i7 6700k oc to 4.5 at 1.25 is getting up to 78ºC with a Corsair H100i v2. Is this normal?

[mercador]

Guys, hello.

My build composition is as follows:

• 
  CPU: Intel - Core i7-6700K 4GHz Quad-Core Processor
  CPU Cooler: Corsair - H100i v2 70.7 CFM Liquid CPU Cooler
  Thermal Compound: Arctic Silver - 5 High-Density Polysynthetic Silver 3.5g Thermal Paste
  Motherboard: Asus - ROG MAXIMUS VIII HERO ALPHA ATX LGA1151 Motherboard
  Memory: G.Skill - Ripjaws V Series 32GB (2 x 16GB) DDR4-3200 Memory
  Storage: Samsung - 950 PRO 512GB M.2-2280 Solid State Drive
  Storage: Mushkin - Reactor 1TB 2.5" Solid State Drive
  Storage: Seagate - Barracuda 3TB 3.5" 7200RPM Internal Hard Drive
  Video Card: MSI - GeForce GTX 1080 8GB GAMING X 8G Video Card (2-Way SLI)
  Video Card: MSI - GeForce GTX 1080 8GB GAMING X 8G Video Card (2-Way SLI)
  Case: Corsair - Vengeance C70 (Black) ATX Mid Tower Case
  Power Supply: EVGA - SuperNOVA G2 1300W 80+ Gold Certified Fully-Modular ATX Power Supply

My CPU is manually overclocked at 4.5ghz on 1.25v through BIOS, set as manual mode.
I have my Corsair H100i v2 set to Performance Mode.

Fans distribution:

• 
  two intake on the front
  one intake on the side panel
  one exhaust on the rear
  two exhaust on the top (attached to H100i v2's radiator)

Some stress test results

Intel Processor Diagnostic Tool

Prime95 version 26.6 10 minute run

Are these normal temperatures?

 

[RememberThe5th]

CpuSupportsAVX=0 in the file local.txt. Because prime95 uses avx instructions and stresses the system more that usual when gaming or video editing. Unless your cpu needs to be delided.

 

[mercador]

CpuSupportsAVX=0 in the file local.txt. Because prime95 uses avx instructions and stresses the system more that usual when gaming or video editing. Unless your cpu needs to be delided.

New 13 minute run with CpuSupportsAVX=0 in the file local.txt

You think my CPU needs to be delided?

 

[RememberThe5th]

How did you apply thermal paste?
Where is the pump connected?
Well for delliding it improves temps but that is last option.

If possible try to swap aio for front and put coolers top for exhaust just for test.

 

[mercador]

How did you apply thermal paste?

A pea-sized drop in the center of the CPU. Thermal paste is a 2 year old Arctic Silver - 5 High-Density Polysynthetic Silver 3.5g.

Where is the pump connected?

Pump is connected to the w_pump dedicated header. Fans attached to the radiator are connected to the headers cpu_fan and cpu_opt.

Well for delliding it improves temps but that is last option.

I delided my cpu 10 days ago so unfortunately I don't think that would be very helpful.

If possible try to swap aio for front and put coolers top for exhaust just for test.

The problem with my case is that if I want to put the radiator on the front I won't have the bays for my HDD and SSD
I can try only for testing though

 

[RememberThe5th]

Weird thing is that the CPU Voltage spikes up to 1.3V

 

[mercador]

Weird thing is that the CPU Voltage spikes up to 1.3V

I've also noticed those spikes. Very weird considering that I've set the vcore to manual.

 

[RememberThe5th]

Do you use adaptive voltage or? Im unsure abput overclocking since I dont have K cpu at all, Im OC'ing on 775.

 

[mercador]

Do you use adaptive voltage or? Im unsure abput overclocking since I dont have K cpu at all, Im OC'ing on 775.

Manual mode. Like the image below but at 1.25v instead of 1.5v

 

[devbiker]

For a chip that's not been delidded, what you seeing is pretty much as expected.

Your HWInfo screens do not, however, show the vCore ... the is the voltage actually being delivered to the CPU. You do have VID; this is the voltage requested by the CPU. The actual vCore will depend on a number of factors, one of which, of course, is your voltage settings (things like Load Line Calibration also have an impact). That said, Manual Mode should nail it to 1.25v (if that's your setting).

A couple of things to look for - watch the performance of the cooler through this. You should see temps for the cooler slowly rise and then level off at some point. Where this is will depend on your fan settings and your ambient temperature; in the summer, it'll be warmer. You also want to watch your pump speed. The pump speed should be pretty constant; you may see variations of about 100-200 RPM. Any more than that is an indication that the pump isn't powered correctly. This particular pump gets its power from the fan header that it is connected to - typically the CPU_FAN header - and this needs to be configured to supply a constant 12V to the pump. The easiest way to do this is to set it to full speed. In the Asus BIOS, you can disable the Q-Fan control for the header and that'll do the trick as well.

 

[mercador]

For a chip that's not been delidded, what you seeing is pretty much as expected.

Your HWInfo screens do not, however, show the vCore ... the is the voltage actually being delivered to the CPU. You do have VID; this is the voltage requested by the CPU. The actual vCore will depend on a number of factors, one of which, of course, is your voltage settings (things like Load Line Calibration also have an impact). That said, Manual Mode should nail it to 1.25v (if that's your setting).

A couple of things to look for - watch the performance of the cooler through this. You should see temps for the cooler slowly rise and then level off at some point. Where this is will depend on your fan settings and your ambient temperature; in the summer, it'll be warmer. You also want to watch your pump speed. The pump speed should be pretty constant; you may see variations of about 100-200 RPM. Any more than that is an indication that the pump isn't powered correctly. This particular pump gets its power from the fan header that it is connected to - typically the CPU_FAN header - and this needs to be configured to supply a constant 12V to the pump. The easiest way to do this is to set it to full speed. In the Asus BIOS, you can disable the Q-Fan control for the header and that'll do the trick as well.

Screenshots of the cooler statistics after Realbench v2.43 stress test (15 minutes)

Pump had a minimum 2760 rpm and maximum 2910 rpm
Cooler temperature had a minimum of 31.6 ºC and a maximum of 37.0 ºC

To be able to see the CPU Core Voltage, I've taken a screenshot during the Realbench v2.43 stress test

We can see that the Core Voltage is 1.296v.

As per your comments @devbiker:

• 
  I have 1.25v on Manual Mode and LLC on default. Maybe I should lower LLC?
  
  This is the second H100i v2 that I own. The first one I thought it was performing poorly due to some problem and I sent it to the RMA. Corsair Brasil acknowledged the faulty cooler and sent me this one which is brand new from the box, installed yesterday. The previous one I had up to 3200 rpm on Performance mode, but apparently this one can only get up to 2910 rpm. Is this a bad sign?
  
  I have connected this cooler to the W_PUMP header of my motherboard. I think this specific guarantees a constant supply of 12V. The radiator fans I connected to the CPU_FAN and CPU_OPT. Unfortunately Corsair Link does not recognize them like that since they are not connected to the cooler directly. Should I disable the Q-Fan control for the CPU fans even though the cooler is connected to the W_PUMP header?

 

[devbiker]

Just because they sent you a new one isn't necessarily acknowledging a fault; it's acknowledging that you weren't happy and replacing it regardless. Corsair's customer service is typically quicker than most to do that.

As for your setup, I would recommend putting the fan header on the CPU_FAN header and disabling Q-FAN. This will give you a CPU Fan warning in the event of a total pump failure (this is good to have). Run the fans from the pump header; it'll control the fan speed based on the coolant temperature, which is the appropriate temperature source for the fan speeds.

Now, there are a couple of things that we don't know just yet that may be some factors in what you are seeing. First, RealBench is actually a good test for your thermal system because it stresses both CPU and GPU. The GPU, in particular, is a major source of heat and, depending on your configuration, could actually be dumping its heat into the radiator, which will impact your cooling performance. This is especially true when the radiator is configured as exhaust on the top - a pretty common installation option. If this is how you have it installed, you will need to do a really good job managing the heat from the GPU or it will cause issues with the radiator. If your motherboard has the option for a temperature probe, this can help understand what's going on ... place it in front of the radiator to get an idea of the radiator intake temperature.

With that said, I'm not seeing anything that indicates that the radiator isn't performing as it should. It seems to be cooling down pretty quickly. I would also be interested in how it handles a 30-45 minute RealBench run. This should get you to thermal equilibrium - where your radiator temp stabilizes and doesn't really go much higher. Your goal should be to keep that under 40C and that's going to involve tuning the rest of your fans as well. My cooler is mounted as exhaust on top and I've found that the rear exhaust fan is actually the most important factor keeping the radiator intake temperature reasonable. (Note: I also run a Commander Pro, so I have 4 temp sensors that measure the key areas, including just above the GPU, the radiator intake and the radiator exhaust.)

 

[mercador]

Just because they sent you a new one isn't necessarily acknowledging a fault; it's acknowledging that you weren't happy and replacing it regardless. Corsair's customer service is typically quicker than most to do that.

As for your setup, I would recommend putting the fan header on the CPU_FAN header and disabling Q-FAN. This will give you a CPU Fan warning in the event of a total pump failure (this is good to have). Run the fans from the pump header; it'll control the fan speed based on the coolant temperature, which is the appropriate temperature source for the fan speeds.

Now, there are a couple of things that we don't know just yet that may be some factors in what you are seeing. First, RealBench is actually a good test for your thermal system because it stresses both CPU and GPU. The GPU, in particular, is a major source of heat and, depending on your configuration, could actually be dumping its heat into the radiator, which will impact your cooling performance. This is especially true when the radiator is configured as exhaust on the top - a pretty common installation option. If this is how you have it installed, you will need to do a really good job managing the heat from the GPU or it will cause issues with the radiator. If your motherboard has the option for a temperature probe, this can help understand what's going on ... place it in front of the radiator to get an idea of the radiator intake temperature.

With that said, I'm not seeing anything that indicates that the radiator isn't performing as it should. It seems to be cooling down pretty quickly. I would also be interested in how it handles a 30-45 minute RealBench run. This should get you to thermal equilibrium - where your radiator temp stabilizes and doesn't really go much higher. Your goal should be to keep that under 40C and that's going to involve tuning the rest of your fans as well. My cooler is mounted as exhaust on top and I've found that the rear exhaust fan is actually the most important factor keeping the radiator intake temperature reasonable. (Note: I also run a Commander Pro, so I have 4 temp sensors that measure the key areas, including just above the GPU, the radiator intake and the radiator exhaust.)

Now that you have pointed that out about their customer service I agree with you. They said that they acknowledged the faulty cooler but they never told me what was wrong with it, which I find odd.

Anyways, I have followed your suggestions and I have put the fan header on the CPU_FAN header. Then I connected the radiator fans to the pump header, which is the Y cable coming off the pump.

Since I have two GTX 1080 on SLI I think this is a great source of heat, I agree with you again. My radiator is indeed configured as exhaust on the top. I was planning to put as intake on the front but then I would lose my bays to the HDD and SSD as well as blu-ray reader, so for now I have installed as exhaust on the top.

I have disabled the Q-FAN on BIOS as you can see in the screenshot below.

All the Chassis Fans have the same configuration, DC mode, as the screenshot below shows. Shoud I change anything about that?

As per your suggestion, I have run a 30 minutes Realbench stress test. Here is a screenshot at the 27 minutes mark of the stress test

• 
  We can see that the cooler is very stable at 36 ºC, which is marked as the maximum temperature recorded for it.
  Pump rpm is showing the same readings as the last benchmark run.
  Core Voltage shown on ROG CPU-Z as 1.296 V. This seems a bit high and maybe ignoring my manually set 1.25 V vcore voltage?

 

[devbiker]

Overall, it looks like you are in pretty good shape. Your max temp is 81C and your averages, across all cores, is under 70C. Intel's TIM is notoriously bad and that's what you are seeing here. It's also not unusual for 1 core to be a bit hotter than the others. You likely won't get any better unless you delid and replace the lousy Intel TIM with some liquid metal. That'll drop your temps by 10-20C, easy.

As for the voltage reading, this is likely due to the Load Line Calibration. This will boost/overshoot voltage based on load. It does take your manual setting into account (or it'd be much higher) and a couple of other factors. The Asus boards, in particular, seem a bit happy with applying voltage. I would suggest that you read through their overclocking guide (https://rog.asus.com/articles/guides/the-kaby-lake-overclocking-guide/). Everything in there applies to Skylake and Coffee Lake as well (at least, in relation to the settings, not the clocks).

Having the cooler stabilize at 36C is good. That indicates that your intake is getting cool air to the radiator and you are exhausting heat from the GPUs primarily out the back (I bet if you put your hand over the rear exhaust, it'll be pretty warm). For the case fans, DC mode vs PWM mode is really based on the type of fan that you have. If they are 3-pin, DC fans, then keep it on DC mode. The only additional thing that I would suggest is to get a 10K thermal sensor (like this: https://www.amazon.com/XSPC-XS-10KSEN-Wire-Sensor-10K/dp/B00CMR38LC/ref=sr_1_cc_1?s=aps&ie=UTF8&qid=1530114276&sr=1-1-catcorr) and connect that to the thermal sensor pins on your motherboard. You could then place this in the intake for the radiator and use this as the temperature source for your fan curves (they'd be custom curves too). This will give you really tight control over the thermal environment and the noise if the fans. The challenge you may have is while gaming, which really won't stress your CPU but will heat up your GPUs, you won't get enough air to keep the coolant cool since the CPU won't get very warm at all). Having the thermal sensor in there solves that issue.

 

[mercador]

Overall, it looks like you are in pretty good shape. Your max temp is 81C and your averages, across all cores, is under 70C. Intel's TIM is notoriously bad and that's what you are seeing here. It's also not unusual for 1 core to be a bit hotter than the others. You likely won't get any better unless you delid and replace the lousy Intel TIM with some liquid metal. That'll drop your temps by 10-20C, easy.

As for the voltage reading, this is likely due to the Load Line Calibration. This will boost/overshoot voltage based on load. It does take your manual setting into account (or it'd be much higher) and a couple of other factors. The Asus boards, in particular, seem a bit happy with applying voltage. I would suggest that you read through their overclocking guide (https://rog.asus.com/articles/guides/the-kaby-lake-overclocking-guide/). Everything in there applies to Skylake and Coffee Lake as well (at least, in relation to the settings, not the clocks).

Having the cooler stabilize at 36C is good. That indicates that your intake is getting cool air to the radiator and you are exhausting heat from the GPUs primarily out the back (I bet if you put your hand over the rear exhaust, it'll be pretty warm). For the case fans, DC mode vs PWM mode is really based on the type of fan that you have. If they are 3-pin, DC fans, then keep it on DC mode. The only additional thing that I would suggest is to get a 10K thermal sensor (like this: https://www.amazon.com/XSPC-XS-10KSEN-Wire-Sensor-10K/dp/B00CMR38LC/ref=sr_1_cc_1?s=aps&ie=UTF8&qid=1530114276&sr=1-1-catcorr) and connect that to the thermal sensor pins on your motherboard. You could then place this in the intake for the radiator and use this as the temperature source for your fan curves (they'd be custom curves too). This will give you really tight control over the thermal environment and the noise if the fans. The challenge you may have is while gaming, which really won't stress your CPU but will heat up your GPUs, you won't get enough air to keep the coolant cool since the CPU won't get very warm at all). Having the thermal sensor in there solves that issue.

devbiker, thank you so much for your in-depth explanations, opinions and patience to help me out. Really appreciate it.
I am reading through Asus's overclocking guide and I am planning to get a 10k thermal sensor. Besides that, I did not know what deliding the CPU meant and I will read more about it too. Again, thank you!