Announcement

Collapse
No announcement yet.

Everything You Need To Know About The NVIDIA Jetson TX1 Performance

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

  • #11
    APM Mustang can take more than 8GB RAM. I have 16GB in mine ;D

    Comment


    • #12
      Price will kill this. What is Nvidia thinking? Design something for small embedded solutions and price it like a high end desktop CPU??

      At the same time, I think we all know that producing a CPU is expensive. Maybe this should strictly be for Nvidia's high end (large and expensive) solutions? But even so, Nvidia probably makes something better in that class.

      If Nvidia would only partner with some more people maybe? They need somebody that can push out lots of volume of product. You used to find a few Tegras in non-Nvidia items, but lately, on their newer stuff, seems pretty limited to only things actually made by Nvidia (?).

      Two choices:

      1. Get out

      2. Partner with somebody that can push high volumes of something that everybody wants.

      Just my opinion. Nice ARM CPU... that we'll never see in anything we can afford...

      Comment


      • #13
        Yeah the Nvidia hardware is nice... if you like being stuck on ancient obfuscated or closed software. My Nexus 7 tablet runs CM 12.1 with a 3.1 kernel! My Ouya is basically a doorstop now. Both uses Tegra 3, which Nvidia apparently no longer support.

        I can see the Nvidia has "updated" to a 3.10 kernel for the Jetsons.... I wouldn't touch it with a ten foot pole...

        Comment


        • #14
          Dear Phoronix,
          I'm a little bit curious why in some benchmarks TK1 is better than TX1. Let me ask is your TK1 board based on Denver dual-core Tegra or is it 4-core Cortex-A15 version? If the later then I'm afraid TX1 should definitely be faster. Could you be so kind and verify that code you run on TX1 is 64bit? The thing is, if you compile for 64bit you do have way much more registers in CPU and generally speaking compilers like that and produce faster code.
          Thanks!
          Karel

          Comment


          • #15
            I am having problems running the SHOC tests. The tests just fail instantly and produce no result. I'm not sure how to debug this. Could you help? Also, is there a test profile number I could use to test against the Alexnet benchmarks? I would like to see where my 2630QM and my nVidia GTX570M place in these tests.

            Comment


            • #16
              I have a NVidia TV Shield station and I could install Ubunt 14.04 on it. I have two questions:

              a) TX1 processor should be an eight core. Why there are only four cores?

              b) I wanted to compile the kernel 3.10 from git://nv-tegra.nvidia.com/linux-3.10. I get code but I can't compile it. It is always saying that while compiling vdso32 there are errors: make[2]: *** [arch/arm64/kernel/vdso32/vgettimeofday.o] Error 127 How can one create a kernel for it?

              Comment


              • #17
                Hausi14, The TX1 is a standard big.LITTLE configuration, which has 4 high performance cores and 4 low power / high-efficiency cores for different situations. You can only use 4 cores at a time, unless do some hardware voodoo like Samsung and MediaTek, which have modified versions of the big.LITTLE architecture that can use all cores at once. I know that Samsung refers to this as Heterogeneous Multi Processing with global task scheduling. It's not easy to implement, which is probably why nVidia's chip doesn't have it.

                Comment


                • #18
                  Originally posted by SuperIce97 View Post
                  Hausi14, The TX1 is a standard big.LITTLE configuration, which has 4 high performance cores and 4 low power / high-efficiency cores for different situations. You can only use 4 cores at a time, unless do some hardware voodoo like Samsung and MediaTek, which have modified versions of the big.LITTLE architecture that can use all cores at once. I know that Samsung refers to this as Heterogeneous Multi Processing with global task scheduling. It's not easy to implement, which is probably why nVidia's chip doesn't have it.
                  Thank you very much for this information. In fact, on the ODROID XU4 I have eight cores and all cores run very well. I once read HMP but did not exactly realize that this is needed to get all cores.

                  Comment


                  • #19
                    Originally posted by phoronix View Post
                    Phoronix: Everything You Need To Know About The NVIDIA Jetson TX1 Performance

                    While last week we were able to write about the NVIDIA Jetson TX1 development board, at that time we weren't able to share any benchmarks or hands-on experience with this ARM board powered by NVIDIA's Tegra X1 SoC. The embargo on that has now expired and as such this morning there are a lot of benchmarks to share with you. There are many benchmarks looking at different areas of the Jetson TX1 including power consumption and thermal. For kicks I've also done some comparisons against the Tegra 2 and Tegra 3 as well as other ARM hardware like the now defunct Calxeda ARM server and Raspberry Pi 2.

                    http://www.phoronix.com/vr.php?view=22414
                    Great and interesting article! Left you a tip

                    Comment


                    • #20
                      Awesome little board! If only it were cheaper, i'd have one to replace my aging core2duo-based media center...

                      Michael, have you considered taking the pcie slot for a spin? It seems like nvidia's desktop arm driver is only 32-bit only at the moment, but I still wonder what would happen if you stuck a desktop geforce or radeon in that board and tried to use them with the oss drivers.

                      Comment

                      Working...
                      X