No announcement yet.

Weird Result on Passmark Rating

  • Filter
  • Time
  • Show
Clear All
new posts

  • Weird Result on Passmark Rating

    I am doing a research to test the performance of FTP server ubuntu 12.04. I use Passmark to obtain the benchmark rating. The test is repeated 30 times. The average result is shown below. However, I obtain weird result as follow:
    Number of Clients CPU mark Memory mark Disk mark
    1 874.67 534.40 486.08
    2 892.67 538.40 657.90
    3 896.60 540.53 680.20

    I read passmark’s document and it said “the bigger the number, the faster the computer.” It is quite strange since the higher number of Client result on the higher rating. Based on my study, it should be otherwise? Anyone have any idea about this strange result?

  • #2
    What are the clients doing?
    What is the hardware spec?

    If this machine is running Ubuntu, how are you running PerformanceTest, as it doesn't support Ubuntu.


    • #3
      Actually I ran FTP server on VMware ESXi. I used wine to run PerformanceTest in ubuntu FTP Server. The skenario is that while some clients download a file in FTP server, I run passmark.
      Here is my hardware spec :
      AMD phenemom X4 BE @3.2 GHz
      8GB DDR3 PC10600
      Motherboard Gigabyte 880GM-USB 3L

      I've tried running Passmark using windows server 2003, and it got the same result. The higher number of Client result on the higher rating. Do you have any idea about this strange result?


      • #4
        We have never tested the behavior in Wine. So that might be having some effect.

        You are also working on the assumption that VMware's behavior is consistent and resource usage is linear with client load. This might not be a valid assumption.

        For example, it might be the case that VMware is limited to 1 CPU core. So as you add clients, each client gets slower and slower, but the load on machine is constant.

        Also I don't think differences of 1% or 2% are all that significant. I suspect you might repeat the whole set of tests and find the opposite result (for those that were very close). Even with 30 runs, one bad run is enough to pull down the average by a percent or two. What do the results look like if you take the max, rather than the average?