×

Discussion Board

Results 1 to 2 of 2
  1. #1
    Registered User
    Join Date
    Mar 2003
    Location
    UK
    Posts
    35

    Nokia vs Siemens JVM performance

    Has anybody compared the performance of a midlet (standard MIDP 1.0 - no extensions) running on Nokia and other phones (e.g. Siemens). I'm wondering how Nokia's JVM implementation compares to other implementations.

    I have a game which is maths and graphics intensive (it uses 3D raycasting). The game runs fine on a Nokia 6310i and 3410. On a Siemens MT50 it runs 3 times slower! Apart from the speed difference the Siemens MT50 also appears to have a buggy JVM implementation.

    Do Nokia phones implement a JIT for extra performance?

    It certainly looks like the Nokia phones have a much better JVM than Siemens phones. Unfortunately the potential customer is not aware of this big difference in quality - all they see in the shop is the marketing message 'supports Java', which would suggest to the customer that all Java phones are the same.

    Any thoughts?

  2. #2
    Registered User
    Join Date
    Mar 2003
    Location
    UK
    Posts
    35

    RE: Nokia vs Siemens JVM performance

    Here are the results of the TaylorBench benchmark. I had to reduce the original midlet's size to under 30K so it would load on the Nokia phones - to do this I removed the storage test code and the large image test file (the developer of the TaylorBench www.poqit.com has the source code available at their site.)

    The only conclusion I can reach from these tests and running my game is that the Siemens MT50 is quite a lot slower than the Nokia phones for standard MIDP midlets. This means that I've got some extra work to do to get my game running at a reasonable speed on the Siemens phones. Hasn't anybody else had a similar experience?

    Times are displayed in seconds (the Taylor benchmark displays times in milliseconds) with longer times indicating slower performance.

    1. Low-level graphics tests

    Using the Taylor benchmark default settings, with 2000 iterations. 2000 iterations was too much for some phones to handle in some cases, so 200 iterations were used in all tests and the result multiplied by 10.

    Line test
    Nokia 6310i: 1.72
    Nokia 3410: 2.17
    Siemens MT50: 3.92

    Rectangle test
    Nokia 6310i: 5.16
    Nokia 3410: 7.63
    Siemens MT50: 5.12

    Ellipse test
    Nokia 6310i: 3.2
    Nokia 3410: 202.44
    Siemens MT50: 10.38

    Arc test
    Nokia 6310i: 2.45
    Nokia 3410: 110.32
    Siemens MT50: 8.58

    Small image test
    Nokia 6310i: 2.31
    Nokia 3410: 2.1
    Siemens MT50: 5.17

    Small font test
    Nokia 6310i: 32.94
    Nokia 3410: 4.13
    Siemens MT50: 10.3

    Control test
    Nokia 6310i: 0.02
    Nokia 3410: 0.07
    Siemens MT50: 0.4


    2. CPU/VM tests

    Using the Taylor benchmark default settings, with 5000 iterations. 5000 iterations was too much for some phones to handle in some cases, so 500 iterations were used in all tests and the result multiplied by 10.

    Array copy test
    Nokia 6310i: 0.834
    Nokia 3410: 3.36
    Siemens MT50: 11.63

    VM test
    Nokia 6310i: 0.1
    Nokia 3410: 0.49
    Siemens MT50: 2.31

    Control test
    Nokia 6310i: 0.042
    Nokia 3410: 0.21
    Siemens MT50: 1.2


    I've uploaded the modified version of TaylorBench to:

    www.abstractworlds.com/bench

    You can download from this site with either your web browser or your phone's WAP browser.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
×