Checking performance at 94°C using an i9 9900k and NZXT Kraken x72 during stress tests
Checking performance at 94°C using an i9 9900k and NZXT Kraken x72 during stress tests
the_fun_juice
09-28-2018, 12:14 PM #1Hey, I'm just getting started with overclocking and I've boosted my i9 9900K to 5.0GHz. On 3DMark (TimeSpy Extreme) I'm seeing 90-94 on the CPU stress test. It's a bit puzzling since I have a 360mm radiator and the Kraken x72. Anyone have tips on how to bring down these temps while keeping strong performance? Could there be an issue with a faulty component?
My 9700k becomes unstable during the p95 small FFT test, consuming 250 watts. With a 240 mm AIO, temperatures can reach up to 100°C before triggering an error. This occurs at a 5.0 GHz clock speed with MCE enabled. For my setup, the maximum safe OC is around 4.7 GHz while running the small FFT/AVX2 stress test. At that point, the CPU draws about 220 watts and operates near 95°C.
It's possible that an AVX2 offset setting in the OC BIOS alters performance when the AVX2 unit is active. I haven't needed to adjust it much as long as I avoid AVX/FPU stress tests. Under normal conditions, the AVX2 unit functions well at 5.0 GHz.
The 9900k can consume a bit more power because of its extra hyperthreads, but I suspect the AVX2/FPU stress tests will behave similarly. I don't recall whether the 9900k includes additional AVX hardware compared to the 9700k, but I wouldn't expect that.
The temperatures are nearing their maximum before throttling occurs, and you shouldn't keep running for extended periods. The best method to identify these issues is by monitoring both power consumption and temperature readings together. If your system is drawing around 250 watts, it should function normally. However, if the temperatures fall below 200 watts, a cooling issue may be present. Minor FFTs or other stress tests that heavily utilize the AVX/FPU unit can cause significant power draw and high heat output. Using CPUID HWmonitor or similar tools can provide a clearer understanding of power usage in relation to temperature.
My 9700k becomes unstable during the p95 small FFT test, consuming 250 watts. With a 240 mm AIO, temperatures can reach up to 100°C before triggering an error. This occurs at a 5.0 GHz clock speed with MCE enabled. For my setup, the maximum safe OC is around 4.7 GHz while running the small FFT/AVX2 stress test. At that point, the CPU draws about 220 watts and operates near 95°C.
It's possible that an AVX2 offset setting in the OC BIOS alters performance when the AVX2 unit is active. I haven't needed to adjust it much as long as I avoid AVX/FPU stress tests. Under normal conditions, the AVX2 unit functions well at 5.0 GHz.
The 9900k can consume a bit more power because of its extra hyperthreads, but I suspect the AVX2/FPU stress tests will behave similarly. I don't recall whether the 9900k includes additional AVX hardware compared to the 9700k, but I wouldn't expect that.
) to reach 90c on 3 fan aio cooling. I own a Corsair H150i 360mm aio which performs slightly worse in benchmarks and I need to surpass 1.475v on gen1 i7 cpu to achieve 90-95c. What thermal paste did you use? When you installed the heatsink, how much force did you apply with the screwdriver?
That's a very high temperature, I usually have to exceed any intel voltage limits for the voltage (with i7 cpu) to reach 90c on 3 fan aio cooling. I own a Corsair H150i 360mm aio which performs slightly worse in benchmarks and I need to surpass 1.475v on gen1 i7 cpu to achieve 90-95c. What thermal paste did you use? When you installed the heatsink, how much force did you apply with the screwdriver?
KCDarkRanger
10-04-2018, 07:23 PM #6In fact, those are the expected temperatures from this rig under the given conditions. Without power readings, a proper comparison isn't possible. Temperature equals ambient plus power times thermal resistance. You mentioned a Gen 1 CPU, which is Nehalem, and it doesn't support AVX extensions. Those instructions are used in the FFT stress test for 9th generation, making the story less relevant. Gen 1 might mean something different to you, as I've seen some articles with varying processor classifications, but in either case, there will be fewer cores and a bigger radiator.
Waleed_PvP_
10-04-2018, 09:30 PM #7Yeah, I actually missed all this AVX stuff during the stress tests, didn’t realize it impacted even 3DMark. Now I see that all my Gen1 CPUs—like i7-875K, i7-870, i7-970, etc.—don’t have these AVX extensions. It’s probably not possible to overclock the i9s with even Kraken X72; it’s insane how high the temperature goes for just 3DMark (sorry for the bad English).
I'm not certain about the 3D benchmark, since I haven't run it. However, if this also increases the onboard graphics load, it will also affect power consumption. The gen 1 model used PCI 2.0, DDR3, and lacked onboard graphics, with SSE4(+). As mentioned, providing power versus temperature data helps determine if the system is functioning properly or has an issue.
Team, Utilities that don’t cause your processor to overload or underload will establish a reliable thermal baseline. Below is a review of utilities categorized by thermal and stability evaluations based on TDP percentages, averaged over six processor generations with standard settings, rounded to the nearest five percent. Even though these tests cover workloads from 70% to 130% TDP, Windows Task Manager treats each as 100% CPU utilization, which reflects resource usage rather than actual demand. Core temperatures directly mirror power draw (in Watts), which is influenced by the workload. Prime95 version 29.8 Small FFT (AVX disabled) delivers a consistent 100% workload, even when TDP surpasses normal limits through overclocking. Should core temps stay below 85°C, your CPU can handle demanding real-world applications without overheating. • AVX Instruction Sets were added with the second-gen Core CPUs, followed by AVX2 in the fourth generation and AVX512 in later high-end desktops (such as certain X-Series, Extreme, i9, and i7 models). Executing Prime95 with AVX enabled creates an overly demanding workload that may harm stability and push CPU usage to 130% TDP. The second and third generations are somewhat less impacted, but temperatures on the fourth through ninth generations can exceed 20°C higher. Many motherboards in the sixth through ninth generation include “offset” adjustments (downclocking) in BIOS to address AVX issues. A minimum of -3 (300 MHz) may be required to cap core temps at 85°C. Since the fourth and fifth generations lack AVX support, you can set separate BIOS profiles for AVX and non-AVX software. Except for a handful of utilities and specialized computational tools, AVX-heavy real-world applications (like rendering or transcoding) and recent games with AVX should not surpass Prime95’s test workload without AVX. According to Intel’s datasheets, TDP and thermal specs are confirmed “without AVX.” In Prime95 versions 27.7 to 29.4, AVX can be turned off by adding CpuSupportsAVX=0 in the local.txt file found in the Prime95 directory after the first run. Nonetheless, since core temperatures remain unchanged at 29.8 without AVX, it’s simpler to use the standard 29.8. You can also opt for version 26.6, which doesn’t support AVX. Core i 1st, 2nd, Pentium, and Celeron processors do not include AVX instruction sets, so they remain unaffected. Refer to Page 87, Section 5.1.1 Thermal Considerations - The first paragraph, second sentence: The 8th and 9th Generation Intel® Core™ Processor Families Datasheet, Volume 1 https://www.intel.com/content/www/u... This release offers a straightforward method to disable AVX, AVX2, and AVX512 by simply checking the boxes. Also consult the Intel Temperature Guide https://forums. Consider reading: Section 11 Thermal Test Basics Section 12 Thermal Test at 100% Workload CT 😎