It seems with BurnInTest Pro ver 5.3 Build 1005 (in both demo and registered mode), the "Test Pattern" setting in the preference dialog has no effect. I've tried several patterns (ones, sequential, etc) but the default cycling pattern is always used regardless. Same result with "Standard" and "Address Windowing Extensions" mode in Windows XP Pro and Window XPe.
Background:
I'm trying to diagnose an odd 'memory' problem on several Intel 945G chipset-based embedded motherboard where running the 3D test and the RAM test together often produces a high frequency of memory errors. Running the RAM test alone results in few or no errors overnight. I've run other non-windows memory tests overnight with no errors detected. Replacing RAM has not solved the problem.
The memory errors detected during a single run always occur within a small address range (~1MB or so), but that range shifts when the computer is rebooted (and sometimes the errors are not detected at all). The expected value vs the read value differ significantly (for example: Read: 0x03e70000 Expected 0x00006014) and the 2 least significant bytes of the read value are always zero. I cannot confirm these values using a RAM editor because the allocated test memory block is zero-ed as soon as the test fails. Seems like there is some kind of memory addressing failure that is exacerbated by 3D processing the Intel Graphic Media Accelerator.
Note: a nice addition to the test pattern would be "expected memory address" to help diagnose memory addressing problems.
Any assistance would be appreciated.
Background:
I'm trying to diagnose an odd 'memory' problem on several Intel 945G chipset-based embedded motherboard where running the 3D test and the RAM test together often produces a high frequency of memory errors. Running the RAM test alone results in few or no errors overnight. I've run other non-windows memory tests overnight with no errors detected. Replacing RAM has not solved the problem.
The memory errors detected during a single run always occur within a small address range (~1MB or so), but that range shifts when the computer is rebooted (and sometimes the errors are not detected at all). The expected value vs the read value differ significantly (for example: Read: 0x03e70000 Expected 0x00006014) and the 2 least significant bytes of the read value are always zero. I cannot confirm these values using a RAM editor because the allocated test memory block is zero-ed as soon as the test fails. Seems like there is some kind of memory addressing failure that is exacerbated by 3D processing the Intel Graphic Media Accelerator.
Note: a nice addition to the test pattern would be "expected memory address" to help diagnose memory addressing problems.
Any assistance would be appreciated.
Comment