Iso9001:2000

[Comark Corp]

Hello, I was wondering if your BurnInTest software has any certification that states that each test (CPU, memory, hard drive, etc) has been tested and validated that it will actually find defective parts. Since our company is ISO9001:2000 certified, we need to provide documents when we are audited that state that each diagnostic actually works as expected.

If not, do you know where we might be able to procure defective parts that actually function well enough to boot windows, yet fail your tests?

Thank you.

[Comark Corp]

Hello, I was wondering if your BurnInTest software has any certification that states that each test (CPU, memory, hard drive, etc) has been tested and validated that it will actually find defective parts. Since our company is ISO9001:2000 certified, we need to provide documents when we are audited that state that each diagnostic actually works as expected.

If not, do you know where we might be able to procure defective parts that actually function well enough to boot windows, yet fail your tests?

Thank you.

Anything at all?

Thanks again.

[David (PassMark)]

ISO9001 doesn't ensure that anything works correctly. I just ensures that you have lots of documentation about on your business processes.

We do our own testing our course, but we don't make these test documents available to the public.

If should be no problem to simulate some defective parts. Try the CD test with a scratched CD. Give an (old) hard drive a good hard shake or a nasty drop during testing, cut the shielding from a USB cable and wrap the data lines around the antenna of a GSM mobile phone, remove the heat sink from the CPU or turn off the CPU fan after booting, etc...

[Comark Corp]

ISO9001 doesn't ensure that anything works correctly. I just ensures that you have lots of documentation about on your business processes.

We do our own testing our course, but we don't make these test documents available to the public.

If should be no problem to simulate some defective parts. Try the CD test with a scratched CD. Give an (old) hard drive a good hard shake or a nasty drop during testing, cut the shielding from a USB cable and wrap the data lines around the antenna of a GSM mobile phone, remove the heat sink from the CPU or turn off the CPU fan after booting, etc...

Hello again, and thanks for your reply.

Granted that some of the tests that you mention would be easy to simulate a failure. Others would be extremely difficult (CPU, SIMD, FPU, etc.) as removing the heatsink or turning off the fan would likely result in a hang of the system rather than a specific failure in the CPU test suite. We don't require your procedures or in depth test results or how the test works, but just a certification. Our ISO process at Comark requires us to maintain proof of validation of the proper operation of any test software used in the manufacturing process. Since we have adopted BurnInTest as our manufacturing test software we need to either maintain a document from the manufacturer certifying that the current version meets the product specifications or we must validate the proper operation of the software ourselves. Can Passmark provide us with such a document?

Thanks again.

[David (PassMark)]

We can certify that the tests operate in the manner as specified in the users guide.

It would be foolish to claim BurnInTest can detect all possible faults in all possible components from all manufacturers. It can't. And we won't be producing a document that could give people a false sense of security. BurnInTest can give you a increased level of confidence that everything is OK. Not 100% assurance.

If this is not sufficient then you might need whatever you need to do to produce the documents you require.

[thanh]

need more information

We can certify that the tests operate in the manner as specified in the users guide.

It would be foolish to claim BurnInTest can detect all possible faults in all possible components from all manufacturers. It can't. And we won't be producing a document that could give people a false sense of security. BurnInTest can give you a increased level of confidence that everything is OK. Not 100% assurance.

If this is not sufficient then you might need whatever you need to do to produce the documents you require.

Hi Passmark,

I once read somewhere in this forum about the percentage of defective parts detected by BIT over a running time.
I need this information to determine a properly burn-in time in our production floor, but I could not find it.
Could you please provide this value statistic for my reference only?
I absolutely agree that there is no 100% assurance.

Thank for your help,

Thanh

[David (PassMark)]

We don't have any figures like this. At least nothing that covers a range of hardware. The rates would vary dramatically in any case between different components and different manufacturers.

In our opinion, the chances or finding a problem in the first hour are relatively high, (the system gets hot, it's the first run across the disk / CD and the first use of some of the drivers). Then every hour after that, the chance of finding a hardware problem drops significantly. The extra benefit of doing 12 hours compared to 6 hours is thus probably not great.

[Comark Corp]

We can certify that the tests operate in the manner as specified in the users guide.

It would be foolish to claim BurnInTest can detect all possible faults in all possible components from all manufacturers. It can't. And we won't be producing a document that could give people a false sense of security. BurnInTest can give you a increased level of confidence that everything is OK. Not 100% assurance.

If this is not sufficient then you might need whatever you need to do to produce the documents you require.

Well, I've finally managed to get through doing our own validation on BurnInTest, and I have found that the BurnInTest standard memory test is not especially good at finding memory faults. Since you only do FF,00,55,AA patern testing, there are many faults it would never catch; including addressing errors where writes to one location changes one or more bits at another location. I have a number of defective DIMMs that were not detected. For windows based testing, I have found that this program:
AleGr MEMTEST
Is much better at detecting memory problems in the DIMMs that I have. See the section titled "Memory test sequence" Since this test does pseudo-random and walking zero/walking one it would detect more types of errors. It also has support for defining your own patterns. Perhaps this would be a good thing to add to BurnInTest? Best of all seemed to be memtest86+, but that is not suprising as it is not a Windows based product, and is better able to manipulate the hardware.

Thanks.

[David (PassMark)]

Yes you are correct that not having Windows loaded into RAM does allow better testing of all the available RAM.

Your comment about the test patterns is not completely correct. We also do a sequence test, 00, 01, 02, 03, FF in consecutive bytes. So it shouldn't be a bad as you are suggesting. But I agree we could probalby add more patterns.

[Comark Corp]

Yes you are correct that not having Windows loaded into RAM does allow better testing of all the available RAM.

Your comment about the test patterns is not completely correct. We also do a sequence test, 00, 01, 02, 03, FF in consecutive bytes. So it shouldn't be a bad as you are suggesting. But I agree we could probalby add more patterns.

I stand corrected. With that test pattern, however, you would only detect an addressing error within a 256 byte (A0 - A7 or MA0-3 as seen by the DRAM) address window. A pseudo-random pattern would be better at detecting address errors over a much larger range of addresses, thereby checking the upper address lines as well. Just a suggestion...

Thanks again.