CPU Burn-in Homepage

 
What is CPU Burn-in?
 
CPU Burn-in v1.01 by Michal Mienik is the ultimate stability testing tool for overclockers.   The program heats up any x86 CPU to the maximum possible operating temperature that is achievable by using ordinary software.  This allows the user to adjust the CPU speed up to the practical maximum while still being sure that stability is achieved even under the most stressful conditions. The program continuously monitors for erroneous calculations and errors ensuring the CPU does not generate errors during calculations performed under overclocking conditions.
 
New! Version 1.01 adds second mode that stress tests CPU even further.
 
Download CPU Burn-in v1.01 (Win95/98/NT/2k)
 
Download CPU Burn-in v1.00 (Linux)
 
Why CPU Burn-in is better:
 
In the past overclocking stability was tested by running intensive software such as Distributed.Net or SETI@home. Running either piece of software for 24 hours would generally show any possible instability. A looping Quake3 timedemo was also a good choice.
 
However, there are inherent limitations in these tests:
 
  • Not every error caused by overclocking causes a program to crash or the system to hang. Some errors may be more subtle, such as a slight miscalculation. If such an event occurs and causes a pixel to render a slightly different colour in Quake3 for example, the user is unlikely to notice and overall this is no big deal. However such small errors can have a potentially devastating on distributed projects such as SETI@home, which rely on the reliable processing of data.

  • As shown in the graphs below, CPU Burn-in consistently delivers a higher CPU operating temperature than the above mentioned applications, and other competeing CPU stress test programs. This allows CPU Burn-in to be particularly effective at testing stability and cooling effectiveness.
How does it work?
 
CPU Burn-in constantly cycles FPU intensive functions for a user specified period of time.  The resultant calculations are constantly checked for data integrity.  If the program detects erroneous data the user is immediately informed.  Applications such as SETI@home and Distributed.Net perform no such data checking.  The user must rely on those programs to crash or the system to hang before a problem can be noticed.
 

Benchmarks:

All tests were conducted on the same system at a uniform case temperature of 27° celsius. Temperature readings were read from the Celeron's internal temperature probe using standard Asus monitoring software. Each test was run for 5 minutes to allow the CPU temperature to stabilise at the maximum possible temperature under load from each program.
 
Temperature readings were collected at one sample per second.
 
System specifications:
CPU Celeron 566 (overclocked to 765Mhz) 1.75v
Motherboard Asus P3V4X BIOS revision 1003
Memory 192Mb PC133 Infineon SDRAM
Memory bus speed 90Mhz (PCI = 30 / AGP = 60)
Hard drive Fujitsu 13Mb ATA66 5400rpm
Operating system Windows 2000 Professional (build 2195)
CDROM Panasonic DVD / Ricoh CDRW
Video card TNT2 M64 32Mb AGP
Cooling Medium size socket7 heatsink/fan
 
SETI@home (version 2.4 CLI)
 
SETI@home temperature benchmark
Maximum temperature achieved: 71°C
 
Due to the high FPU load SETI@home is a good general overclock test utility. However as shown in the graph the temperature fluctuates widely as the size of FFT calculations changes. This does not provide the optimum constant heat generation overclockers are after.
 
Distributed.Net client (running OGR-25)
 
D.Net OGR-25 temperature benchmark
Maximum temperature achieved: 59°C
 
OGR is a new distributed project run by Distributed.Net which generates approximately as much heat as the ongoing RC5-64 does.  Unfortunately as it uses the integer unit almost exclusively it does not generate much heat at all.  The maximum achieved temperature of 59°C is well below that achieved with SETI@home and others.
 
CPU Burn - P6 version by Robert Redelmeier (redelm@ev1.net)
 
BurnP6 temperature benchmark
Maximum temperature achieved: 76°C (peak)
 
CPU Burn is a suite of highly optimised programs by Robert Redelmeier written in assembly language. The program comes in various flavours including versions optimised for the AMD K6, Intel Pentium, Pentium Pro family, BX, and MMX.  The program exhibited the highest average temperature of any program, but still did not reach the record temperature of 77°C sustained average by CPU Burn-in.  The peaks in the temperature graph are interesting and helped boost the average which would have otherwise been a disappointing 58°C.
 
It should be noted that attempts to run the BX version of the program for benchmarking purposes failed, with the Asus monitoring software recording 'off the scale' temperature readings.  As my Celeron continued to function normally for more than half an hour during this test, I can only conclude that the reading was false and not a true representation of the temperature reached.
 
CPU Burn-in by Michal Mienik (v1.01)
 
CPU Burn-in temperature benchmark
Maximum temperature achieved: 77°C
 
CPU Burn-in v1.0 is the overall winner in temperature generation and thus the most likely utility to show heat related problems due to overclocking under the most demanding conditions.  Not only does it generate the most heat, but it also includes comprehensive error and integrity checking.
 
CPU Burn-in achieved the highest peak temperature in addition to exhibiting the highest overall average temperature. Click below to download CPU Burn-in and try it for yourself.
 
Bug Reports:
 
Please report all bugs and problems to info |at| cpuburnin.com
 

Download:

Try CPU Burn-in for yourself.  Download the Windows and Linux versions below
 
Download CPU Burn-in v1.01 (Win95/98/NT/2k)
 
Download CPU Burn-in v1.00 (Linux)