extremely high temp with h100v2

I've just built my first computer and the normal temperature of CPU is ~30 C. However, the temperature in game can reach 100 C even I use the performance profile in Corsair Link, and when I use Intel burn test, all the core temperature can jump up to over 90 C. I've tried to reapply thermal paste but the same problem still happens. What should I do now?

First, stop using Intel Burn Test or Prime 95 on Kaby Lake or any other recent processor. There just isn't much utility in it and most of the key component manufactures advise against it as well. You are always going to be limited at the CPU with those or any other synthetic test. Additionally, you are probably on standard default auto settings and you should not run synthetic stress tests without making some BIOS modifications. Since you have an Asus board, read through their guide. It's a bit technical, but the key things for you are this: 1) even if you are not overclocking, you probably want to set a specific defined voltage for Vcore and absolutely if you are going to run synthetic stress tests. 2) On page 3 are some bits about the IA Load Line calibration. I would follow their settings. Kaby Lake CPUs are generally warm and a bit jumpy by nature. That takes some getting used to.

 

All that said, it is a bit weird to see a higher game peak temp and than in Intel Burn Test. It could have been a freak moment without optimized settings, but that still sounds like a lot.

 

1) Using Link or any other hardware monitor, what is the peak Vcore value you normally see?

 

2) When you took the pump off and re-pasted, what did the TIM on the cold plate/CPU lid look like? Was it evenly distributed and thin? Or were there solid chunks of sections of thicker material?

 

3) Make sure the CPU fan header or wherever you connected the H100i v2 lead is set to Full Speed/100% in Q-Fan section of the E-Z BIOS or disabled in the Advanced BIOS monitoring section. The H100i v2 needs a constant 12v supply at all times.

 

The cooler itself likely has no impact on what's happening (save #3). If there is a cooler issue, you will see slow to moderate escalating temperatures, even when doing nothing but reading this post. CPU temps go up and never come down until you power off and wait. Bad contact between the CPU and pump cold plate is different. You get slightly warm idle temps, but huge peak numbers and generally erratic and spiky temp behavior in between (launching browser causes all 4 cores to spike to 60C+, then drop back down as the voltage kicks off). Unfortunately, to some extent this is also part of the Kaby Lake personality and it is not always clear as to the cause. However, 100C is definitely out of bounds. Do not run any more synthetic stress tests until we have a better idea of what's going on. When it comes time to test it, use something mild like Intel XTU. A moderate high load for longer is a better evaluation tool than a frying pan test like IBT/Prime. It also gives you more time to react if there is an issue.

1. If the VCPU shows voltage of CPU, I think the peak is frequently 1.39 and most of time it's around 0.7-0.8, In game, I think it can be over 1.4.

2. http://mattbellis.github.io/turn-key-cloud-chamber/assets/images/peltier_with_thermal_paste.jpg It looks like this, but I see more peaks. First, I applied thermal paste with the x shape but used a spatula to spread it in the second time. The peak temperature decreases a little bit but still can surpass 90 when in game.

3. I connected the pump fan to AIO pump header and case fan to CPU fan.

Indeed, I can see a temperature jump to over 60 when opening Chrome and then it will bound back to 33. When I plays game for 20 minutes, the water temp is ~40 C.

I also have a problem with h100 driver. I have to disable driver signature enforcement to get my cooler realized by Windows, but when I restart windows, it can't recognize the cooler again.

That is a lot more TIM than you need. On a tiny CPU like the 7700K, a pea sized drop in the middle will spread out over the CPU heat plate. The TIM is an inferior conductor compared to the metals involved. It's purpose is the cover the micro gaps or imperfections in the metal surfaces. Thin is all you need.

 

However, I do not think that is the prime offender and that title goes to the 1.39 Vcore. That is more than enough to run 5.0 GHz and this is a common complaint on Kaby Lake. When the chip first launched, there seemingly were 10 posts a day about the cooler not working. It is not uncommon to see auto voltages around 1.37v, more than than the safe recommended value from most board makers. Google up some user overclock guides in addition to the Asus linked one above. They can probably suggest a more appropriate base Vcore for the standard 4.5 Turbo frequency. It will likely be in the 1.20-1.25v range and that will take a ton off the top end of the core temps. In working with people in the past on this, it can be tricky to get the voltage to stay below the auto table if using adaptive voltage. You may need to use a fixed voltage, with EIST and C-states enabled to get the frequency to drop at idle and keep energy use/heat lower. That is most certainly a better arrangement than what you have now and still would use less energy. The other option would be to just go ahead and set-up your overclock now. No matter what frequency you choose, you will be at less than 1.39v and it will be cooler with a performance boost. For both choices, you will need to do the usual trial and error to find the right voltage, so be prepared to for a BSOD or two.

 

Coolant temp at 40C while gaming may not be overly high, depending on your room temperature and definitely not with twin 1080's. Most of your "coolant temp" is going to come from a rise in case temperature as a direct result of the GPU load. Your two GPUs can produce as much as 500-600 watts when loaded. That's a lot of heat. Unfortunately, any rise in coolant temp is passed on to the CPU. So if the coolant temp goes up +5C, then the base CPU temp will also go up +5C. This is likely how you wound up with higher game peak CPU temps vs. IBT. Once you get this sorted, you will definitely want to make your own custom fan curves in Link. The 40C=100% fan speed line is not required and has it's basis in the coolant temp rise from a standard room temp. It won't be 22C inside your case very often. Figure out where the normal baseline coolant temp is. Then your normal load. Set a comfortable fan speed for the normal load, with a sharper rise 3-4C higher up. This should give you an auditory warning if things are out of the normal zone.

 

There seems to have been a rash of driver problems this week. I am a chemistry/physics guy and I can't help you much with this. Take a look at the posts in the Corsair Link section and see if any of those suggestions help.

Ok, I will do research about overclocking. Indeed, after updating Bios and using the stock setting, the peak temperature reduces significantly. Thanks a lot! I'm also a chemist.

For overclocking: http://www.overclock.net/t/1621347/kaby-lake-overclocking-guide-with-statistics

And this: http://edgeup.asus.com/2017/kaby-lake-overclocking-guide/

 

1.4v for your vCore is very, very high and likely far higher than you need to go. c-attack is absolutely correct - that's what is causing your high temperatures. It sounds like you are using the "Auto" overclocking settings; the Asus boards seem to be a bit aggressive in the max voltage in the Auto overclocks. I'm at 5.1 Ghz and have my vCore in the BIOS set to 1.347 and see it actually peak at 1.376.

For testing, consider using ROG RealBench ... this tool will test your full heat load (CPU and GPU) and give you a pretty good indication if your entire cooling system is up to the job. Start with everything at stock and see how it handles the temperature loads. And keep in mind that when it comes to your coolant, the relevant temperature to be concerned about is the temperature of the air that is being blown over the radiator fins. If your radiator fans are set up as exhaust, that air is going to come from your case and, with 2 GPUs, that could get quite warm. Again, if your radiator is configured as exhaust, consider using your motherboard's thermal sensor connector ... this will allow you to place the sensor right in front of the fans. You can then control your case fans according to that temperature. This is how my system is currently configured (my case kinda requires me to have the radiator in exhaust configuration) and it works quite well.