H100i V2 high temps i7 4790k

[devils_oc]

Hello,

 

I hope it’s not a problem that I start a second thread, because the first one I started was in the german forum.

 

at the moment i use this rig:

• I7 4790k@4,4ghz 1.170VID
  
• GTX 1080
  
• Asus MAXIMUS VII RANGER
  
• H100i V2 (connected to the cpu fan header)
  
• Windows 7 64 bit
  
• Prime Version: 26.6
  

 

http://i.imgur.com/Gn8am5K.png

 

Problem:

In the image you can see that my temps go up to 90°C (load)

When I use Aida64 with the FPU test oder Prime with 1344 the temps goes really fast to 90-100°C.

 

In the most games like Overwatch I get temps like 50-60°C with an unlocked framerate.

 

But when I use a Software like Open Broadcaster (which is using the x264 encoder), the temps goes up to 78°C.

 

My old 45€ CPU air cooler could cool the cpu better under load with 4,6ghz@1.210v

The H100iv2 is hotter on 4,4ghz@1.170v

 

That is my first AIO and I don’t know what I’m doing wrong, but I see many users with better results at a much higher voltage or oc.

 

With the Corsair Link software, I can easily control the fans of the radiator but it makes no difference if I use the maximum performance or the lowest (800rpm).

 

The liquid temperature stays nearly the same under load (27°C-29°C)

 

 

I think it could be a mounting problem, but I already tried to remount the thing 3 times (and I always use new thermal paste).

 

One time I used washers for the backplate but every remount doesn’t make a different.

 

Tl;dr:

• My H100i V2 give me better temps than my 45€ cpu air cooler on IDLE
  
• But on load it’s the other way around (better temps with the 45€ cpu air cooler)
  
• Liquid temperature don’t change much
  
• Pump/Radiator fan speed don’t make a difference in temperature
  

[c-attack]

I think you are placing a lot of value in the results of these synthetic stress tests and are operating under the assumption your CPU should be able to easily handle all of them. The problem is these tests do not have any individual value for most people and I am wondering if you are basing your comparisons on unattainable numbers put forth by others.

 

I am not surprised your Haswell hits 90C in seconds on Prime. So do all the others. The merits of running that program in its base form without modification have been discussed at length elsewhere. But if you missed it, you don't need to run Prime 95 unless you are actually hunting for Mersenne Prime numbers or you are trying to bin a bunch of CPUs fairly quickly. The limiting factor on almost all modern CPUs is the physical properties of the chip. With these types of programs, you can easily heat it up faster than you can conduct heat out. This is why you hit 90C in 2 seconds. The cooler has nothing to do with it, other than the cold plate conductivity. That rate of conductivity is material related and fairly standard between brands and models, until you get into high end exotic metals. Regardless, using Prime 95 is an unrealistic test, there is no connection to actual use for most people, and is only an indicator of what happens when you run Prime 95.

 

AIDA64 FPU only is also a very difficult test. This is substantially different than the blend or CPU only tests. My 5930K (4.5@1.295) will hit 78C on the first tick, compared to never breaking 65C on the full blend. It is just a difficult test. Do you have programs you need to run that stress the FPU in this manner? If not, then I am not sure this should weigh on your decisions. Not many people even run the FPU only test because it is so difficult. Make sure you are not comparing numbers to someone running the full blend, cache, or CPU only.

 

Your game temps seem normal and that is an indication you don't have a contact problem. If you've taken it off three times, it should be evident if the TIM is not spreading. Your coolant temperature does not heat up very much during Prime because you likely do not run it for very long and quit after a few moments at 90C (as you should). I would be curious about coolant temp range when gaming. It takes a specific amount of wattage to raise the volume of fluid by 1C. You may find it has a greater range while gaming under longer sustained conditions.

 

The x264 software results are a little more complicated. That is something that has real world use. For these tests, did you revert the BIOS back to factory default conditions? If so, that may be part of the problem. Auto voltage and these kinds of tests tend to produce excessive amounts of voltage as the CPU tries to compensate. Those auto voltage tables are designed to make sure it does not crash, regardless of the load. Temperature efficiency was not the goal. If you were running 4.6 with a specific voltage (adaptive or otherwise), it might perform better because the actual voltage was lower than when on full Auto. It is difficult to believe you can put a cheap air cooler on and run Prime95 below 80 at the stock settings.

 

I don't think you have a cooler problem, but you may need to re-evaluate your expectations or perhaps just your needs. You don't need Prime 95. Don't base your decisions on a set of circumstances that will never occur. The x264 temps are worth evaluating, but I need to find some comparison temperature data. Most sites and people do not list the peak CPU temp on their x264/Handbrake runs.