Announcement

Collapse
No announcement yet.

NVIDIA Jetson AGX Xavier Benchmarks - Incredible Performance On The Edge

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

  • #11
    Originally posted by enihcam View Post
    At 15W, X86-64 can do a better job.
    When counting total system power? Not likely. x86 + decent GPU does not fit in this envelope.
    As a total performance measurement? Then this will get crushed like a bug.
    As a cost efficient option? Nah. x86 is probably the more cost efficient option.

    But that's not the question here now is it?
    I think this is a sweet package.
    Size, power envelope and computing power.

    Sometimes a unique sizing like this is hard to come by.

    I envision the AGX as a base for a lot of vector math signal processing using tensor units.
    Various sensors in cars for example.
    Medical equipment etc.

    Comment


    • #12
      What is the noise level of the AGX Xavier? Do the fans get loud on idle, under load, etc. at the 15W and 30W configurations?

      Comment


      • #13
        Originally posted by fawkes View Post
        What is the noise level of the AGX Xavier? Do the fans get loud on idle, under load, etc. at the 15W and 30W configurations?
        Very quiet if/when it turns on... I actually had to look to see if there was a fan within the heatsink.... I wasn't able to hear it from like 4 feet away.
        Michael Larabel
        http://www.michaellarabel.com/

        Comment


        • #14
          I also would have liked to see how it compares with an x86.

          Comment


          • #15
            Originally posted by Michael View Post

            The max performance mode was used on the Jetson boards tested.
            Sorry but you are definitely not using the right mode... My results will be up within 1h. Many of the tests are more than 30% faster.

            Comment


            • #16
              Originally posted by Michael View Post

              The max performance mode was used on the Jetson boards tested.
              Here you go: TX2 properly configured
              As anyone can see some of my TX2 results are twice better than what you get. This doesn't change the fact AGX is great, but you should fix your TX2 setup.

              EDIT: for the sake of completeness I reran with performance governor https://openbenchmarking.org/result/...SK-1812248SY17
              Last edited by ldesnogu; 12-27-2018, 11:48 AM.

              Comment


              • #17
                Hi Michael.
                Can you perform tests of Snapdragon 850 based laptop using Windows Subsystem for Linux to compare with Jetson Carmel ARM cores.

                Comment


                • #18
                  Originally posted by jold View Post
                  Hi Michael.
                  Can you perform tests of Snapdragon 850 based laptop using Windows Subsystem for Linux to compare with Jetson Carmel ARM cores.
                  I don't have any current Qualcomm hardware, that's why you don't see any Qualcomm results in this article.
                  Michael Larabel
                  http://www.michaellarabel.com/

                  Comment


                  • #19
                    Hi,

                    how was jetson TX2 benchmarked with INT8 models?
                    Because what is from nvidia documentation int8 models are not hardware supported on tx1 and tx2


                    Thanks.

                    Comment

                    Working...
                    X