No announcement yet.

3Ghz P4 faster than 3.2Ghz P4 in CPU test. Why?

  • Filter
  • Time
  • Show
Clear All
new posts

  • 3Ghz P4 faster than 3.2Ghz P4 in CPU test. Why?

    Support Question:
    I am getting strange PerformanceTest results for the CPU test. I have a Intel Pentium4 running at 2.8Ghz (or 3.06Ghz) that gets better scores in the CPU test than the new 3.2Ghz P4.

    The 3.06Ghz CPU has a front side bus (FSB) speed of 533Mhz and 512KB of cache. The 3.2Ghz P4 has a FSB of 800Mhz and 1MB of cache. So the new CPU should get better scores than the old ones.

    There are many other factors that effect performance besides clock speed. In this case it is probably the extended 31 stage CPU pipeline in the 3.2Ghz Prescott didn't work as well as the 20 stage pipeline in the previous Intel CPUs (like the Northwood). It is well documented that long pipelines are not always good for performance. The larger cache might also be slightly slower in speed despite being bigger in size.

    Longer pipelines are good for increasing the clock speed however. But in this case the slight increase in clock speed doesn't seem to compensate for the inefficient long pipeline.

    What are CPU pipelines?
    Pipelining is an implementation technique used by CPU manufacturers in which multiple instructions are overlapped in execution. It also exploits parallelism among instructions in a sequential instruction stream.

    Unfortunately the CPU often needs to make a guess about what code is going to executed in the future in order to load up the pipeline with new instructions. If it guesses incorrectly or the result of one instruction depends on the results of a previous instruction, the pipeline starts to become inefficient and overall performance decreases.


  • #2
    I am also running a 3.2 E processor which is geting either the same or lower scores than my old 2.8 C processor (both a socket 478 P4, 800FSB).

    However, the computer would seem to be somewhat faster with the 3.2GHz over the 2.8GHz processor.

    Also, my CPU and memory scores seem to be wildly different from test to test with scores usually being very low and the occasional (after 10 to 20 tests) high score.

    We really need some standard way to run PT or the numbers seem to have little meaning.



    • #3
      When you say, "seem to be somewhat faster". How did you measure this? I assume you didn't use PerformanceTest as this would contradict your 1st statement. Or was it just a gut feel that it might be running faster that particular day? Maybe you were in a good mood or maybe you were doing some particualar task where the CPU performance was in fact better.

      The CPU and memory results should not vary by more than 10% from one run to the next. In most cases the variation is less than 5%. Maybe you have some tasks running in the background using up CPU time.



      • #4
        but my 3.2c is slower than...

        I have the Northwood "c" version of the P4 3.2, and it too is slower than the 3.06 non-HT. Mine is HT and it is enabled. My mobo is the D865PERLk. Comparing mine to the HT enabled 3.06 and my 3.2c falls further behind. ?


        • #5
          Check the number of processes when benchmarking HT CPUs

          Be sure to edit your preferences and change the processes to reflect the
          number of logical CPUs to test when testing hyperthreaded enabled CPUs.