Jump to content
Corsair Community

Problems with h100i setup, cpu spikes, and decreased performance

Recommended Posts

Hello All!


I have been trying to find answer for making my CPU work properly and am having some trouble finding direct answers so I figured I would make a post.


I just upgraded from an AMD stock cooler to an H100I Elite Capellix. I mounted it on the top of the case because I like the fans i have in the front and it seemed like the best place to put it. Everything works and run but When I am at idle the CPU temperature will slowly fall down to 35c the spike up to 45c and repeat the process.


I also ran Time Spy as a test before switching coolers and after and got a worse score for CPU and GPU.


Any help would be appreciated, especially on fan curves or settings that I should check. I can provide photos of tests and temperatures if that is helpful.



Link to comment
Share on other sites

The 35-45C CPU temp cycling is pretty common on AMD CPUs. Watch your CPU voltage at the same time. It surely is momentarily increasing as well. This is not something that can be addressed with fan or pump settings. The "cooling" for any type of cooler is conductive. The radiators, fans, pumps, and fluids are all tools for dumping that heat somewhere else. But when the voltage to the underside of the CPU increases, so will temperature.
Link to comment
Share on other sites

Interesting! so that 10c jump is something that just naturally happens and shouldn't be worried about.


if that's the case Id like some help setting up my fan curves just to make sure I am setting it properly. especially for the radiator fans. The main thing with those two fans I have not been sure about is a) what temps and percentages to set them at and b) if they should target the cpu temperature (3700x package) or the coolant temperature (Elite Capellix temp)


It does make me curious why I am getting worse scores on Time Spy too.

Edited by frobro0926
Link to comment
Share on other sites

For the radiator fans you should use coolant temperature (H100i Elite Temp). This is the measure of the cooler's performance/efficiency. +1C to coolant temp is +1C to CPU temp. Same thing for cooling and -1=-1C. On a 3700X you are not going to see a large change. The CPU's power output (watts) and die size are the main contributors to coolant temp change. Just because the CPU is hot does not mean the coolant should automatically go up and the CPU's first cooling stage is conductive. The crude example is 'you can't heat a tea kettle with 2 core duo at 100C'. It would just make a very tiny spot on the bottom of the kettle warm to the touch. It has neither the watts or surface area to effective change the water temp.


So the good news is your CPU happens to very energy efficient and small in die size. But it does mean most of your cooling is conductive and the radiator and fans likely can dissipate the added wattage without too much effort. Remember that when the CPU is 80C, but the coolant has only gone up +3C. CPU temperature and voltage are the dominant limiters for all CPUs. Even the extreme overclockers cannot get around this and that is what they use liquid nitrogen for -- to reduce the initial starting temperature of the CPU. They can't make the CPU not increase by +100C when they apply 1.90v to it.


Now the tricky bit is getting the curve set up. Your room and case temperature will set the baseline and this is different for everyone. As such, each set-up is unique and everyone should make their own curve. There is no way to make a universal preset for the entire planet. Set your radiator fans to a quiet level, what you would like them to be while quietly working on the desktop. Let it sit/desktop work for 5 minutes and note the coolant temp. That is your baseline. Set the bottom of the curve there (28C=750 rpm or whatever). Put a second dot at the same speed and move it about 3C down the line (31C=750 rpm). This creates a small buffer zone since your room temp is going to shift throughout the day. After the +3C start ramping up slowly. The next part will take some figuring, but essentially you want to the fans to still be reasonable when you are at load. If you mostly do CPU intensive tasks, you can use a simple stress test to calculate this. However, a large percentage of people will see their highest coolant temps in mixed load scenarios like gaming. The GPU waste heat will increase case temp and coolant temp. Not much you can do about that, but you want the fans to be set for that real use situation and not some benchmark. Most people will see about +10C over their idle when in a closed case and long GPU load. When you discover where your normal peak is, set a fan speed you can tolerate (39C=1300 rpm) or whatever it is. Save one final point as the emergency notification. You should never reach 50C coolant temp. Set the last dot to 50C=2400 rpm or max speed. That will give you an immediate audible warning if something is wrong.


For the rest of the case fans, you have some options. You can use the same coolant temp for balancing out opposing sides. For open front intake and radiator top with an air cooled GPU, I might use GPU temp on the front and rear fans. That is the prime heat mover in the case and the radiator fans will always have enough air. You also have a single temp probe slot on the Commander Core and you can pick up a cheap 10K thermistor wire to run to the back and get an exhaust temp, but that might be something to try later if you are not happy with other control methods. It also has it's own unique temp properties and it is likely better to keep it to one new variable at a time.

Link to comment
Share on other sites

As for the performance drop in Time Spy, it may very well be the CUE program itself. It is a full spectrum monitoring program and it uses both CPU and GPU resources for various things in the UI. You wouldn't want to run tests with the KB UI up and the system rendering 100+ little RGB keys. Most pages are far less than that, but it is still a monitoring program and will affect your score, since any resources not committed to 3D Mark means less for the benchmark.
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Create New...