it and polished it as best I could in about 1/2 hour's worth of time. Wasn't perfect but was probably 80% closer to perfect than it was before. Got it to a mirror finish albeit with the deeper sand scratches left. Applied the ol' grain of rice dab of Antec Nano Diamond 7 (not the best, not the worst) and put it all back together.

Now at 100% load the temp has dropped to 74c max (though it only hits 74 briefly, usually hovers at 73c). And if I'm not mistaken, ambient temps are up 2 or 3 degrees from last time. So a good polish and re-application of some better quality TIM knocked about 5 degrees off the H110i GT.

Next time I have some free time on my hands, I'll pull it out again and polish it for another half hour and get some even BETTER heatsink compound. I'll let everyone know what it does from there.

All in all, I'm much happier with 73-74c than I was with 78c (or worse). I wasn't having any stability issues, but who wants their cpu running THAT warm. Thanks once again to all who contributed.

 

[BonzaiDuck]

Sorry it took so long to get back to this thread, life happens has it will I guess. Like to thank everyone who replied for their valuable input. Even as long as I've been working on computers, sometimes even I get a head scratcher.

So I finally got around to removing the water block from my i7. Cleaned off all the stock thermal gunk (for lack of a better term). What I expected was a nice shiny surface, and boy was I wrong. I work in a body shop, so I kind of have an idea about varying types of sandpaper grit. The surface looked like it had been DA'ed with some 600 grit at best. The pattern in the surface looked exactly like a DA pattern, and the sand scratches were deep. No wonder they lay such a thick layer of TIM down.

So anyway, I took the Never Dull to it and polished it as best I could in about 1/2 hour's worth of time. Wasn't perfect but was probably 80% closer to perfect than it was before. Got it to a mirror finish albeit with the deeper sand scratches left. Applied the ol' grain of rice dab of Antec Nano Diamond 7 (not the best, not the worst) and put it all back together.

Now at 100% load the temp has dropped to 74c max (though it only hits 74 briefly, usually hovers at 73c). And if I'm not mistaken, ambient temps are up 2 or 3 degrees from last time. So a good polish and re-application of some better quality TIM knocked about 5 degrees off the H110i GT.

Next time I have some free time on my hands, I'll pull it out again and polish it for another half hour and get some even BETTER heatsink compound. I'll let everyone know what it does from there.

All in all, I'm much happier with 73-74c than I was with 78c (or worse). I wasn't having any stability issues, but who wants their cpu running THAT warm. Thanks once again to all who contributed.

Before you polished it further, did you examine the waterblock for flatness with a metal ruler and a strong light?

The "polish" isn't much crucial to anything: you should do fine using 400 grit wet-or-dry to still show some scratches in the light.

 

[Micrornd]

You still have something wrong with your installation.
You have an 84w CPU and 73c @100% load
I have std. H110s on my E5-2696s (120w CPUs) and 55c @100% load with less airflow @ ambient temp of 26c.

Either the block is warped (as the Duck said), the block is not seated right, or the airflow is restricted or fowled up.
But something is just not right if my CPUs push 50% more wattage and run 20c cooler @ 100% load with the same exact radiators.


btw - "fowled up" was an intentional snicker for the Duck :whiste:

 

[BonzaiDuck]

You still have something wrong with your installation.
You have an 84w CPU and 73c @100% load
I have std. H110s on my E5-2696s (120w CPUs) and 55c @100% load with less airflow @ ambient temp of 26c.

Either the block is warped (as the Duck said), the block is not seated right, or the airflow is restricted or fowled up.
But something is just not right if my CPUs push 50% more wattage and run 20c cooler @ 100% load with the same exact radiators.


btw - "fowled up" was an intentional snicker for the Duck :whiste:

It now occurs to me that he should see better than a nominal 72C with the 4790K OC'd to maybe 4.6+Ghz. But it depends on airflow in and out of the case, and airflow through the radiator.

Would there be any chance of a defective pump, perhaps working at some percentage of its expected performance?

 

[MongGrel]

Sounds better.

Maybe lap that baby on a mirror, is on old trick if you don't have a granite surface plate around to lay the sandpaper/polishing cloth on.

I even did that with just paper and diamond paste on a True Copper long ago to mirror finish it. I don't go that far these days, there gets to be a point of diminishing returns.

Just getting it flat helps more.

 

[Micrornd]

It now occurs to me that he should see better than a nominal 72C with the 4790K OC'd to maybe 4.6+Ghz. But it depends on airflow in and out of the case, and airflow through the radiator.

Would there be any chance of a defective pump, perhaps working at some percentage of its expected performance?

I would also question the Corsair software.

As I have posted before, when I asked in the Corsair forum, I was told by a Corsair rep. that on the H110 coolers, the pump should be connected to a constant 12v source with the radiator fans connected to the "CPU fan" header to vary their speed, and that is directly opposite what the directions that come with the H110 indicate as to how it is to be connected.
(ref. http://forum.corsair.com/v3/showthread.php?t=121430 )

In my case, with only a total of 252cfm (push-pull) flowing across 2 - H110 radiators, my full load temps (2 CPUs @120w each @55c full load) would confirm that constant speed on the pump and variable speed on the fans works perfectly, confirming that what the Corsair rep. said was correct.

And note that it is with only 68% of the cfm that is supplied with the H110s - 2 x 94cfm fans each H110.
(The H110i and H110i GTX each come with 2 x 113cfm and 2 x 104cfm, respectively)

Even though the pumps are different between the H110 and the H110i, why would the theory of operation be different ?
Why would you not want the same thing on both units ?

Just to push the software ?
Just to run up the price ?
Who knows ? :whiste:

 

[Rubycon]

I remember those (silver arrow), one of if not the best performing tower cooler if it fits in your case!

AIO units, IME, are convenient that you don't have a huge hulking metal structure suitable for holding small craft in position - in your case!

Performance-wise they're seldom much better. You may or may not have an optimal mount (the factory applied TIM is usually pretty decent) or the base machining may be slightly less than optimal, etc.

The only way you will see significantly lower OC/OV temps is with a custom cooling circuit. Even those will still show loaded core temps of 25-30C higher than your coolant temps so sufficient radiator capacity and decent ambient air temps are necessary.

With AIO tests, manufacturer's often show comparison to the factory supplied heatsink so they will always be far better when OC. Some chips (Devil's Canyon) can barely pass on factory heatsinks stock let alone with any sort of OC!

In a system I recently worked on (5960X, 1.25V, 4.3GHz) with a Kraken X61 which is one of the best AIO it was hitting around 80C with AVS Prime95. Switching to a custom circuit based on the new EK EVO block (a Titan X is also in the loop) loaded temps dropped to the mid to upper 50s. Same ambient. System has two 360 rads and a MCP35X PWM pump. (speed varies from 1400RPM to 4400RPM but stays under 2400RPM which is nearly inaudible) Coolant temp averages 26C. GPU temps (from stock cooling) went from mid 80s to lower 30s. That was running Heaven benchmark and the Titan X was OC to 1.4GHz GPU and 3.8 memory (no voltage tweaking!).

WC makes the biggest difference on GPUs and best of all there is no hair dryer noise when loaded and the back of the case where the display port cables are doesn't feel like a freaking furnace! :biggrin:

 

[doyll49]

Silver Arrow (all varients) are among the best coolers made. There are minor differences in cooling between original Silver Arrow and newer SB-E & IB-E variants Original cooled slightly better at low rpm and not quite as good at full speed. I did not find any difference between SB-E and IB-E. Extreme versions have same cooling and noise figures at same rpm as normal versions, but latch the hatches before running them on up to their 2500rpm limits. The most huge amounts of air and are also quite loud at full speed.

NH-D14 is similar as is NH-D15. Main Reason NH-D15 cools a little better is it's fans run 2-300rpm faster. s for the Silver Arrow fans, do a side by side comparison of NH-A14 and TY-140, 143 and 147 fans. They are near identical except for color and housing shape. Performance is also near identical in pressure, cfm and noise levels when running at same rpm.

But to the issue of H110 vs Silver Arrow, cooling is similar but H110 is louder at same temps

Reason review tests vary so radically is most use room ambient for baseline temerature instead of the cooler / radiator intake air temp. While room temp has an effect on cooler intake temp, cooler intake has many other variable affecting it and is always warmer than room .. and can easily be 10, 15 or even 20c warmer under full load. Of course the CPU will also run similar amounts hotter, but the room ambient will not change. End results are we see this huge 20c variance in results between reviews even when using same CPU & mobo.

Honestly, using room ambient, especially when testing in a case, is like going into the kitchen to look at thermometer to know how warm your lounge is. Which bring me to all this hype about "real world enviroment" when testing in a case and using room ambient. They are not testing cooler performance. They are testing their systems' performance with different coolers. And the only "real world environment" is their system in their room .. unless you happen to have the exact same system in the same room. I know I don't.

 

[Micrornd]

Reason review tests vary so radically is most use room ambient for baseline temerature instead of the cooler / radiator intake air temp. While room temp has an effect on cooler intake temp, cooler intake has many other variable affecting it and is always warmer than room .. and can easily be 10, 15 or even 20c warmer under full load. Of course the CPU will also run similar amounts hotter, but the room ambient will not change. End results are we see this huge 20c variance in results between reviews even when using same CPU & mobo.

Honestly, using room ambient, especially when testing in a case, is like going into the kitchen to look at thermometer to know how warm your lounge is. Which bring me to all this hype about "real world enviroment" when testing in a case and using room ambient. They are not testing cooler performance. They are testing their systems' performance with different coolers. And the only "real world environment" is their system in their room .. unless you happen to have the exact same system in the same room. I know I don't.

I understand why that makes sense from your point of view and although I agree to a point, I would say that may be a little too narrow.

In my particular case, room ambient temp is the same as cooler intake temps, as my rads are the sole intake for the case.
In my scenario how the air flows through and exits the case is the main variable that affects the cooling.
How the air flows through and exits the case also is a great factor in your scenarios, as well as what other heat producing sources are in the case.

I've seen "open air" testing for coolers proposed also, but to me, that's a much worse solution than testing in a case, as it has way too many uncontrolled variables, as to airflow restrictions and deviations from what would be a "norm" in a case. (i.e. cables, other cards, other heat sources, other case fans, etc.)

Then you also have to factor in how the cooler is made/designed and mounted compared to a OEM cooler (what direction the air flows through compared to the OEM cooler i.e. up, down, left, right, etc.).
Also as to how it deflects air to the MB and VRMs for cooling them (as almost all OEM coolers are designed to do).
That seems to never be discussed or compensated for with "non-OEM" cooler testing.

There are just so many variables that factor in beyond inlet temp, that there is no really good way to test coolers.

So maybe just being aware of all the variables that enter into the cooling scenario is the most important factor of all, rather than just out of context temp and noise figures.

What do you think?

 

[doyll49]

@Micrornd
I started new thread to discuss cooler testing criteria with both of our posts and my reply.