How much can vary results on the same computer?

[chicot]

I produced the five tests. The system was the same. The results were very different. I mean the ratio of the maximum result for the minimum.

In the tab "Summary" was:

PassMark Rating 1.71
CPU Mark 1.75
2D Graphics Mark 2.74
3D Graphics Mark 1.62
Memory Mark 3.42
Disk Mark 1.028

In separate tests the spread is even higher:

CPU - Encryption 4.78
Graphics 2D - Complex Vectors 6.18
Memory - Write 10.02

Only for disc test ratio max / min was stable.
The worst in terms of the parameters of the spread following section:

Disk - Sequential Write 1.096

My computer is old, but BurnIn test is successful.
No errors.

How to interpret these results?
What ratio of the maximum to the minimum of the results should show your program to a known-good computer?

Sorry for my English.

[David (PassMark)]

In a clean stable system expect multiple benchmark runs to return results with a margin for error of about 5%.

More stable results can be obtained by selecting these settings in the preferences window.

• increasing the test duration
• increasing the number of test iterations (e.g. 5)
• selecting 'keep best result'

There are many external factors that can cause larger variability however. Include software running in the background, thermal throttling, power control / power saving settings, background internet activity, etc..