Various stress test outcomes were observed for Aida64 and prime95.
Various stress test outcomes were observed for Aida64 and prime95.
Hello,
I’ve finally chosen to overclock my 7700k, but the results have been inconsistent with prime95 and Aida64. Right now I’m running at 4.8ghz at 1.3 volts. Aida64 didn’t show any temperature rise above the low 70s after a 30-minute session. Prime95 stayed steady until around the 10-minute mark, then jumped from mid 70s to low 90s right away. I stopped testing when that happened. It seems only temperature affects the outcome with this setup, and all cores and threads were running at full capacity.
When I tried a higher overclock of 4.9ghz at 1.3-1.375 volts, the same problems appeared—temperatures spiked again and stability was shaky.
On Aida64 everything looked normal and stable. Prime95 showed all cores running at 100% until about three minutes, after which one core or two threads stopped using at 100%.
I’m new to this process and wanted advice from others who’ve tried similar setups. I used a real bench for short bursts (around 15 minutes each), and temperatures stayed stable between low 70s and low 80s without any crashes.
Could someone confirm if passing tests on Aida64 would be considered stable, even though Prime95 didn’t behave well? Thank you for your feedback.
Guys
,
The issue with Prime95 is strictly
version
dependent, which is a well known problem. For the record,
there is
nothing
wrong with version 26.6
.
Also, 100% CPU utilization seldom equals 100% workload which means 100% TDP.
Intel tests their processors under carefully controlled conditions at 100% TDP.
Prime95 version 26.6 Small FFT's is ideal for CPU thermal testing
, because it's a
steady
100% workload with
steady
Core temperatures that typically runs Core i variants with Hyperthreading and Core 2 processors within +/- a few % of TDP. No other utility so closely replicates Intel's proprietary test conditions. This is also the utility that Real Temp uses to test Core temperature...
I've been informed that Prime95 may not perform well with the Intel Core i7, potentially leading to CPU issues. Forum members suggested alternatives like RealBench, Aida, and IntelBurn, which provide more realistic testing under actual temperature conditions. There are many discussions on Prime95 and Intel sites highlighting how P95 can impose excessive stress on an i7. I hope others can add their insights.
The problem with Prime95 is mainly tied to the version, which is a common issue. In fact, there is nothing incorrect about version 26.6.
Also, achieving full CPU utilization doesn't always mean 100% workload, which implies 100% TDP. Intel conducts their tests under tightly controlled settings at maximum TDP. Prime95 version 26.6 is perfect for CPU thermal evaluation because it provides a consistent 100% load with stable core temperatures, usually matching those of Core i variants equipped with Hyperthreading and Core 2 processors within a few percentage points of TDP. This tool is also the one Real Temp uses to assess Core temperature sensors. Avoid using Prime95 versions older than 26.6 on 2nd through 8th Generation i3, i5 or i7 CPUs, as these models include AVX Instruction Sets. Prime95 versions beyond 26.6 execute AVX code on the CPU's Floating Point Unit (FPU), leading to unrealistic temperatures up to 20°C higher. AIDA64's FPU test confirms similar outcomes. You can disable AVX in Prime95 versions after 26.6 by adding "CpuSupportsAVX=0" to the local.txt file in Prime95's directory. However, since core temperatures remain consistent with 26.6, sticking with version 26.6 is simpler. AVX doesn't impact Core i 1st Generation, Core 2, Pentium or Celeron chips because they lack AVX support.
Download and run only Small FFT's for a short duration:
• Prime95 v26.6 - http://www.mersenneforum.org/showthread.php?t=15504
If you're overclocked and use AVX-intensive applications like rendering or transcoding, consider lowering Vcore and core speeds or upgrading your cooling solution and case fans to prevent temperatures from reaching 85°C. Many 6th, 7th and 8th Generation boards handle AVX issues by offering BIOS offset adjustments. A value of -2 or -3 (200 or 300 MHz) usually works well. Asus RealBench performs a realistic AVX workload with minor TDP variations, making it ideal for overall system stability checks, regardless of overclock status.
• Prime95 v26.6 is recommended for testing.
Blend is Prime95's default test, which provides a changing workload to evaluate memory stability, while Large FFT combines CPU and memory assessments. Both fluctuate and aren't optimal for thermal analysis. Stability benchmarks like Linpack and Intel Burn Test have peak loads at 120% TDP, making them unsuitable for thermal evaluation. OCCT, which merges Linpack and Prime95, stops CPU tests once temperatures hit 85°C.
SpeedFan's "Charts" display thermal patterns over 13 minutes, illustrating how each test generates distinct heat profiles. The steady thermal behavior of Small FFT's is key for precise core temperature readings. Maintaining a consistent 100% workload is essential for accurate thermal testing.
Steady thermal readings from Small FFT's, Intel Extreme Tuning Utility CPU Test, and AIDA64 CPU Test are highlighted. Intel Extreme Tuning Utility also offers variable loads; although AIDA64's test runs steadily, its TDP values are too low for meaningful thermal analysis. All other AIDA64 combinations are fluctuating, which is not ideal for this purpose.
Other benchmarks such as Linpack and Intel Burn Test feature cycles that reach 120% load, further unsuitable for thermal assessments. OCCT combines elements of Linpack and Prime95, automatically halting CPU tests at 85°C.
The "Charts" in SpeedFan illustrate how different tests produce unique heat signatures.
Observed from left to right: Small FFT's, Blend, Linpack, and Intel Burn Test.
Pay attention to the consistent thermal pattern of Small FFT's, which ensures reliable core temperature measurements. A stable 100% workload is vital for accurate thermal evaluation.
Thank you Computronix... I appreciate the detailed information.