What is the optimal load for testing OCs
What is the optimal load for testing OCs
I recently acquired a liquid cooler and have been conducting overclock tests. The current setup at 4.7 Ghz with 1.295V appears stable during the stress tests I've run. ROG Realbench and Intel XTUs both passed the tests, showing similar temperatures. I suspect neither is truly putting significant strain on the CPU. For more realistic testing, I should consider using higher-end components. The new Prime 95 model features AVX instructions, which could impose a heavy workload on the CPU. Additionally, I have an AVX offset, making it unlikely to be a genuine stress test. At standard clock speeds (4GHz), temperatures only reach around 55 degrees during load in both tests. I’m uncertain if this is due to the cooler’s performance or the actual thermal limits. My fans are set conservatively—typically between 25% and 35%—and they remain at or below 65 degrees.
Prime95 is widely regarded as one of the most popular stress tests available. However, I don't prefer it because it demands more from your hardware than other programs. Aida64 is another choice for testing under stress. You're correct, Intel's XTU isn't particularly tough, but it's useful for keeping an eye on temperatures.
Personally, I've noticed that running a straightforward Cinebench R15 can reveal system stability and temperature performance. If your processor finishes the Cinebench test without crashing or experiencing extreme temperatures, it's a solid option.
mcgge1360 :
clutchc :
The Intel Burn Test will push it to the brink.
I assumed IntelBurnTest used AVX, which is why I’m avoiding the new prime95.
Someone mentioned that IBT turns off AVX before execution on certain Intel chips.
I’ve been looking for this detail but haven’t found a source yet.
So maybe my advice should wait until I locate a reference.
The key factors influencing comparisons are ambient conditions, hardware setups, and stress testing tools. To ensure fair evaluation, it's crucial to define terms clearly. “Load” or “full load” can be ambiguous and may refer to anything. Activities like gaming, app use, rendering, conversion, and streaming represent variable tasks with changing demands, making them unsuitable for assessing thermal behavior under consistent heat. Not all tasks are equal; stress tests fall into two types—stability tests involving changing workloads, and thermal tests with constant demand.
According to the datasheets, Intel evaluates its processors using a constant 100% TDP load to confirm thermal specs. Prime95 version 26.6 Small FFT is recommended for CPU thermal analysis because it maintains a steady 100% workload, resulting in consistent core temperatures typical of Core i variants with Hyperthreading and standard CPUs within a few percentage points of TDP. Other tools that don’t overload or underload the processor offer reliable baseline readings.
Tools that avoid either overloading or underloading the processor provide accurate thermal data. Here’s a summary grouped by thermal and stability categories based on TDP percentages, averaged across six CPU generations at default settings, rounded to the nearest five percent: Higher TDP tests lead to increased core temperatures. All tests display 100% CPU utilization in Windows Task Manager, which reflects processing activity rather than % TDP. While actual power consumption changes with speed, voltage, and workload, Prime95 v26.6 Small FFT consistently delivers a stable 100% load, whether running standard or overclocked configurations.
Note: 2nd and 3rd generation i7, i5, and i3 chips support AVX instructions, whereas 4th through 9th generation i9, i7, i5, and i3 models do not. Prime95 versions after 26.6 incorporate AVX code execution on the CPU’s floating-point unit (FPU). Second and third generation processors experience minimal impact from AVX, but fourth to ninth generation models with AVX2 may see temperatures rise by up to 20°C.
Many 6th through 9th generation motherboards handle the AVX challenge by enabling BIOS offset adjustments (downclocking) at 300 MHz or higher to cap core temps at 85°C. For overclocked systems, consider upgrading the CPU cooler and case fans. AVX instructions can be disabled in Prime95 versions beyond 26.6 by modifying the local.txt file in the Prime95 directory. However, since core temperatures remain consistent with version 26.6, sticking to that version is simpler. AVX does not affect Core i1st generation, Core 2, Pentium, or Celeron chips, which lack AVX/2 support.
Intel’s specifications confirm TDP and thermal data are verified “without AVX.” Stability tests like OCCT run cycles surpassing 120% workload, making them less ideal for CPU thermal evaluation. Nevertheless, OCCT will automatically stop at 85°C.
Displayed below are the tools: Small FFT's, Blend, Linpack, and IntelBurn Test. The consistent thermal behavior of Small FFT's is essential for precise core temperature readings. A stable 100% workload ensures that CPU, cooler, socket, motherboard, and voltage regulators can maintain proper heat dissipation.
Shown above: Small FFT's, Intel Extreme Tuning Utility CPU Test, and AIDA64 CPU Test. Intel Extreme Tuning Utility presents a fluctuating workload, which is not optimal for thermal analysis. AIDA64 offers 15 stress test options, each producing varied core temperatures. Although the individual CPU tests are steady at 70% TDP, they fall short for thermal testing. Only the CPU/FPU combination delivers a true 100% load. All other AIDA64 selections involve variable workloads, which are similarly unsuitable for thermal evaluation.