Announcement

Collapse
No announcement yet.

Test stability

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Test stability

    I built a new PC recently and figured I would give PerformanceTest a try. My intention was also to use it to give feedback on any minor overclocking I did. Having actually done some "overclocking" (setting the RAM to the correct 7-7-7-24 timings as per the packet since I just found the BIOS thought a 9-9-9-24 default was a good idea...) actually gave a considerably lower score of 1667.8 vs the original 1733.6(even for largely unrelated things like the CD test). I then ran the entire benchmark several times with the same exact setup and found the results varied a lot.

    Most of them varied +- 10%, the CD Mark +- 20% and I got one 30% off for the "Windows Interface" test and 25% for the "Complex Vectors".

    I can understand that with windows its impossible to get the same exact results twice, but getting results on the same exact setup with a 30% difference between them is just ridiculous. Is this an inherent issue with PerformanceTest, or some issue with my system? Also is there anything much I can do about it (Ive been thinking that maybe killing stuff like explorer would be a good idea, I guess also that PerformanceTest updating its windows in the background during the test doesn't help either)?

  • #2
    Window updates to performance test are done between tests. So not taken into account.

    CD and HDD testing are subject to many external factors, like the position of the test file on the disk, fragmentation and if the CD drive is already spun up to full speed.

    Windows background task can also have an impact. So if you do enough runs you are sure to see some bad dips in the results more or less at random.

    Of course there are steps you are take to remove some background tasks. e.g. wait for idle after booting, disconnect your network connection, turn off windows updates, turn of the desktop search indexing process, etc...

    But is is easier to ignore the range. Instead if you want to do several runs take the max of each test.

    But if you really want to see the impact of RAM timings it is pointless looking at the HDD and CD test. Instead use the Advanced memory test and compare the graphs.

    Comment

    Working...
    X