PassMark Performance Test: Take Care Other Vista Discussions Discuss PassMark Performance Test: Take Care in the Windows Vista forums; Hi Friends: Take care in interpreting the results of your PassMark performance tests. Because of some anomalous results, I decided to take a closer look at my PassMark composite-score ...

[Znod]

Hi Friends:

Take care in interpreting the results of your PassMark performance tests. Because of some anomalous results, I decided to take a closer look at my PassMark composite-score results. I found several interesting things.

First, and most importantly, it looks like different test CD's can produce widely varying composite-score results. I used the same CD in all my tests where my CD / DVD results are in the 290's (see the snips below--64-bit tests B and D and 32-bit tests A and B). I will use only this CD in future tests for comparability.

My widely varying CD scores rule out across-computer comparisons useless the same CD (possibly different copies though) is used in testing--or unless the results of the CD tests are tossed out as illustrated via my snips (i.e., via the use of my Composite - CD / DVD scores) . Note that if the second procedure is used, then it appears that I have reasonable comparability on all tests except for 64-bit test B.

This result lead me to my second observation. It is important not to assume that a best score, for example, is really a best score even after controlling for the CD used. As implied below, after considering multiple test results, one can identify composite-test results that appear to be spurious--e.g., 64-bit test B. The composite result for this test appears to be spurious since the CPU test result of 1807.4 appears to be too high.

Ignoring 64-bit test B one can draw a reasonable conclusion about what appears to be the difference between 64-bit and 32-bit on my machine. That is, focusing on my Composite - CD / DVD scores, it seems reasonable to conclude that 64-bit speeds my desktop up about 105-107 PassMark composte points vis-a-vis 32-bit.