The Corsair User Forums

The Corsair User Forums (
-   Cooling (
-   -   H115i and 2700X - CPU Temp high, Liquid Temp Low (

ChromeD2 06-16-2018 02:27 PM

H115i and 2700X (Contact Issue?)
Hey everyone!

A few days I finally upgraded my rig from AM3 (FX-8350 @ 4.7GHz, never went past 56C in roughly 1yr of H115i goodness) to AM4 (2700X).

I tested things out with the Stock Cooler while waiting for the AM4 Bracket for the H115i to arrive and temps seemed high (using Ryzen Master temps as reference here as they seem to be what most people advise using, up for any suggestions though).

Optimized Defaults loaded on BIOS with MSI's Core Boost enabled went up to 4150, throttled down to 4000 or 3900 after a while on air. Not bad to begin with so I was stoked to try it out properly.

Bracket arrived, I installed the H115i and.. temps were just a couple of degrees lower if that.. and the Core Boost OC was about the same.
I tried reseating the cooler as maybe I missed something or applied the Thermal Paste wrong or w/e.. but nope. Same result.

I tried several manual and automatic OC setups on the BIOS and the results were not encouraging. Any voltage above 1.42v on the CPU means I'm looking at a potential thermal shutdown (gets to 100c fairly fast).
I've since changed the cooler's orientation twice, retested the Stock Cooler (again, only marginally worse than the H115i) and tried both "loose'ish" and tight fits with the screws.

The Liquid temp (measured by AIDA64, iCue and Link on separate occasions) rarely goes past 32C and doesn't seem to really rise much, if at all, under load.
I've tried different pump/fan speeds as well, with no significant difference in the results and the air coming off from the fans doesn't really feel very warm either.

I'm running out of ideas here, anyone got any suggestions?


c-attack 06-16-2018 03:05 PM

Type 2700x into the search box and see how your fellow AMD owners are doing. Corsair has yet to make a Link/iCUE update that can accurately read the 2600/2700 and everyone is getting absurd temps in the meantime. The liquid temp is accurate. Package temp is not. Depending on how much other Link dependent stuff you have, it’s probably best to fully exit Link or iCUE and use another program until this is sorted or you become comfortable with the cpu’s normal temperature behavior.

ChromeD2 06-16-2018 03:37 PM

Thanks for the suggestion, mate. Already had a look but will look and read a bit further, can't hurt.

Should iCue/Link's inaccuracy matter much when I'm using Ryzen Master to read CPU temp and have fans/pump set to max though? Is there any hidden automated curve that I'm missing?

c-attack 06-16-2018 03:42 PM

The fans should react to coolant temp, so it is more of a worrying nuisance than an actual control issue. Most people are reporting idle temps in the 80-100C range, which is probably enough to stop your heart on a brand new system. Make sure you leave fan curves set to the H115i Temp Group. Aside from that, it doesn’t matter. The one final catch is Link and iCUE don’t always get along with other monitoring programs. If using Link, the custom fan curve for the cooler will stick, so you can quit and use other things if needed.

ChromeD2 06-16-2018 04:23 PM

Most 2700X threads seem to focus on inaccurate readings from iCue/Link and things related to those being off.. or to which temp group the fans are assigned to. I just set it all to max instead, no curve, so I can test things thoroughly.

My issue is more that with the H115i on full blast I get a couple of degrees under the stock cooler, measured with AMD's own Ryzen Master under full load.
It's accurate enough that I now had one thermal shutdown while testing (it went over 105c).

Now.. I know the Wraith Prism isn't quite as bad as stock coolers used to be, but if the H115i can only shave 2-3C off during full load by comparison.. something could be off?

The liquid temp not going up much also makes me somewhat suspicious as I'd expect it to rise far more with heavier loads than it currently does and the air coming from the fans to be significantly hotter (it was the case with my 8350).

All this being said.. I'm still out of ideas. Going to run a full load test with fans set to 40% and pump to extreme to see if the liquid warms up any more than with the fans on max.

Again, thanks for the replies, mate. Anything helps at this point.

c-attack 06-16-2018 04:32 PM

Wait, so you are saying the package temp is 80C+ on the AMD software?

I would expect a coolant rise of +6-8C for a 100% CPU only load. Probably go up for 10 minutes, then level off.

What is the max difference between coolant temp and real package temp when at load? (e.g., 32C coolant/85C package)

ChromeD2 06-16-2018 05:29 PM

That's correct. On Link/iCue I disregarded temp readings entirely as they were 90+C idle and a very cozy 130-140C under load (maybe more, even).

I've now with tested the fans on fixed 40%, pump on extreme. The highest Liquid temp (as measured by Link) was 32.6C. Usually it sits between 29.8 and 31.5. AMD's tool (Ryzen Master) reports package temps of around 36-37 idle, around 85C on full load (spikes up to 90C sometimes though).

The 80+C temps were with the settings on Auto, which downclocked me to 3.8-3.9GHz and downvolted me as low as 1.25.

So max difference between coolant and real package is around 54C full load (31C coolant/85C package).. and the coolant never really rises more than 2.5C.

DevBiker 06-16-2018 06:00 PM

That sounds like a possible contact issue. What did you use for paste? Can you double check the mount and make sure it's tight?

Note: I messed up my mount once; one of the stand offs wasn't tight. Saw the same kinds of symptoms.

ChromeD2 06-16-2018 06:26 PM

Good old Arctic MX-4 (brand new tube). I hand-tightened the 4 standoffs the first time around. 2nd time around I did that and gave it a little extra very gentle squeeze with some pliers. I also fiddled around with the mounting pressure on the thumbscrews up top, first just the hands, then a little extra with a screwdriver.

I have mounted the cooler 4 times now (different paste amounts and shapes too) because I wanted to make sure it wasn't a contact issue. I'm still not entirely convinced, though.. I'll redo the whole thing at least once more.. still got half a tube of MX4 left to play with.

To be 100% accurate here.. I'm using the H115i ( CW-9060027-WW ) with the AM4 Bracket ( CW-8960046 ).

c-attack 06-16-2018 06:41 PM

I am not familiar enough with the new 2700x to be sure of the expected max package, but that 50+ delta is pushing the normal limits. The 2.5C rise in coolant temp is certainly too low. Both of those do suggest the heat is not transferring out as it should. Not disastrous enough to cause terrible idle and meltdown when you open the browser, but not quite good enough either. Unfortunately, I just don't use AMD CPUs and have zero experience with that bracket to know the normal pitfalls. We need someone who has been using it.

If you don't get some more responses in the next 24 hours or so, re-post this as a new thread and a title that reads something like "AMD 2700X contract issue" or "AMD 2700x bracket help". The erroneous temp readings on the new CPUs have been a constant posting topic and I am afraid many people might overlook this as an already seen & read issue, based on the title alone. My mind was already leaping there as soon as I saw it.

ChromeD2 06-16-2018 06:47 PM

Edited the Title for now, will repost if needed (don't want to litter the forum with threads, etc :p). Edit: Just noticed it doesn't change the topic's title on the subforum. Oh well.. I'll repost later if needed.

I'll finish a couple things here then go and redo the whole mounting process back to back.
Really hoping I messed up something nice and obvious and that it isn't just something like the backplate that comes with this motherboard.. or anything else that's similarly annoying.

ChromeD2 06-16-2018 10:32 PM

Alrighty.. tested a whole bunch more in the last few hours.

99.9% sure it's a contact issue, can't tell why it happens, though. I was 90% sure going in, the 9.9 extra % came from the fact that in the majority of attempts, I verified the thermal paste covered the edges but the middle often had very little to none. This is regardless of amount of paste as smaller amounts meant the center would have almost no paste and heat would be much worse.

I've tried changing orientations, mounting pressure, all thermal application patterns I could think of (smiley face not included, but I was tempted).. and results were about the same.
Liquid temperature is always below 33C, even when running blender benchmarks and keeping package at a toasty 88-91C for a few minutes.

I'll start a new thread with a clear title and a summarised version of what we have so far.

Thanks again for the suggestions so far, guys.

c-attack 06-16-2018 10:57 PM


Originally Posted by ChromeD2 (Post 959393)

I'll start a new thread with a clear title and a summarised version of what we have so far.

Thanks again for the suggestions so far, guys.

You still can if you want, but now that top has changed and the issue narrowed, it may not be necessary. Besides, even if they think it’s the erroneous temp issue, a lot of 2700x owners may pop in to add the typical “me too” and those are the people who you want to read this - ones using the bracket.

ChromeD2 07-20-2018 12:16 AM

Solved: After much testing I RMA'd the H115i unit. Replacement came in today and.. it's no longer happening.

I'm fairly sure it was indeed a contact issue due to a deformity, very likely caused by cooling the overclocked 8350 for 1yr'ish.

Hope no one has similar issues but if you do.. RMA is the way to go.

zealelf 09-05-2018 10:54 PM

Same Problem
I had to RMA my h115I and now after 6 months its doing it again, i think remounting it warps it . I recommend using the graphite cooling pad so thermal past dosent need to ever be replace. because when you do it will warp again. 3rd rma her i come. also this was the fix for my wife's comp that is the same build as mine she hasten had any issues.

Thermal Pad

All times are GMT -4. The time now is 11:47 AM.

Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2019, vBulletin Solutions, Inc.