Corsair H115i high temps.
Corsair H115i high temps.
I began adjusting my CPU's overclock yesterday and noticed the thermal barrier limited further increases. Currently, gaming loads perform adequately. My setup runs at 4.6GHz on a 6700k with average voltage settings (1.305V in BIOS). While it fluctuates, I observed peaks up to 1.338V at 4.7GHz. This suggests some flexibility in that range. However, pushing to 4.7GHz caused temperatures to reach 87°C and the same core failed a Prime 95 short FFT test twice consecutively before reverting to 4.6GHz. Now my maximum stable temperature is around 83°C (even though gaming loads rarely approach that), which aligns with the thermal limits most recommend for this chipset.
Regarding cooling, those temperatures seem quite high. I suspect a cooler that achieves around 23°C idle should perform significantly better. It cools quickly too—after testing, it drops to 50°C within a second and to 27°C after a few seconds of inactivity. Gaming loads typically sit in the high 40s to low 50s.
With other users' insights, are these temperatures typical? Or should I consider reapplying thermal paste or adjusting fan placement/fan speed?
Hello!
After reapplying your paste, observe a few details!
If you're running a newer prime95 version, avoid using IntelBurnTest since high AVX versions over 26.6 can cause overly unrealistic heat unless you turn them off.
I own a Corsair H105 which performs worse than yours; with my 7700k, 5GHz and 1.3 vcore I reach max 75-77°C on IBT, 50-65°C during CPU-heavy games (like ME:Andromeda), especially in 25-30°C ambient conditions... (summer is here!)
When you mention your H115i is exhausting, does it draw air from the radiator or force air through it and out of the case?
Based on your case's airflow, the intake solution is always the...
if you used your CPU's heatsink without removing the old thermal paste and then re-applied new stuff, that should be the first step. That sounds like what I needed to hear, even though it wasn't what I expected. It'll take about 30 minutes or so, but I'm a bit lazy. I've looked into it a bit and think it's due soon, so I'll do it this afternoon and update you on the progress.
Hello!
After you finish reapplying your paste, observe a few details!
If you're running a relatively recent prime95 build, avoid using IntelBurnTest since versions over 26.6 produce overly exaggerated heat due to AVX operations unless you turn them off.
I own a Corsair H105 which performs slightly worse than yours; with my 7700k, 5GHz and 1.3 vcore I reach peak temperatures between 75-77°C on IBT, around 50-65°C during CPU-heavy games (like ME:Andromeda), especially in warmer environments (25-30°C).
When you mention your H115i is exhausting, does it draw air from the radiator or expel it through it and out of the case?
Based on my experience, directing airflow through the radiator and up from the case works best overall. This approach helps maintain cooler temperatures even in a large case where GPU heat doesn’t drastically raise internal temps.
If you're curious, give it a try—just note that with pull-exhaust fans set to exhaust, I saw a noticeable rise of 5-7°C max temperature.
Also, it seems unlikely it would fail at 87°C, so the issue might be related to voltage stability, memory problems, or something else.
Maebius:
Hello!
After applying your paste again, observe a few details!
If you're using a relatively recent prime95 version, avoid... IntelBurnTest since versions over 26.6(?) generate overly exaggerated heat because of AVX instructions unless you turn them off.
I own a Corsair H105 which performs slightly worse than yours, and with my 7700k, 5GHz and 1.3 vcore I reach max temperatures between 75-77°C on IBT, 50-65°C during CPU-heavy games (like ME:Andromeda), especially in ambient conditions of 25-30°C... (summer is here!)
When you mention your H115i is exhausting, does it draw air from the radiator or expel it through it and out of the case?
Regarding your intake solution, it depends on your case's airflow—optimizing it for the CPU is ideal but less so for your GPU and other components.
Based on my tests, forcing air through the radiator and out (up) proved to be the most balanced choice, especially since I had a sufficiently large case where the GPU's heat didn't drastically raise internal temps.
If you're not too interested, give it a try—just as a reference, I experienced 5-7°C higher max temps when my radiator fans were set to exhaust mode (up).
Also, even at 87°C, it shouldn’t fail; it might be due to voltage issues (like vdroop), memory problems, or both.
Thanks for the advice—it’s a valuable perspective. I was worried about voltage being too low since it was set to 1.305, while others with higher values and adaptive mode still succeed.
I’m using Prime 95 version 28.0 along with LinX. When I run a specific Prime 95 Test (1344 Max), I see much lower temps, and in games I never approached 80°C. But I failed Prime 95 so I reduced the multiplier by 1 and stopped pushing, even though I feel I could go higher (up to 1.338V max at 4.6 with 100% stability).
The fans are directed to exhaust case air from the top, which Corsair recommends. I have all SSD drives, so the case doesn’t get much heat from them, but my Zotac 1080 Ti can get hot—though it’s in silent mode, allowing me to adjust the fan curve if needed.
I’m keen to push the chip safely while maximizing performance.
What stress tests would you recommend? (Try 30 minutes of 1344 Min/Max FFT, run in place, 15 minutes each, plus about 10 minutes Small FFTs, and two LinX tests each time I tweak.)
I rely on
IntelBurnTest,
AsusRealbench
for initial stability checks.
If the system remains stable for an hour, it's generally acceptable.
Following that, I perform some extended 4+ hour encoding or file conversion tasks (such as converting an mkv to avi using WinAvi).
Then I run a few 3d Mark Firestrike tests, though I haven't encountered any issues there—provided everything else is fine, the main focus is on observing progress while increasing the frequency.
Of course, nothing compares to real-world usage, so playing demanding CPU games like BF1 or ME:Andromeda will eventually cause crashes or freezes if stability isn't maintained.
Regarding the LLC setting during overclocking, if you leave it on auto it tends to be too gentle, and voltage drops noticeably under stress, leading to instability. (On Asus boards, 1 is the lowest and 7 the highest.)
For example, with auto mode I set vcore around 1.31 in BIOS, under load it actually dropped to about 1.27.
With LLC 5 and vcore at 1.305, it falls to just 1.296, achieving perfect stability. Of course, not every chip behaves identically, but the principle holds.
If you swap your radiator/fans to act as an intake (I didn't see your PC's image since I was at work), your CPU will likely run cooler.
In summary, your internal components will probably be warmer due to having two intake fans and just one exhaust—especially with the exhaust placed next to the intake. This could affect airflow, but you might want to test it out... perhaps using a windtunnel simulation as a reference.
Cheers o/
The settings haven't changed much so far, making it fairly simple. Referring to the guide for Asus' UEFI BIOS on their website is straightforward. The LLC settings aren't listed there, and I don't see them in any other manual I've found—it seems Asus uses different terminology. The VCORE or multiplier option isn't available on the extreme tweaker page. I haven't noticed the voltage dropping below my limit; it appears to be functioning as expected, staying above it. More testing is required to confirm this.
I planned to use AsusRealBench today, as it's recommended in the Asus guide. For gaming tests, even on Battlefield 1, CPU usage remains moderate. It spikes significantly on Fallout 4 and a single-core performance in World of Warcraft (around 90% usage). I don't have the time or patience to run lengthy tests, so I'm using about an hour of listed tests with some gaming to verify stability. The system started stuttering severely and has since stopped, which suggests the OC is working but I'm getting advice to consider raising voltage or lowering temperature. At 80°C, the throttling at 4.5 GHz slows down the LinX test and causes major freezes. I still have a lot of work ahead, but I can't run tests longer than a few hours right now. No time or patience left.
You should have LLC... it's CPU Load Line Calibration. It's either in the external digi power control or the internal CPU power management section of the extreme tweaker menu... even on my old 2500k motherboard there were LLC settings to adjust (it had auto, low, medium, high or similar options... your board should have numbers from 1-7 or 1-9 other than auto). What tools are you using to monitor your real Vcore? Asus Rog's CPUz displays ViD (requested voltage) instead of actual. Use HWMonitor and check the VCore voltage number during stress tests... if there are no voltage drops then it's fine. The 6700k supposedly doesn't throttle at 80c so you're probably dealing with a memory issue, even so... if you're not, verify your overclocking potential first using stock (2133GHz) memory frequency rather than XMP. I don't play FO4 anymore but big multiplayer maps in BF1 were quite demanding (also it depends on the refresh rate you're using... at 60fps it's not too heavy). ME: Andromeda has poor optimization... running it feels like stress testing the system hehe. WoW is very basic, only uses one core and it spikes only at the start of a fight with bloodlust on and recount showing. 😛
It includes SVID, which acts as the external voltage controller. There were no other alternatives available. It was turned on, and the recommendation was turned off, so I disabled it. The voltage was set to 1.34v and increased it back up to 47 on the multiplier, which resolved the issue with the Small FFT on Prime. So, it appears the voltage was the cause.
I'm about to download Asus and check the results. Reapplying the paste had minimal effect, suggesting it wasn't the main problem. I was maintaining similar readings before and after. (possibly not applying enough? I placed a tiny dot smaller than a pea-sized mark in the center and then applied the block, following all the directions I've seen.)
Anyway, with the voltage at 1.34v, the spike temperatures were significantly higher. I recorded 96°C on one core spike.
HOWEVER, my (balanced) normal and idle temperatures are lower. Idle temps are around 23°C at these settings, whereas I was seeing nearly 30°C before. During stress testing, they dropped by about the same range [4-7°C] (maybe the paste helped after all?).
[Thinking out loud here]
So overall it seems stable so far at 4.7GHz, and I plan to gradually lower the voltage while keeping the frequency, to see what works best.
UPDATE: Increased the multiplier to 48, LinX failed right away, so I returned to 47 and began reducing the voltage. Progress so far at 4.7GHz. RealBench max temperature is 79°C thus far. I intend to keep lowering it after reaching the gym, but it seems the thermal paste has also contributed. So far, things are heading in the right direction!