Keep Corsair H100i V2 or get Noctua NH-D15

[zarlaan]

Hello,

 

I've had my system for about 1.5yrs and I swear that I had better idle temp at 28-30c previously, but now my idle is around 35c. My real concern however are my temps during gaming. Recently there was a BIOS update for the Spectre & Meltdown, and just yesterday I thoroughly cleaned all dust from my system which there was hardly any as I clean it regularly.

 

My temps during games such as GTA5 will range from 65c to 80c according to the Link graph. When running the ROG Realbench stress test for 15min with 8GB RAM my CPU temp is hitting 95c and water temp close to 50c.

 

Are these temps typical for the H100i V2? I have two Noctua NF-F12 fans instead of the stock Corsair fans and I also has two Noctua NF-A14 in the front of the case for intake. I'm debating swapping the H100i out for a Noctua NH-D15.

 

CPU: I7 6700K @ 4.2Ghz (Stock)

MB: Asrock Fatal1ty Z170 Gaming ITX/ac

GPU: MSI GTX 1070 Seahawk X w/ Noctua NF-12

Case: Phantek Enthoo Evolv ITX

RAM:16GB G.Skill Ripjaw V 3200Mhz

 

First screenshot is from idle.

 

https://ibb.co/b2sJby

 

Now after 15min randomly driving around in GTA5 and running its benchmark twice. I think the CPU temp was slightly higher than it would be while in the game due to tabbing to the desktop. You can see that my case temp from the "Temp Sensors" is only at 34.5c.

 

https://ibb.co/fnM0pJ

 

My system: https://ibb.co/gVuEGy

[c-attack]

Hello,

 

When running the ROG Realbench stress test for 15min with 8GB RAM my CPU temp is hitting 95c and water temp close to 50c.

 

Are these temps typical for the H100i V2?

 

We need the coolant temp starting point to put this into context. You can't be any lower than the room/case temp and it seems like that is in the 32-34C range. That is not uncommon for a small case in Summertime. That said, a +16C rise in coolant temp during RealBench is a lot, especially with both the GPU and CPU exhausting waste heat. You might be able to rein in CPU temps a little with some BIOS tuning, but the coolant temp rise is way off the mark.

 

Can you run a different CPU only stress test? Use something mild like in Intel XTU or AIDA. Just not Prime95 or anything else that is going to cook the CPU regardless of cooling capability. The same 10-15 min duration is fine. Coolant temp should go up 2-3C in the first minute they slowly rise at maybe +1C/min for 3-5 min. Rises after that tend to be case related.

 

I can't see under the top fans, but hopefully they are set to blow air up and out as exhaust.

[zarlaan]

We need the coolant temp starting point to put this into context. You can't be any lower than the room/case temp and it seems like that is in the 32-34C range.

 

My room temp is 23c and I have a thermostat sitting on my desk. The H100iV2 temp (water) can be seen my screenshots at idle of 31.7c. I'm using a commander pro with one of the temp sensors to get the case temp which can also be seen in the images at idle 28.7c. The sensor for case temp is located hanging just below the radiator fans which are used as exhaust.

 

The Intel XTU would not run, but I ran the AIDA stability test for about 5min and had to stop it as my cpu temp was hitting 100c.

 

Before AIDA: https://ibb.co/cTVPzJ

 

5min running AIDA: https://ibb.co/bMqDtd

[c-attack]

Well, at least one issue then. Nearly +10C coolant temp rise in 5 min on a AIDA blend test is too much. Usually this suggests some kind of building restriction inside the cooler preventing the liquid from making the expected number of cycles. You may also be able to see this if it takes a really long time for the coolant to drop back down post test. It should drop 3C within the first minute, then slowly start dropping about 1C/min back to just above the original start temp. Another way to demonstrate this is kicking the pump into the higher speed, either at idle or load. It really should make little difference, but if you turn the pump to high in the middle of a test or on the post test cool down and then see the coolant temp take an instant dive, that probably means there is too much resistance.

 

However, there appears to be a secondary issue. Even if the coolant delta had been the expected +5C or so on the blend test, your end core temp to coolant temp delta is enormous (about +60C). This is often do to high voltage, but it looks like you only peak out around 1.28v at 100%. Look in the Link graphing are for Vcore to confirm. At 1.28v, you should not see a +60C coolant/core differential. That is more like 1.40v+. So, if the voltage is reasonable and hasn't gone crazy on the default auto settings, then you probably have a contact problem between the cold plate and CPU. Did you move anything around recently or re-apply new TIM?

 

The problem with this is if you did have poor contact, you would not expect to see high coolant temps. However, if both are true, one can mask the other. I think checking the voltage peaks and checking the liquid cool down rate are the least intrusive ways to sort this out.

[zarlaan]

The Vcore doesn't seem to be an issue as I never see it rise above 1.25-1.28.

 

As of right now my CPU is 36c, water temp 35c, and a case temp of 31c. This is about 15-20min after stopping the AIDA stability test just idle. Something just isn't right.

 

In the last 1.5yrs I've had this setup I have never removed to the heatsink. I slide my computer out every few weeks to open it up and blow out any dust and that's the only movement of my system. The paste is still the same stock past that was pre-applied to the H100i. I did order some Thermal Grizzly Kryonaut thinking maybe that stock paste has somehow failed. I won't get that until Tuesday.

[c-attack]

The cooler behavior is suspect, but short of the clearly disabled point that makes all other metrics meaningless. That means you might want some more clear evidence before submitting the RMA claim. I am concerned they will get hung up on the core to coolant delta and claim contact issue. That does seem to be a problem. At 1.5 years I would generally say its time for a TIM redo, but you don't normally see performance degrade to such poor levels. Perhaps give it a rest for tonight. The problem with testing in this state is if there is a flow problem, each result gets worse and worse and you lose touch with a meaningful baseline. After a few hours of PC nap time (sleep or shutdown), you can try running a short 5-10 min test again. Also try switching the pump to Performance mid test or post test and see if the H100i v2 temp takes a quick drop or you hear the "whoosh" at quiet idle.

[zarlaan]

Update...

 

I removed the H100i v2 and cleaned away any trace of the stock Corsair thermal compound using 91% isopropyl alcohol. Afterwards I applied small amount of Thermal Grizzly Kryonaut. Waited 24hrs and then ran AIDA64. My CPU package idle temps were still at 35c with a coolant temp of 29c and a system case temp of 27c. This of course is not bad, but its still higher than what I was getting a few months ago.

 

However, my max CPU package temp after running AIDA64 for 10min had a marked improvement maxing out at 82c with the Corsair H100i v2 and fresh TIM applied.

 

Results after applying fresh TIM to the Corsair H100i v2: https://ibb.co/gOVigK

 

Today, I have replaced the H100i v2 with a Noctua NH-D15 and now my CPU idle is 26c. Running AIDA64 for 10min my max CPU package temp only hits 67c. On top of this both of the CPU fans are only set to run at 600rpm.

 

Results with Noctua NH-D15: https://ibb.co/n4pjMK

 

System with Noctua installed: https://ibb.co/m1D8Fe

[c-attack]

Every time I don't copy a long message before posting, this forum glitches and its lost. Here is condensed try #2.

 

I am glad it is working well for you. However, if I gave you a 720mm radiator and custom water cooling set-up, we couldn't take 15C off your temp or anyone else's. That is probably a combination of two things. 1) Slipping performance on the H100i (looks like about 3-4C); 2) Difference in room temperature. We discussed above that perhaps your cooling was deteriorating to some degree. In your screenshots, the DH-15 set up shows a substantial reduction in case temperature compared to prior h100i shots. It looks like about 8C less on both the T-sensor and the DRAM data. I am not sure if that is a reduction in room temp when you were testing or a result of the direct blow through effect of the air tower and removal of the 240mm radiator. The 8+4C at least makes sense to explain some of the differential, as opposed to contact problem or something along those lines. It would also explain the idle temps, which aside from CPU specific voltage and power saving features, is most impacted by room/case temp. Cooler is certainly better, but I might be curious if it is the physical set-up or just a cool day in the house. Hopefully this will hold up for gaming or whatever else you like to do. The necessary relocation of the GPU radiator to front intake has me a little concerned. You may see exhaust temps off of that in 45-50C range, depending on the load/game.

[zarlaan]

The room temp has not changed and is a consistent 23.5c. I have a thermometer hygrometer sitting on my desk which is very accurate. Plus my desk is in main hallway and not a closed room.

 

I think this can be explained by two things.

 

1) Declining performance of the H100i as we have both stated.

2) Poor top exhaust in the Evolve ITX case.

 

I had noticed during some testing that increasing the fan speed on the radiator which was mounted to the top as push/exhaust would also slightly increase the case temp by a couple degree celsius even at an idle cool case temp, e.g. case temp of 28c would increase to 30c.

 

This indicated to me that the design of the Evolve ITX has a slight flaw in that the roof is poorly ventilated, and some of the exhaust from the radiator will actually end up getting pushed back down into the case. I think you would need to modify the Phantek radiator mounting bracket by completely sealing it up to prevent this. I was very close to switching out my case to the Define Nano S for just this reason. I suspect the results would be better if the H100i was mounted as an intake at the front of the case, however I have a 3.5 HD so I cannot try that configuration with the H100i. Again, another reason I was tempted to switch to the Define Nano S.

 

I was concerned at first as well about mounting the GPU radiator as a front intake, and in fact was going to mount it to the top as an exhaust. However, with the push/pull config the core GPU temp never exceeded 47c after 30min of playing GTA5. The case temp only increased to 34c, and the CPU temp maxed at 54c. This is also with all my fans running in the "Quiet" profile mode from the Corsair Link using a Commander Pro for the exhaust, GPU and 2 front intake fans. The CPU fans are set in the BIOS to "Silent" which puts the fans at 600rpm until it gets to 75c. At that point the CPU fan will ramp up, but I have not encountered that scenario yet :)

[c-attack]

The room temp has not changed and is a consistent 23.5c. I have a thermometer hygrometer sitting on my desk which is very accurate. Plus my desk is in main hallway and not a closed room.

 

 

I was concerned at first as well about mounting the GPU radiator as a front intake, and in fact was going to mount it to the top as an exhaust. However, with the push/pull config the core GPU temp never exceeded 47c after 30min of playing GTA5. The case temp only increased to 34c, and the CPU temp maxed at 54c. This is also with all my fans running in the "Quiet" profile mode from the Corsair Link using a Commander Pro for the exhaust, GPU and 2 front intake fans. The CPU fans are set in the BIOS to "Silent" which puts the fans at 600rpm until it gets to 75c. At that point the CPU fan will ramp up, but I have not encountered that scenario yet :)

 

 

OK, then a positive end to this. I generally prefer water cooling to air for multiple reasons, but there are absolutely some scenarios in which I would want an air tower for total system performance, not just a straight up CPU bench results. The only real knock is the big air towers can be exceedingly difficult/tedious to install. Also, there are still a lot of cases out there that may not be completely water cooling friendly. I usually have concerns about the solid panel "noise baffle" type cases and we have seen people run into problem with a heat layer trapped above the radiator keeping it artificially warm when the exhaust can't escape.

 

Since it looks like the case temp drop is genuine, that should offset any additional heat coming off the GPU radiator. I don't think the GPU will be affected, but when it is 47C then the exhaust temp is going to 45-47C as well and that could be passed onto the CPU. However, since your actual usage looks good, that does not seem to be a problem and you appear to be in a better state of affairs. Enjoy.

[zarlaan]

Thanks c-attack. Your assistance was greatly appreciated.