Announcement

Collapse
No announcement yet.

Why does duel processor performance vary so much over single processor?

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

  • Why does duel processor performance vary so much over single processor?

    Just curious what one's thoughts are on why a dual processor system does not always perform that much better than a single processor system. I took all the numbers on the Passmark site and dropped them into Excel and charted the performance gain of the dual processor systems over the single processor systems of the same model CPU. The performance ranged from 0% (AMD Opteron 6272) and as high as 65% (Xeon E5-2630). I know there are different bus designs, supporting bridges, and other overhead factors. What I'm specifically looking for is a detailed explanation, a deeper dive, than the superficial variables I mentioned. Anyone have any insight for me?

  • #2
    The causes probably vary from one CPU to another.

    For the Opteron 6272 however the reason is as follows. Most of the dual and quad CPU systems are fairly rare. For rare CPUs the charts on our web site still include some results pulled in from PerformanceTest V7.

    As a result you are effectively comparing PT8 (Multi-CPU chart) with PT7 results (single CPU charts).

    In most cases the PT7 and PT8 results are fairly well aligned. But some CPUs show a larger difference. What we have done today is remove a bunch of the PT7 results from the single CPU charts for the newer CPUs. We don't want to remove all the PT7 results as it would mean dropping around 500 different old CPUs from the charts and we think it is important to have some indication of performance for these old CPUs.

    So if you re-run your comparison, you should get different results now.

    Comment


    • #3
      Hi Allen,

      See what might be the fix here:

      http://www.passmark.com/forum/showth...Opteron-6180SE

      Comment

      Working...
      X