Announcement

Collapse
No announcement yet.

App: Connect IQ benchmark

Collapse
X
  • Time
  • Show
Clear All
new posts

  • App: Connect IQ benchmark

    Connect IQ benchmark application
    https://apps.garmin.com/en-US/apps/7...b-1d81a7808ba0

    Application displays Connect IQ version, partnumber, device name, screen resolution, CPU score benchmark and software version installed. It can help to develop Connect IQ applications.

    Currently known versions:
    forerunner 235
    4.70, Connect IQ 1.2.6 | 0.4/10.5 Pips
    5.07b, Connect IQ 1.3.0 | 0.4/10.0 Pips
    5.20, Connect IQ 1.3.1 | 0.5/9.2 Pips
    5.40, Connect IQ 1.3.1 | 0.4/9.8 Pips
    6.10, Connect IQ 1.3.3 | 0.5/9.5 Pips
    6.30, Connect IQ 1.3.3 | 0.4/9.9 Pips
    7.10, Connect IQ 1.4.1 | 0.4/9.7 Pips
    forerunner 630
    4.70, Connect IQ 1.2.6 | 0.5/10.8 Pips
    fenix 3
    7.20, Connect IQ 1.2.6 | 0.6/10.3 Pips
    7.34b, [Connect IQ 1.3.0] | 0.6/10.8 Pips
    fenix 3 HR
    3.20, Connect IQ 1.2.6 | 0.6/10.8 Pips
    3.62, Connect IQ 1.3.3 | 0.6/10.2 Pips
    4.20, Connect IQ 1.4.2 | 0.6/9.8 Pips
    fenix 5
    2.72b, Connect IQ 2.2.3 | 5.0/15.9 Pips
    2.91b, Connect IQ 2.3.0 | 4.9/14.9 Pips
    FR 735XT
    4.09b, Connect IQ 2.1.1 | 5.6/19.6 Pips
    4.40, Connect IQ 2.1.2 | 5.6/20.0 Pips
    5.08, Connect IQ 2.1.4 | 5.6/19.2 Pips
    FR 920XT
    7.10, Connect IQ 1.2.6 | 0.5/9.5 Pips / 0.6/8.5 Pips
    8.03b, Connect IQ 1.3.0 | 0.7/8.1 Pips
    Oregon 7xx
    2.41b, Connect IQ 1.2.9 | 23.3/71.4 Pips
    2.43b, Connect IQ 1.2.10 | 20.8/71.4 Pips
    2.50, Connect IQ 1.2.10 | 22.7/71.4 Pips
    2.60, Connect IQ 1.2.10 | 23.3/76.9 Pips
    2.62, Connect IQ 2.2.1 | 22.2/71.4 Pips
    2.70, Connect IQ 2.2.1 | 22.2/66.7 Pips
    2.80, Connect IQ 2.2.1 | 20.8/62.5 Pips
    3.10, Connect IQ 2.2.3 | 22.2/66.7 Pips
    epix
    3.20, Connect IQ 1.2.1 | 1.0/15.4 Pips
    3.40, Connect IQ 1.2.1 | 1.1/13.3 Pips
    vivoactive
    3.30, Connect IQ 1.2.1 | 0.8/10.3 Pips
    4.20, Connect IQ 1.3.1 | 0.8/9.9 Pips
    vivoactive HR
    2.60, Connect IQ 1.2.7 | 6.3/20.8 Pips
    2.65b, Connect IQ 2.1.0 | 6.0/19.6 Pips
    2.71b, Connect IQ 2.1.0 | 6.1/21.3 Pips
    2.74b, Connect IQ 2.1.1 | 5.9/20.4 Pips
    2.80, Connect IQ 2.1.1 | 6.1/20.8 Pips
    3.30, Connect IQ 2.1.4 | 6.0/20.8 Pips
    3.40, Connect IQ 2.2.1 | 6.1/20.4 Pips
    Edge 1000
    8.30, [Connect IQ] | 18.5/66.7 Pips
    8.34b, Connect IQ 2.1.0 | 19.6/62.5 Pips
    Simulator 0.00, Connect IQ 1.2.11 | 66.7/1000-1200 Pips (different results from time to time, minimal timer tick is 15 ms)

    ... more devices welcome!

    Changelog:
    1.0.1 - Added CPU score in "Pips", that means calculations Pi per second. Application utilizes CPU by operation with float numbers. First value is 64-bit float operation score (should be low), second value - 32-bit float used in calculation.

    There is a data field with the same functionality. Please visit https://apps.garmin.com/en-US/apps/9...6-68458b597a7d to get it.
    Last edited by a1234.garmin; 04-12-2017, 09:02 AM.

  • #2
    Nice idea!

    Wondering if ConnectIQ version makes a big difference in score (especially v2 versus v1). Too bad this one does not list the score

    vivoactive HR 2.60, Connect IQ 1.2.7

    My FR235 with 5.07b gives me 0.5/10.1 btw (but results vary a bit with each run)
    My ConnectIQ apps

    Comment


    • #3
      Originally posted by FlipStone View Post
      Nice idea!
      Wondering if ConnectIQ version makes a big difference in score (especially v2 versus v1). Too bad this one does not list the score
      vivoactive HR 2.60, Connect IQ 1.2.7
      Tested and found values are pretty much close.
      vivoactive HR 2.60, Connect IQ 1.2.7 | 6.3/20.8 Pips
      vivoactive HR 2.80, Connect IQ 2.1.1 | 6.1/20.8 Pips
      Last edited by a1234.garmin; 08-09-2016, 05:20 AM.

      Comment


      • #4
        Seems the 735xt and vivoactive HR are quite a bit more powerful compared to the 235.

        Cheers, interesting!
        My ConnectIQ apps

        Comment


        • #5
          fenix 3 HR, SW 3.62, Connect IQ 1.3.3, CPU 0.6/ 10.2 Pips

          Comment


          • #6
            Forerunner 235 | SW 7.10 | Connect IQ 1.4.1 | CPU 0.4/9.5 Pips

            Comment


            • #7
              Fenix 5

              Comment


              • #8
                Any chance you'll add support for the Edge 1030?

                Comment


                • #9
                  Updated. Thanks!

                  Comment


                  • #10
                    Sorry for bumping an old thread, but what does Pips mean exactly? The traditional CPU performance metric is FLOPS (floating point operations per second) and MIPS (million instructions per second). But what is Pips? If it is a version of MIPS, then I don't know what prefix P stands for. And the first post says Pi calculations per second (Pi as in 3.14... ?)

                    I would appreciate clarification, thank you!

                    Comment


                    • #11
                      litetaker based on my reading of the OP, the app calculates the value of pi twice using CPU floating point operations: 1st as a 64-bit float, 2nd as a 32-bit float. As the OP pointed out, the speed to calculate the 64-bit value should be slower because of the increased precision.

                      I think "Pips" is just a term the app developer made up.

                      e.g.
                      "Edge 1000
                      8.30, [Connect IQ] | 18.5/66.7 Pips"
                      means
                      The Edge 1000 (8.30 firmware) can calculate pi 18.5 times per second as a 64-bit float and 66.7 times per second as a 32-bit float.
                      Last edited by WillNorthYork; 06-27-2018, 11:25 PM.

                      Comment


                      • #12
                        Thanks for the clarification. Now to figure out how many floating point operations is involved in 1 computation of pi! But I think it won't be a simple conversion as there are probably other delays that add up in that one second.

                        Comment


                        • #13
                          litetaker np. CIQ apps run in a VM (like .NET or Java), so I'm not sure you can do a straight conversion. That's probably why nobody can do a "bare metal" benchmark and instead have to resort to doing operations with a fixed number of steps.

                          Comment


                          • #14
                            True, although if you can figure out the number of computations in the calculation, you can see how much performance you do get after OS and infrastructure to run the app take up the rest.

                            Comment


                            • #15
                              Actually it is hard to predict how many instructions used each time and each SDK can change things significantly (until you have "assembler" instructions for VM). I guarantee only the constant code for the functons. All the optimizations and improvements lay on compiler and VM (firmware) sides. You may notice that Pips for 64-bit floats on sw 2.72 and VM ciq 2.2.3 on device, build with the SDK from 01.2017 is 5.0, but current VM 3.0.0 on 9.71beta with the SDK 2.4.7 from Jun 2018 make only 1.2 Pips! This is dramatically decreased performance, CIQ team has a report regarding this.

                              Comment

                              Working...
                              X