PassMark Software
Global (Go to Australian site)

CPU Heat Generation Benchmarks


Which utility is best for CPU heat generation?

Since the introduction of the Maximum CPU Temp test in BurnInTest 6, it's been a vital tool for PC enthusiasts and overclockers looking to verify CPU and cooling system reliability and stability with a maximized heat generation torture test. In the newest version of BurnInTest (build 1019), we've optimized the heat generation feature to produce higher temperatures in shorter times for torture testing multi-core CPUs.

To test the effectiveness of our new heat algorithm, we've conducted a small benchmark test comparing BurnInTest to other CPU stress test utilities available on the web. The products we have benchmarked are BurnInTest Professional, OCCT, Prime95, Super PI and Hot CPU Tester Pro.
For more information about the products we have tested, our test tools or test methodology, please see below.

Benchmark Results

The graph below compares CPU temperatures in degrees Celsius over the duration of an eight minute stress test:

CPU Heat Generation over eight minutes - Click for the full version

The graph on the left compares Peak CPU temperatures to Average CPU Temperatures for each application over the duration of the stress test. The graph on the right compares the CPU temperature increase from the baseline as a result of performing stress testing with each product.

Peak Temp v Average Temp v Baseline - Click for the full version!
  • Our results demonstrate that:
  • » BurnInTest Professional is more efficient at CPU heat generation than all other tested utilities.
  • » BurnInTest Professional made the CPU 17.1°C hotter than the baseline on average.
  • » At all points of testing, BurnInTest Professional generated a higher temperature than all other tested stress testing software.
  • » BurnInTest Professional generated more heat on average (62.1°C) than the next best application's peak temperature (61.5°C).

Stress Test Utilites

For our CPU Heat Generation benchmark test, we have tested the following versions of utilities:

Program Name Version Number Publisher Name License Type
BurnInTest Professional v6.1019 PassMark Software Evaluation
Prime95 v25.9.4 Mersenne Research, Inc Free
Overclock Checking Tool (OCCT) v3.0.0.25 Tetedeiench Free
Super PI v31.1 Kanada Lab, University of Tokyo Free
Hot CPU Tester Pro 4 (Lite Edition) v4.4.1 7 Byte Computers Evaluation

All tested utilities, including BurnInTest Professional, were either free or evaluation copies of commercial software. Testing was performed on 32-bit versions of software.

Testing Methodology and Tools

A Core2 Duo CPU was chosen for our benchmarking as it represents an extremely common CPU in the current PC population (as of 4 February 2010). For our benchmark testing, we have used the Core2 Duo 6600 @ 2.40 GHz.

To measure the temperature of the CPU, a third party utility named RealTemp (version 3.40) was used.

RealTemp was configured to log the temperature of each CPU core at a rate of one sample per second for a total of 480 temperature samples over eight minutes. Our results show the average of two temperatures (one for each core) for each second over the duration of a stress test.

For each stress test utility, we conducted the benchmark using the "Heat Generation" setting, for example, we used the 'In-Place FFTs' mode for Prime95 which maximizes CPU heat generation. Where a heat generation specific setting wasn't available, we conducted a CPU torture test on default program settings.

The quick test option "Maximum CPU Temp" was selected when testing with BurnInTest Professional.

Super PI was placed on the highest setting, i.e. 32M digits, to guarantee that testing duration would exceed eight minutes.

By default, OCCT produced little to no CPU load during the first minute of stress testing because it conducts "CPU monitoring" prior to stress testing. For OCCT, we have excluded the first minute of results, and increased the test duration by one minute to obtain an equal number of results.

For comparison purposes, we have also obtained a baseline which shows the average temperature of the CPU during system idle. Between testing applications, the machine was left idle until the CPU temperature had returned to the baseline figure (approximately 10 - 20 minutes).

  • Notes on our Testing
  • » Ambient temperature may have an impact on CPU temperatures and results. Our testing was conducted near room temperature (24°C).
  • » Our results are accurate for the Core2 Duo 6600 @ 2.40 GHz with a standard cooling configuration. Your results may differ depending on your hardware and setup.
  • » We have observed that CPUs on 64-bit systems generally run hotter. 64-bit systems or 64-bit applications have not been examined as part of this round of benchmarks, but are likely to be investigated in future.
Current special