Jump to content
Corsair Community

H110i Pump Failed, Upgraded to H115i, Still High CPU Temp


vitalsine
 Share

Recommended Posts

Hey guys, it's been a pretty bad couple of nights. Last night I had my pump fail on my H110i, I was rendering a video in Adobe Media Encoder when it happened, I had no idea, and all of a sudden I experienced a thermal shutdown. Immediately following I could not boot, it would just shutdown as soon as I hit del to enter BIOS. After a few hours of troubleshooting I established that it was the pump in the H110i that failed, it was about 4 years old.

 

While it is still under warranty, I work from home on this PC and I had to get back up and running right away. I ordered a H115i PRO off Amazon and was able to get free next day delivery, thank god.

 

After replacing the H110i with the H115i, I immediately entered bios, and reset my clock speeds to stock. I was able to get into Windows, and initially temps were fine on that first boot, around 28-30C idle on the CPU Package temp, the cores were registering around 23-25C. I ran AIDA64 Extreme and was able to run it for about an hour without any issues, topping out around 80C.

 

So, I rebooted the system and put my clocks back to where they were, a pretty modest overclock of 4.0GHz, 1.3V. When I booted back into windows my temps were a lot higher... idling around 60C Package temp and 40C on the Cores. I reverted back to stock again, and the temps are still 60C+, I can only run AIDA64 stress test for about 1 minute before temps are over 80C.

 

2syRXLG.jpg

 

With all that being said, I have confirmed the pump is running, I have it set to balanced in iCUE so it is at about 2070RPM. Fans were running around 1500-1800RPM, but after tuning the fans in BIOS it seems to have slowed them down, but I've tested them at 100% and it's not affecting the temps at all. Oddly enough, when I push down on the Pump Unit, temps start to fall pretty quickly. My screws are as tight as they'll get though, so I'm not sure how to get it to seat better? I have removed the Pump Unit and reattached it a couple times with no positive or negative results. I have some extra Arctic MX-4 that I could use if I have to reapply thermal paste, but I don't see why I should have to, when removing the Pump and checking the CPU, you can see that it is making flush contact with the CPU.

 

I guess what I'm wondering is, could the CPU have been damaged from the thermal shutdown last night? I would assume it just wouldn't boot if it got fried. On top of all this, my RAID array crapped out on me, 2 drives failed because of the thermal shutdown and I may have lost a decent amount of work. Losing it over here, please help. lol.

Link to comment
Share on other sites

The Pro cooler is working (at least from its perspective). If there were any kind of internal pump, flow, or related issue, the H115i Pro Temp (coolant temp) would just tick upwards every few seconds taking the CPU temp along with it. Slow, steady rise. 29C is a pretty normal idle coolant for anyone in a 23C or so room. *FYI, this model has an ultra low speed Quiet pump (1100 rpm). Make sure you don't use that for rendering. It will produce a very clear and dramatic temp penalty. Most people will have identical results on Balanced and Extreme.

 

You are running iCUE and Link. They are overlapping and you are essentially running dual/triple polling tools. Get rid of one. Generally speaking, we tell people Link has been dead for 2 years, time to move on. However, given you only have the cooler and the RAM, you don't really need the advanced lighting capabilities of iCUE and the lighter resource using Link (it's literally half the program) may make more sense for your purposes. I would uninstall iCUE. H115i Pro control will return to Link.

 

Motherboard fan tuning should have no effect on the H115i Pro. It is SATA powered and the motherboard connector reports a pump speed to the BIOS and will get you past the CPU fan boot error when on that header. It's fans or pump cannot react to BIOS fan control changes. The wire to the motherboard lacks the required pins.

 

Now with all that out of the way, we may not be any closer to troubleshooting the original event. However, it very well may be just that -- an event. I think the best course is to get rid of iCUE, understand you can't run HWinFO at the same time without excluding Corsair and Asetek sensor data, and reboot. If you need to run HWiNFO, make sure you use the portable version that does not have the persistently deployed driver. Reboot. Keep an eye on things. See if there is a occurrence. If CPU temps spike, check load % and Vcore to see if they match. A cooler level issue does not require either and will behave differently.

Link to comment
Share on other sites

Thanks for the reply c-attack, I uninstalled iCUE so that I could gain control over the H115i fans again, I have them running around 1100RPM, set to 100% in Link. I left the PC idling all night, well I left some data recovery software running to see if I can salvage anything from my broken RAID array, and it's still sitting at 55C, pump temp is 24.3C. I'm not using HwMonitor either, just LINK at this point. I don't understand what the issue is here. Using logic it seems like the pump head isn't sitting on the CPU tight enough, otherwise wouldn't the pump liquid be a higher temp?

 

vE3eKaL.jpg

 

Maybe I should mention that when I installed the H115i, I left the backplate from the H110i on the mobo, as it looks to be identical, same with the standoff screws, they appear to be identical. Maybe this is part of the issue? I could try unmounting the unit, changing those things out, reapply some MX4, and see what happens... It just doesn't make sense to me that when I first booted with this thing it was running safe temps, now I can't get it under 50C... even in BIOS it is showing a temp of 44C. Room temp is around 60F.

Link to comment
Share on other sites

...and it's still sitting at 55C, pump temp is 24.3C. I'm not using HwMonitor either, just LINK at this point. I don't understand what the issue is here. Using logic it seems like the pump head isn't sitting on the CPU tight enough, otherwise wouldn't the pump liquid be a higher temp?

 

Maybe I should mention that when I installed the H115i, I left the backplate from the H110i on the mobo,

 

Oh, you are sitting at an elevated temp, not sporadic and expected spikes. Yes, the backplate might be something to look at. The H110i (GT) is a CoolIT OEM part and the H115i Pro is partnered with Asetek. They will have different brackets and I would not expect them to be interchangeable. Unfortunately, millimeters matter. That seems like the best place to start and a constantly elevated CPU temp, with minimal load %, and no coolant change usually spells poor physical contact.

Link to comment
Share on other sites

So, after getting in there, I realized I don't even need a backplate... LGA2011 doesn't require one, and there is no way to attach it anyway. I did swap the standoff screws (no difference whatsoever) and the thumb screws (which were slightly different). I also cleaned the old TIM off and reapplied some MX4. I was able to confirm that it is sitting flush on the CPU, the spread looks nice and shows that there is contact between the pump head and the CPU (I've been building PCs for about 20 years and never really had issues with this). Still running hot, 48C, about 6C less than before but still very high for idle... before the H110i failed on me I was idling around 23-25C, so we're still looking at over a 20C difference. As I am sitting here typing this, the CPU temp has risen to 50C.

 

RRcuRH4.jpg

 

EDIT: Thought I should mention, when I removed the TIM from the H115i, it seemed to be a little hard, not sure if that makes sense... could have something to do with the cold in the Northeast? I remember reading something about this somewhere, that the cold can affect the paste.

 

EDIT 2: So I am running AIDA64 Extreme stability test, it is showing a max CPU temp of 49C under 100% load, which is WAY off from Link or any other monitoring software I have checked... weird.

 

EDIT 3: Here are is a snapshot of AIDA and Link next to eachother...

 

atHB4hV.jpg

Edited by vitalsine
Link to comment
Share on other sites

Ugh... Sorry. I am a not myself today. I am seeing 6850x on the Link screen shot, but it didn't ring the right bell. Yes, no extra backplates for you and it seems it was just the wrong stand-offs.

 

I can't speak specifically about the 6850x, however my X99 Asus boards never had AIDA "CPU temp" match Link data. AIDA package on that series was really socket temp, so it would read about +12C at idle when cores were down, but socket on. Then at load they would come together. I don't know if Link is averaging or taking a different sensor point. AIDA stress tests are also dynamic, so it's possible they are polling at different times. One way to get a better handle on it is to load up something like CPU-Z and go to the bench tab. Run that little app's Stress Test. It is a mild and linear Linpack. It should load at 53C (or whatever) and essentially stick there, only increasing by +1C for each +1C of coolant rise in Link. That might make it easier to sort the various temp sensors. They should all hold steady.

Link to comment
Share on other sites

Thank you so much for your help thus far man. I really appreciate it, truly.

 

So I tried running the CPU-Z stress test and had no issues after about 40 minutes. Oddly enough my CPU Package temp did not change one bit... its been idling at 48-55C according to Link, and even when running the stress test for 40 minutes it didn't go over 55C. I am so confused. Is there a chance the diodes or whatever is measuring my temps were damaged when I experienced the thermal shutdown? That wouldn't really make sense though, as like I said the first time I booted with the H115i my temps were great... after that though they've been stuck up pretty high. Im hesitant to try and do anything with this cause I don't want to damage the CPU, but I feel like I'm running out of options as to how to fix this.

Link to comment
Share on other sites

I think damage is unlikely. The entire point of the safety shutdown it to make it at a point below the "chance of damage" threshold. The limits posted on Intel should be the safety limits, not the actual damage point. Running to 99% of the actual limit is never safe.

 

Your core temps are down and exactly what I would expect. My best guess the Link CPU package is really a socket temp sensor (on Broadwell and Haswell-E). You can always get a third opinion. Try HWiNFO. It has an absurd amount of CPU information. A few caveats.

 

1) Download the portable version that does not use the persistent driver

2) Before opening for the first time, quit Link and kill the Corsair Service in the task manager.

3) You may get a prompt about conflicting Corsair monitoring anyway. You now need to decide if you want to use it along side Link or instead of it. If simulatanous use, then you must disable monitoring of Corsair hardware and tick off the Asetek safety sensor in the settings. That is the cooler. If instead of... run normally without Link or the Corsair service. It should show you every CPU temp sensor.

4) Don't run AIDA and HWiFO at the same time. Probably doesn't matter, but might lead to excessive polling.

Link to comment
Share on other sites

So I ran CPUID stress test again with HWiNFO open to monitor temps. Same results.

 

zw5kX91.jpg

 

I am going to just live with the idle temp I guess. Max temp never goes over 60C at full load, so I guess that works, right? Wish I could figure out why the temps are higher at idle though. Even 30C on the cores seems high to me, compared to before where I was sitting at ~25C... like where is the extra 5C coming from? I would assume the H115i would provide better cooling than the 110i, but maybe I'm wrong? Also the fact that the Package temp is always 20C higher than the core temps, just seems whacky... Do you think it could still be a contact issue? I would assume it isn't because the temps aren't getting too high under load, but I'm not sure what to think at this point... lol. I just don't want to damage my CPU or have another thermal shutdown or anything... My RAID failing really screwed me over, I have found a way to recover the data, but it's put me out of commission now for a few days, and will probably be at least another 24 hours before I am able to recover everything and rebuild the array. Ugh... Sucks when you have to rely on your system for work, I'm sorry if I seem rushed, I really am very appreciative of your help!

Link to comment
Share on other sites

Had someone on another forum suggest that maybe the overheating caused the IHS and CPU to have poor contact, and suggest I delid. Is that possible? Only issue is the 6850k is soldered so I don't really want to delid it... I needed a break last night so I played some games for a few hours and the CPU didnt go over 60c, which I guess is OK, all things considered. Just would like to know what is causing these high temps at this point. Used to be able to OC to 4.3ghz/1.35v and was still only running at 80c in OCCT after 6 hours when I built the machine. Now I am hitting high 90s at stock speeds after 3 minutes... ugh.
Link to comment
Share on other sites

Idle temp changes really only result from a couple of things: 1) change in room/case internal temp; 2) change in processor activity level (power plan - balanced vs performance); 3) physical contact between the CPU + cold plate. The wattage to dissipate at idle even on a 6850x is pitiful. A little 120mm will idle the same as 480mm. If your coolant temp is 26C, then at 0 volts, your cores are 26C. So for them to slow waltz around the 30C mark is as good as it gets for that baseline. Compare that to the idle core temps back in shot 1 when there was a contact issue. I do not think you have a contact issue. I do think you would remember switching to the Performance power plan with a 100% uptime (and it would show in the temps), so that kind of leaves normal daily room temp fluctuations as the most likely difference.

 

Every Haswell-E and Broadwell CPU I have ever seen has this difference in the sensor temp between idle package/socket and the cores. When I first got my 5820 and 5930Ks, I niggled on this forever trying to get some info from Asus about what that sensor was measuring. It is a separate sensor with a different normal range. It should be above the cores at idle, if the socket has power. It also has its own limit range, different from the cores. Back in those days for Haswell, the suggested package limit was 70-80C (cores might be hotter than package at full load). I do not know if that extends to Broadwell, but I would expect it to be similar. It is unfortunate the Asus ROG forums are essentially dead for real information, but there might be more specific info on the relevant chipset forums on overclock.net or somewhere else.

 

I think it is unlikely the silicon was materially damaged during this period. The moment the board senses you have hit Temp X, it shuts down. If you could melt silicon in 1 second at Temp X and fundamentally alter it, then that would be a terrible safety feature. It takes prolonged exposure to high temperature to damage a CPU. When people "fry their CPU", they are usually doing it with current, not heat per se.

 

Well, it looks like I am about 5 generations out of date on OCCT. I just downloaded it. Let me fire it up and see how strong a test this is. Could be your temps are normal for 6850x, unless you were using that same version previously with much lower values.

 

*I ran the standard CPU Auto test (that looks like what you did) with my 8700K, 12 threads, AVX2. It came back at the same 50-55C as my CPU-Z testing. So I do not think there is anything overly difficult about that test. However, I suggest you do not use the Linpack model at this time. I set a CPU record for me at 73C. It power spikes from 80 to 100% to 80 to 100 over and over. This would be extremely unhelpful in the current state for your CPU.

Edited by c-attack
Link to comment
Share on other sites

This is definitely a newer version of OCCT than I used in the past. Just seems crazy that my temps are so much different after this whole ordeal. I did run this thing with an OC of 4.3ghz 1.3v for about 4 years... temps were always safe though, never had issues until the pump failed on the h110i. Maybe it's just coming to the end of its lifespan.

 

The only other thing I can think of that was fishy about this install was that the radiator is a little bigger than the h110i, which resulted in me having to kind of force the CPU plugs into the motherboard. I can confirm they're sitting flush and all the way in though. Not sure how this would affect the temps though, Id assume the computer just wouldn't boot if there was an issue with that.

Link to comment
Share on other sites

Do you get really high package temp on something simple like the CPU-Z stress test? No AVX, linear... it should remove some of the complications. There are a lot of variables these days to using difficult synthetic tests. If there is CPU damage, it should be omni-present at all levels. Not selective about load type, program, etc.

 

Yeah, deliding Haswell-E and Broadwell was a no go as far as I am aware. I am sure someone has done it. Not sure what the results would be. I seem to remember Silicon Lottery would not delid these for you at all. I suspect there is a reason.

Link to comment
Share on other sites

I'll check the CPU-Z stress test again in a moment. I remember that it doesn't get very high in that though, topped out around 64c last time if I remember correctly. Just finishing up dinner.

 

As for deliding, I read somewhere that the 6850k was one of the last Intel CPUs that uses solder, so deliding it is unnecessary/stupid to try.

Link to comment
Share on other sites

The liquid temp to CPU temp math is no so straightforward. Coolant will always be heat in, less heat expelled. The CPU is stuck in the middle, both passing momentary heat and retaining it. However, it is very difficult to take momentary CPU temps and turn them into coolant temp change. It's still watts over time that will make changes.

 

I am of the opinion you are OK and the package temp numbers are reflective of the load type (and instructions). You can try contacting Asus, Intel, or the general 68xx forums at large to see if others have similar peculiarities. If there has been some kind of damage, it will be revealed all the time - not selectively. Both your normal gaming and normal stress test loads seem OK. Just the harder synthetics. These specific versions did not exist when I was using the Haswell-E CPUs and I cannot make a comparison. I suppose the thing that really matters is your normal use. If renders, compiling, gaming, video editing, or whatever you normally do is all OK, then I don't see any reason to loose sleep over the results of OCCT or Prime 95. Still, I am sure you would feel more confident to hear the same from another 68xx owner.

  • Confused 1
Link to comment
Share on other sites

I am going to let it rest for the time being. Cores are idling at 30c which isnt too bad. I am just very persistent lol. I hate when I can't come up with the answer, it drives me insane. I guess time will tell if there is a hardware issue. I am getting to that point anyway, the system was built in Jan 2017 (I know I said 4 years, I was mistaken), and I put some pretty heavy use on it.

 

I really appreciate all of your help brother. It means a lot! I see you out here commenting and helping a lot of other people as well, kudos!

Link to comment
Share on other sites

I feel like an absolute dip**** right now lol. I decided that for the hell of it I would make sure I was using the right standoff screws. I compared the ones installed to the others I had and noticed a clear difference, so I wondered, did I ACTUALLY replace the standoff screws? So I went ahead and swapped them, and low and behold, temps are down... lmao. Ran OCCT for 3 hours at 4ghz/1.2v and never went over 51c. Wow. Thanks for you help man, sorry I am a dumbass. haha!
Link to comment
Share on other sites

 Share

×
×
  • Create New...