Jump to content
Corsair Community

High temps with h100i v2, only ~5c lower then evo 212 when running prime95 smallfft?


mikep7779

Recommended Posts

I wanted to upgrade my old Evo 212 today so i purchased a h100i v2 from best buy. The temps im getting now are only 5c lower then my previous temps i was getting with my evo 212 with a SP120 on it. I have the fans exausting through the radiator out of my case.

 

i7-7700k @ 5.0GHZ @ 1.3vcore

Evo 212 Low/Max = 35/91

H100i V2 Low/max = 30/86

 

Ambient in the room is ~22c, the radiator reads at 26-28c depending on time running. The pump is running at 2k+ RPM and the fans are over 1600 RPM. I dont have a screen shot of the link software, but these numbers were from last night.

 

I have tried the following:

I set the pump and fans to performance via the link application.

Disabled fan control for the CPU fan header in the bios.

Tried the secondary CPU fan header.

Added two SP120 fans on top of my case for more pull. This helped by maybe 1c but was so loud.

Link to comment
Share on other sites

Why are you running Prime small FFT to test your system? The limitation is always going to be on the CPU itself rather than the cooling system and that is why there isn't much difference between the two. It is not a good test for how much heat the system can remove. It is a good test to see at what point your CPU will issue a thermal shutdown command. This has been the case for several generations of Intel models now. Granted, I have not gone looking for a tear down of the new Kaby processors, but they are going to be similar to Skylake in most ways. Your +60C CPU delta also indicates this is a CPU limited test. You would need to keep the ambient temp below 20C for this voltage and load to be reasonable. Fortunately, I suspect you don't need to run this type program.

 

Since the 7700K is rather new, there is not a lot of comparative data from real users out there. The cooler can only reduce the CPU temperature by the by the difference between original and current coolant temperature (H100i v2 Temp). So in your data, you have a coolant delta of roughly 6C. If the cooler was 100% efficient, it could reduce the coolant to the same temp as the ambient case temperature. Of course, no cooler is 100% efficient, but that should give you an idea about the maximum possible reduction in your Small FFT test - 6C. With the worlds greatest cooler, you would still be above 80C.

 

The real advantage of a water system is its ability to hold a greater amount of heat at any one time without adversely affecting the CPU core temps. When you are looking at gaming data (or whatever), focus on the average temperature over the period of time. The min/max values are often nearly the same and usually the result of voltage spikes, abnormal instructions, or something else not related to the cooler function. I would suggest running programs, games, etc. you are familiar with to get a better handle on temperatures and what is acceptable to you. If you insist on running a stress test, try something like AIDA64 or Intel XTU that offer a more progressive build up and/or easy to identify deficiencies.

Link to comment
Share on other sites

I'm running small FFTs because it's the easiest way i know to get the CPU temp high. Isn't that what we are testing? The ability to get temperature away from the cpu? What keeps the temp the lowest?

 

How does the CPU delta end up implying that its a cpu limited test? Even with the best cooler in the world i'm only going to shave off another 6c? Basically I should compare lower average temps when running XTU vs the evo 212?

 

I feel like what you're saying makes a lot of sense, i'm just having a hard time internalizing it. At the end of the day I would expect the h100i to keep the CPU cooler because it has a higher heat capacity then the evo. The evo is a $25 cooler.

Link to comment
Share on other sites

Remember how everything fits together. Your CPU is sitting in the socket. Voltage is applied from the pins on the back side. This is the source of the heat. In order to get into the cooling system, all that heat must pass through the entire CPU package, lid, and whatever TIM is inside, then through your TIM material and the radiator's cold plate. The rate at which the heat passes through all of that (faster is better) is down the thermal conductivity of the materials, plus any wacky design choices in the CPU (hello Haswell 4690). The metal is a better transmitter than any of the thermal interface compound, but the TIM is necessary to make sure there aren't any gaps. Air is not a good conductor for these purposes. If you turn up the voltage/heat high enough, the rate at which heat can pass into the cooling system becomes the limiting factor. Prime95 small FFT is like turning the burner to high. The heat is piling up faster than it can pass through the door into the cooling system. In essence, I am asking your to test with the burner on a lower setting, so that the CPU material is not the bottleneck. If there is an issue with the cooler, that is the way to find it.

 

Why did is used to work? Different CPU designs and different basic CPU behavior. I have an old 930 somewhere that is clocked out to 4.4@1.35v. One more tick and it will crash incessantly. That is the limit. I could run Prime 95 all day long with an old H80 and never break 60C. The new CPU's react differently to that type of full throttle load, regardless of cooling system.

 

Run Intel XTU for 5-10 minutes. You can do more if you want, but that is long enough to provide the necessary information. XTU is an easy stress test and temps will not be sky high. The load applied is very even and predictable. It should resemble a smooth sine wave. Before you begin, click the wrench at the bottom and configure the line graph to show your the 4 individuals cores. CPU temp average is not as helpful and can mask problems. If you see odd spikes or a very scribbled line, that might suggest a contact problem or TIM mess. That is something that can be fixed. If it is smooth and at reasonable temperatures, the cooler is likely working as intended. Take note of your starting H100i v2 Temp (coolant temp) and then again at the end of the test. This is the coolant delta and is a good indicator of how the unit is working or if there is an issue.

 

Once that is determined, I still think you should make your evaluation based on your actual use. I have heard people mention 7700 can run 5.0 in air. Perhaps there is a reason for that. Regardless, your actual use and the temperatures derived from it should be the determination on whether you want to keep the cooler or send it back. Because spikes are often program related, I do suggest you look at the average over a period of time. If it does not meet your expectations, than take it back. It should not be a problem unless you decide to become a professional Mersenne Prime hunter and then you will need more than a H100i.

Link to comment
Share on other sites

I really appreciate the effort in making that clear. Makes a lot of sense.

 

I ran intel XTU for 10 min:

26.8 starting cool temp, 27.8 high = delta of 1C

CPU temp were 30-79C on one core, 78C on two others and 74C on a third.

 

Appears to deal with the heat well enough, but now i have to retest the evo. I still feel like id expect it to be lower based on where the air cooler was able to get it too. In reviews the difference between the evo and the h100i was 10+c.

Link to comment
Share on other sites

Hmm... the 1C delta has me a wondering a little. I would like to think the H100i is that efficient, but that is an awfully low delta for any cooler. I was going to say you don't need to test the Evo again, however when (if) you take off the H100i block, take a good look at the TIM spread on the CPU and block. Did it spread out? Are there any uneven patches?

 

How did the XTU line graph look? Was it a relatively smooth wave? Or were there some spikes or stretched values?

 

Unfortunately, you are something of the guinea pig right now for 7700K. I don't have a exact sense of where you should be at 1.30v. If any one else has data, please share.

Link to comment
Share on other sites

Perhaps I was too hasty. Reading through this fairly detailed review, those guys are getting about the same temp range as you, but 0.25v lower at the stock setting with a competing brand 240mm cooler. You may have already read it, but it has some more interesting and useful benchmarks. It also seems 1.29-1.30v is about the cliff for Kaby's before you hit that nasty voltage wall to gain another tick. I could be you are just a smidge on the other side and something you are willing to tolerate to keep the glossy 5.0GHz number. Either way, I am not quite so concerned about a potential contact issue now. Still, if you do take the H100i plate off, look for an even spread of TIM or the opposite.
Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...