i7-6700k overheats with H115i?

Been using this setup for a little over a year. Running 6700k on Rog Maximus VIII Hero board. The only settings for overclocking that I changed is multiplier set to 46 and manually set the voltage to 1.26 (voltage sensor reads 1.28). I know this is a rather low limit for voltage, I know people go with as high as 1.35 with Skylake CPUs.

The CPU is cooled by H115i upgraded with Noctua fans. Idling at around 30-35C, and I had this setup for over a year and it is rock solid stable. However in the past when I ran stress tests in CPU-Z and Aida, the CPU never went above 65C or there about. These days when I'm running CPU-Z stress the temperature spikes to 85C almost immediately and keeps climbing to about 90C. In Aida 64 stress test temperature spikes and climbs to 95C+ and I get CPU throttling.

Obviously I check the radiator and the fans, they are functioning properly, everything is clean and free of dust.

I have no idea what is happening. If the performance of my h115i has degraded I think the CPU idle temps would be much higher, but they are barely above room temperature. Did Aida64 and CPU-Z stress test get more advanced in the last year or so where they can stress my CPU a lot more than older versions?

Here is a screenshot of the stress test. CPU was at 100C and throttling down, and Aida automatically shut off the test after just 4 minutes. H115i liquid was just barely above 40C.

Here is another screenshot. See the spike in thermal throttling right in the beginning, which goes does and stays down? This is after I set H115i pump to turbo mode instead of quiet default mode. Meanwhile, in the past it always stayed in quiet mode and temperatures never exceeded 65C. I'm guessing either the pump is failing or the liquid is full of bubbles.

If you have Corsair Link installed, open it up and look at the H115i Temp (coolant temperature). This is key to separating cooler issues from other problems. Typically, the coolant temp might be 4-7C above the room temp at idle and should be more or less the same as your CPU idle temps (with zero load and processor stepped down). If there is an issue with the pump or flow, you normally see high idle coolant values and very high H115i temps (50C+) with even the slightest load. The coolant temperature is quick to rise, but very slow to come down in those conditions.

I am not sure that is your issue. In your AIDA test, it starts throttling almost immediately. This could happen with a severally blocked or failed pump, but you would like see very high temps all the time in those conditions. Use something relatively mild like AIDA64 or Intel XTU and start the stress test. Keep an eye on the coolant temp. If should move up slowly and come down gently as well. For AIDA, I might expect a +6C or so coolant delta for a 10min test. If it's a lot more, something is wrong with the flow. If it is almost non-existent, that seems more like a contact issue between cold plate and CPU.

If you have Corsair Link installed, open it up and look at the H115i Temp (coolant temperature). This is key to separating cooler issues from other problems. Typically, the coolant temp might be 4-7C above the room temp at idle and should be more or less the same as your CPU idle temps (with zero load and processor stepped down). If there is an issue with the pump or flow, you normally see high idle coolant values and very high H115i temps (50C+) with even the slightest load. The coolant temperature is quick to rise, but very slow to come down in those conditions.

 

I am not sure that is your issue. In your AIDA test, it starts throttling almost immediately. This could happen with a severally blocked or failed pump, but you would like see very high temps all the time in those conditions. Use something relatively mild like AIDA64 or Intel XTU and start the stress test. Keep an eye on the coolant temp. If should move up slowly and come down gently as well. For AIDA, I might expect a +6C or so coolant delta for a 10min test. If it's a lot more, something is wrong with the flow. If it is almost non-existent, that seems more like a contact issue between cold plate and CPU.

Thank you for the response. Yes, I do have corsair link installed and at idle both CPU and coolant temp are usually in mid 30ies.

The temperature spike that you see is not limited to only Aida, even in CPU-Z test the temperature spike to 80-85 almost instantly, and AFAIK the CPU-Z stress test is relatively mild.

During these tests coolant temperature rises slowly and then comes down slowly, just like you described.

OK, then that suggests there is some type of contact problem between the cold plate and CPU. It could be the TIM has lost its consistency and you need to clean and re-apply. Probably more likely is the bracket or backplate has slipped or shifted and the plate and heatspreader are no longer in good physical contact. If you wiggle the pump block, is there any play right now? If you press down with two fingers, does the temperature drop? It seems like you are going to have to take the block off for an inspection and that means a clean and reapply of TIM anyway.

What's bothering me a little is your idle baseline just before the AIDA test starts shows core temps in what appears to be the mid to upper 30's. Nothing overly unusual about that for an overclocked 6700K, but I also think it looks a little jagged, pre-test. I live with AIDA up in my monitor all the time and my Hasewell-E CPU are basically dead flat at idle. 6700K and certainly 7700K are a bit more jumpy, but it could also be the CPU was actually doing something notable prior to the start in those few seconds. How does your core temp graph look when relaxed on the desktop, browsing, etc.? Excessive jumpiness usually indicates issues with the TIM, but that is not a certainty.

So these are my developments. I removed the pump off the CPU, and cleaned off the old thermal compound (which seemed just fine by the way).

I found that the studs that are holding backplate to the motherboard have some slack in them. They are tightened as much as they can be, but there is some play. I saw some videos of people adding washers in the back, but I don't feel like removing my motherboard at this time.

I applied a small dab of compound to the center of the cpu and retightened the pump, and then removed it. The compound was nicely distributed all over the cpu, missing just a tiny bit in on of the corners. This tells me that the contact is nice and tight. I reapplied compound again, reinstalled the pump and went testing.

I observe that at idle my cpu sits in the 30ies. Core 1 and 4 are usually 28-32C and Core 3 and 4 are usually 35-39C. Also at idle h115i sits at arounds 34C.

Then I ran Aida64 stability test, and with pump on quiet mode the temperatures spiked into 80ies and 90ies right away, but there was a lot less throttling than last time. Last time the throttling was 20%, this time it was about 1-3%. Temperature of H115i climbed up to 46C. After a while I set the pump speed to performance and CPU temps dropped to 70ies. No more throttling as well. H115i temperature dropped to 41C.

As far as H115i goes, one tube seems to be pretty warm and the other one is cold. I think this is normal - the fluid from CPU gets hot and travels through one tube, then it gets cooled down in the radiator and returns thought the other one. The pump does make some sounds on start up, as if there are air bubbles in the system, but quiets down after a minute or two.

I don't even know what to think at this point.

+12C coolant delta seems like a lot for the AIDA blend test. Mine would normally be +6C, although on 280mm but with a much larger power draw (5930K - overclock 215w), and the fans only at 1350 rpm. The only other thing would be if you were on Auto voltage and it scaled heavily on the synthetic tests, but you already stated the voltage is peaking at 1.26-1.28v. (that 0.02v overshoot is completely normal).

I think the most probably answer is the cooler is developing a blockage. You should not see that kind of dramatic difference between the two pump speeds and the faster setting is able to push more fluid through the obstruction. I am a little surprised how quickly things turn sour and that likely means the blockage is more than mild now. Contact Tech support through the ticket system and request an RMA. They key details to tell them are the coolant temp delta, that is goes down a lot when you increase the pump speed, and of course the CPU is throttling at that voltage. Also mention the contact was good. It gets into the throttle zone so quickly that seemed to be the answer, but clearly is not. Good luck and try and do the advance RMA if you can where they send you a new unit and then you ship the old back. You have to ask for it and it is not always available.

+12C coolant delta seems like a lot for the AIDA blend test. Mine would normally be +6C, although on 280mm but with a much larger power draw (5930K - overclock 215w), and the fans only at 1350 rpm. The only other thing would be if you were on Auto voltage and it scaled heavily on the synthetic tests, but you already stated the voltage is peaking at 1.26-1.28v. (that 0.02v overshoot is completely normal).

 

I think the most probably answer is the cooler is developing a blockage. You should not see that kind of dramatic difference between the two pump speeds and the faster setting is able to push more fluid through the obstruction. I am a little surprised how quickly things turn sour and that likely means the blockage is more than mild now. Contact Tech support through the ticket system and request an RMA. They key details to tell them at the coolant temp delta, that is goes down a lot when you increase the pump speed, and of course the CPU is throttling at that voltage. Also mention the contact was good. It gets into the throttle zone so quickly that seemed to be the answer, but clearly is not. Good luck and try and do the advance RMA if you can where they send you a new unit and then you ship the old back. You have to ask for it and it is not always available.

That is what I'm thinking, there's got to be some sort of blockage. When the unit was brand new, I specifically ran the same test with pump set on quiet and performance, and there was no difference in CPU temperature at all.

Hi,

I literally joined Corsair forum just to reply to this thread.

I had the exact same problem. High temps with i7-6700k (not overclocked) with Corsair H115i. I installed windows 10 from scratch because of some issues and after that for some reason my CPU temps were like 75ish when I was playing Battlefield 4, Battlefield 1, etc. What I found and did to fix my issue:

1. After reinstalling Windows, I found these https://imgur.com/a/nzeKDaA services using CPU like crazy which caused the Throttling when I ran AIDA64. I know it sounds wired but trust me after I disabled this well no throttling. (Still high CPU temp max 85C)

2. Reapplied Thermal Paste and this time I went with Arctic MX-4 https://www.amazon.com/gp/product/B0795DP124/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1 (been using Arctic Silver 5 and still didn't fix it, high CPU temp max 82C)

3. Watched this [ame=

This done the trick.

Ran AIDA64 max CPU temp was 61C with h115i coolant temp at 36.6C w/ corsair cpu fan speed at 850-900 RPM. I up the RPM at 1250-1300 (Quiet Mode with Max pump speed) and few minutes later temps drop to 59C at 100% load.

Proof:

https://imgur.com/a/tJ9fREE

I've attached a pic of my PC, its not the most clean clutter free build but as my first build I'm proud of myself :)

I wrote this in a hurry but hope this helps and do let me know.

Thanks and Good Luck!!

Hi,

Did you eventually manage to resolve the issue? I am literally having the same problem. I have the same CPU and the same H115i cooler.

Stock fans were replaced with Corsair ML140.

My OC is mild with 4.4Ghz and my voltage is 1.3.

Pump is set to high performance.

While testing I'm running the fans on full speed.

As soon as I start the Aida stress test my CPU spikes and after half a minute reaches 100C and starts throttling. At the same time my H115i water temps rise above 40C. They stabilize back to 30C after a while (10-15min) but something is very strange here.

As soon as I start the Aida stress test my CPU spikes and after half a minute reaches 100C and starts throttling. At the same time my H115i water temps rise above 40C. They stabilize back to 30C after a while (10-15min) but something is very strange here.

Can you elaborate on this part? The key to differentiating a cooler issue from all the other kinds of problems is what happens to the coolant temp. When you start AIDA or any other stressor, it should take about 5 minutes for the coolant to up about +6C. It's about +1C every 30-45 seconds. Is that what you see? Or does the coolant tick up rapidly?

There is nothing magical about 40C and it is the total rise that puts it in perspective. It would be normal to get to 40C is you started at 33-34C. What was the starting coolant temp?

You mentioned when you stopped the test, all went back to normal levels in 10-15 minutes. When you first quit the test, there should be an immediate drop in coolant temp during the first 30-60 seconds. The last couple of degrees often takes longer because the case has heated up that same amount and the entire system must stabilize. However, one of the hallmark signs of a troubled cooler is the coolant temp does not drop when you quit the test, but takes an eternity to come back down to normal levels.

Finally, bad contact between the CPU and cooler is usually evident in everything. CPU temps become spiky and exaggerated for mundane tasks and almost instantly unsustainable for higher loads. A stress test would lead to shutdown levels in 2-3 seconds, assuming the the full load is delivered on initialization of the test. Some testers have a delayed start. A problematic cooler will allow you to do most things, but higher loads and gaming will lead to unacceptable temperatures in a few minutes as opposed to a few seconds.

Can you elaborate on this part? The key to differentiating a cooler issue from all the other kinds of problems is what happens to the coolant temp. When you start AIDA or any other stressor, it should take about 5 minutes for the coolant to up about +6C. It's about +1C every 30-45 seconds. Is that what you see? Or does the coolant tick up rapidly?

 

There is nothing magical about 40C and it is the total rise that puts it in perspective. It would be normal to get to 40C is you started at 33-34C. What was the starting coolant temp?

 

You mentioned when you stopped the test, all went back to normal levels in 10-15 minutes. When you first quit the test, there should be an immediate drop in coolant temp during the first 30-60 seconds. The last couple of degrees often takes longer because the case has heated up that same amount and the entire system must stabilize. However, one of the hallmark signs of a troubled cooler is the coolant temp does not drop when you quit the test, but takes an eternity to come back down to normal levels.

 

Finally, bad contact between the CPU and cooler is usually evident in everything. CPU temps become spiky and exaggerated for mundane tasks and almost instantly unsustainable for higher loads. A stress test would lead to shutdown levels in 2-3 seconds, assuming the the full load is delivered on initialization of the test. Some testers have a delayed start. A problematic cooler will allow you to do most things, but higher loads and gaming will lead to unacceptable temperatures in a few minutes as opposed to a few seconds.

Hi, thank you for taking your time to respond :)

Just as additional info: PC case is Corsair AIR 540 and I have 3x ML120 RGB fans as frontal intake, 2x ML140 on the h115i rad pushing hot air through the radiator at the top of the case and one 120mm fan from the 980ti hybrid pushing air out of the back through the radiator.

I have attached screenshots of the temperatures in idle / load (test) and after. My ambient temperature is 22-23C.

Coolant starts off around 30C. After a few minutes the throttling starts as cores start to hit 100C in spikes. Taking your calculation in equation the temperature rise of the coolant is a lot faster, +1C every 22 seconds.

After stopping the test the CPU temperatures drop and the coolant starts cooling off and reaches 30C in less than 10min. The coolant temperature is dropping faster at the start just as you mentioned.

I have ordered MX4 TIM and will be applying it tomorrow to see if it helps with the issue.

At first glance this looks like a contact issue to me and while it is normal to have one core hotter than the others, your one core is doing something all on its own. Take the pump off, clean it, put on new TIM, and then carefully remount with deliberate attention to tightening opposing corners a little at a time.

However, I don't like this +8C rise in coolant at high pump and fan speeds with 6700K. I don't care what you have it clocked at, you should not be able to reach +8C in 4 minutes with absolutely zero case temp gain. The second thing bothering me is your coolant to core delta at idle is a bit off. The heat transmission back and forth across the cold plate goes both ways and we often talk about your coolant temp being the theoretical minimum CPU temperature. In reality, there is physical space between the coolant and cores so it is not a literal "32C must equal 32C" kind of rule. You will see variance and it is a general expectation. That said, you cores are 6-8C below the coolant while the case temp is the same as the cores. To me this suggests the thermal transfer is not great between the cold plate and CPU or the cooler is retaining extra heat it should not --- or both.

There is nothing conclusive yet and it is near certain if you start a support ticket with Corsair toward replacing the cooler, they will pursue the contact approach first. It would be logical to rule that out now or get the materials you need to do this and avoid delays later. That said, I ran a 540 with 280mm top CPU radiator and a standalone 120mm in the rear slot for my Titan GPU. Your case temps are exactly as expected. The coolant temp on the CPU radiator is not. The only time I idled at 30C coolant was when the room was 28C. My much hungrier 5930K that pulled about 215W never made it to +8C in 5 or 50 minutes. Something is off.

At first glance this looks like a contact issue to me and while it is normal to have one core hotter than the others, your one core is doing something all on its own. Take the pump off, clean it, put on new TIM, and then carefully remount with deliberate attention to tightening opposing corners a little at a time.

 

However, I don't like this +8C rise in coolant at high pump and fan speeds with 6700K. I don't care what you have it clocked at, you should not be able to reach +8C in 4 minutes with absolutely zero case temp gain. The second thing bothering me is your coolant to core delta at idle is a bit off. The heat transmission back and forth across the cold plate goes both ways and we often talk about your coolant temp being the theoretical minimum CPU temperature. In reality, there is physical space between the coolant and cores so it is not a literal "32C must equal 32C" kind of rule. You will see variance and it is a general expectation. That said, you cores are 6-8C below the coolant while the case temp is the same as the cores. To me this suggests the thermal transfer is not great between the cold plate and CPU or the cooler is retaining extra heat it should not --- or both.

 

There is nothing conclusive yet and it is near certain if you start a support ticket with Corsair toward replacing the cooler, they will pursue the contact approach first. It would be logical to rule that out now or get the materials you need to do this and avoid delays later. That said, I ran a 540 with 280mm top CPU radiator and a standalone 120mm in the rear slot for my Titan GPU. Your case temps are exactly as expected. The coolant temp on the CPU radiator is not. The only time I idled at 30C coolant was when the room was 28C. My much hungrier 5930K that pulled about 215W never made it to +8C in 5 or 50 minutes. Something is off.

I'll definitely try re-seating and reapplying TIM, however after what happened I am having doubts. I have launched PUBG and while I was in the main menu my temperatures skyrocketed to over 90C in minutes and the coolant temperature rose over 47C!!! (screenshot attached).

I will be starting the RMA process as that definitely doesn't look like normal behavior, more like pump failure and/or blockage.

Wow, hopefully that 6700K is fine after that torture, at 62C mine went south fast, the IHS thermal paste didn't like getting to that temp and it caused damage, at least that is as far as I saw it get. But, the board the OP is using, there is a set of capacitors near the rear of the board that are inside the area where the pump could hit and it could explain the higher temps on the two cores, if the OP has or hasn't, may want to invest in a telescoping mirror (google that, Amazon has them for like cheap, also the magnetic grabber is nice too for finding screws), attach the pump and then put the mirror down to the board and shine some light, see if those capacitors have any clearance between the pump and the tops. Now ASUS is usually pretty good about supporting water coolers when their board lists it can and making sure things clear, but manufacturing and component changes can make that a problem, make sure they clear, if not, you may need to shave their tops off some (dangerous because of the filings are conductive).

Interested to see what happens, though 600MHz seems underwhelming of an improvement to go through this, most overclocking of this chip (are usually done by a delid) hit 5.6-5.7, that itself is more of an overclock that I would expect to see temps like those, hope the thermal paste inside the 6700K didn't weaken due to those temps or the OP is either looking for a new processor earlier than later or doing a delid to better it.

EDIT: YouTube video of 6700K OC'ed to 4.6GHz, temps before/after Delid[ame=https://www.youtube.com/watch?v=Gp23ZY5DimM]YouTube video of 6700K OC'ed to 4.6GHz, temps before/after Delid.[/ame]

GuruSR.