checking 9-9900KS lowering temp and undervolt configurations
checking 9-9900KS lowering temp and undervolt configurations
I tend to be a bit prone to anxiety, especially when it comes to gaming temperatures. Occasionally, the CPU temperature can reach as high as 94°C on stock BIOS settings. That’s quite alarming, so I’ve decided to explore undervolting options. However, compared to other results like i9-9900K and i9-9900KS, undervolt seems to be more popular. Therefore, I opted to apply OC settings while keeping the Vcore at 1.3v without altering the clock ratio, so I stick with the stock 5GHz configuration.
I started testing using the configurations I used (see screenshots below):
*Please note I live in a hot climate; daytime temperatures are around 31-34°C.
*I verified the stock BIOS settings: my Vcore is 1.404v, which I believe is too high.
@1.3v = No BSOD on Cinebench R20 and during gaming on demanding titles like BFV and COD Warzone. It wasn’t stress-tested on P95 because it’s considered an extreme test. No AIDA64.
@1.28v = No BSOD during gaming, with temperatures staying below 77°C sometimes, spiking up to 81-82°C. Cinebench R20 is around 90°C if I’m correct. No AIDA64.
@1.25v = BSOD after several hours of gaming. Cinebench R20 passed. The AIDA64 stability test lasted over an hour and a half, with no throttling at the maximum temperature of 95°C.
@1.26v = No BSOD so far; it seems I haven’t run an AIDA64 test since I played for 5-7 hours without issues. Temperatures are usually under 65°C, except occasionally spiking to 71-72°C, then staying below 65°C again. Idle temps are about 5°C higher than my stock BIOS settings, around 42-44°C. If BIOS settings depend on liquid temperature, a reading of 35°C typically means the CPU is at 36-37°C.
Screenshots (Album) showing current BIOS settings:
https://imgur.com/a/aMs4Mg4
View: https://imgur.com/a/aMs4Mg4
My main aim is to reduce my CPU temperature. So, my question is: Are my BIOS settings within acceptable limits? If I want to pass the AIDA64 stability test, what adjustments would be necessary?
I’ve read that changing VCCIO and VCCSA values might help, but I have no prior experience with this. Could you clarify?
If any of these settings are incorrect, please let me know so I can update them.
Full Specifications:
i9-9900KS
Z390 Aorus Master | BIOS version F11c (Latest)
G.Skill Trident Z 32GB(8GBx4) 3200Mhz CL14 (8GBx2 Trident Z RGB + 8GBx2 Trident Z NEO) B-Die)
NZXT Kraken x72 with Noctua NF A12x25
RTX 2070 Super
EVGA G+ 1000w
Lian Li PC-O11 Dynamic top mount radiator (side and bottom as exhaust, others as intake).
HVNtime,
It seems you've made a good start, but Phaaze88 is right; Prime95 isn't overkill if used properly. P95 v29.8
Small FFTs with all AVX options turned off
represent a consistent 100% load that matches Intel's Datasheets for thermal validation. According to the documents,
TDP and Thermal Specifications are confirmed “without AVX”.
If you choose Prime95 v26.6,
it lacks AVX, but its Small FFTs match the same workload as the v29.8 version without AVX. Should OCCT's initial CPU test be set for
Small Data Set and No AVX,
then it indicates a steady-state 97% load...
P95 isn't excessive. It's designed to evaluate thermal resilience under cooler conditions—small FFT, AVX support allowed. Verify error reporting in the options section.
Cinebench R20 'infinite loop' tests voltage stability.
1) Use an AVX offset of 2 or 3; some apps and stress tests rely on AVX, and matching its frequency to standard SSE helps.
They consume more power and need higher voltage for stability compared to SSE.
2) I lack deep knowledge of Aida64, so someone else should clarify this.
3) It aids memory overclocking, such as reaching 4000mhz, though it's not essential at 3200mhz.
4) Only the AVX offset information is clear from what I understand.
Temperatures on my i9 9900KS range from 29 degrees at the coolest core to 33 degrees C at the hottest core, with the room temperature at 24 degrees C while I'm typing. During this process, I completed 11 runs of IntelBurnTest v2.54, and the maximum spiked temperatures reached 74 degrees in the coolest core and 84 degrees C in the hottest core. The cooler model I used is a Noctua NH-U14S air cooler, paired with a Cooler Master HAF 922 case featuring three 200mm fans (one 140mm and one 120mm). I monitor temperatures, voltages, etc., using HWiNFO64. The IR35201 VR OUT voltage was 1.342 at idle, rising to 1.287 and then 1.318 during IntelBurnTest execution. My Vcore is significantly higher than yours. I set the CPU to 4.8GHz across all cores because I can't distinguish between 5GHz and 4.8GHz in games, prioritizing stability over performance. This isn't the best silicon for an i9 9900KS, and I faced challenges getting it to run smoothly, which was a frustrating experience. My motherboard matches yours (F11c BIOS) and includes 32GB of G.Skill 3600 RAM. I tried a different BIOS variation from someone else, but I don't remember the details or have a link for instructions. If you're interested, I can provide some screen shots of my BIOS settings.
I just completed another set of 11 IntelBurnTest runs at 5GHz Vcore in BIOS 1.35v. The highest spiked temperatures reached were 85 degrees C, with the coolest core at 77 degrees. The IR35201 VR OUT voltage was recorded at 1.342 during idle and ranged from 1.283 to 1.316 while the test was active.
For comparison, you should download IntelBurnTest and HWiNFO64 to analyze your own readings.
HVNtime,
It appears you've made a solid start, but Phaaze88 is accurate; Prime95 isn't excessive when used properly. P95 v29.8
Small FFTs with
all AVX options turned off
represent a consistent 100% workload that aligns with Intel's Datasheets for reliable thermal evaluation. According to the documents,
TDP and Thermal Specifications are confirmed “without AVX”.
If you choose Prime95 v26.6,
it lacks AVX, yet its Small FFTs match those of v29.8 without AVX. Should OCCT’s initial CPU test,
named "OCCT", be set for
Small Data Set
and
No AVX,
then the steady-state workload reaches approximately 97% and closely mirrors Prime95's results without AVX.
A consistent workload is essential for thermal analysis, ensuring the power supply, VRMs, processor, socket, motherboard, and cooler can maintain stable temperatures. This approach helps Intel remove software influences when defining thermal guidelines.
AIDA64 includes four CPU stress test options (CPU, FPU, Cache, Memory) with 15 possible configurations,
producing 15 distinct workloads and 15 unique core temperatures. That’s a complex mix of variables.
The dedicated FPU test demands around 115% of TDP, the CPU/FPU combo about 90%, combined all four tests roughly 80%, while the standalone CPU test sits near 70%. Other AIDA64 options involve fluctuating loads ideal for stability checks but not thermal ones.
“Stress” evaluations differ greatly and can be split into two types: stability tests with changing workloads, and thermal tests with steady demands. Tools that avoid overloading or underloading the processor provide a reliable thermal baseline.
Below is a comparison of utilities categorized by thermal and stability tests, based on % of TDP across six processor generations at default settings, rounded to the nearest 5%:
These tests span from 70% to 130% TDP. Windows Task Manager treats each test as 100% CPU usage, which reflects processing activity rather than actual workload. As shown at the top of the scale, if P95 is set for AVX, it becomes a demanding 130% load. Without an AVX adjustment to cut power draw, core temperatures rise sharply.
If you only play and don’t expect heavy CPU tasks like rendering or transcoding that use AVX, you won’t need to worry about AVX settings. Still, note that the recent game engine’s AVX usage is less intensive, so it shouldn’t surpass Prime95 Small FFTs without AVX.
For a clearer view, consider using
HWiNFO
"Sensors Only" to monitor Vcore, Package Power (Watts), and Package temperature (hottest core) during tests. Even logged data can be misleading; right-click these metrics to view graphs. Keep in mind that frequency, Vcore, and workload directly influence power use, which in turn affects core temperatures.
CT