• 1
  • 4
  • 5
  • 6(current)
  • 7
  • 8
  • 11
Nvidia Jetson Nano, superfast Kodi device?
#76
Can also be reproduced with 'Clannad After Story OPA - sample HEVC main10 1080p.mkv' from https://kodi.wiki/view/Samples

15. Hi10 Anime Samples (thanks @Morien)
Reply
#77
https://github.com/xbmc/xbmc/issues/17124

Someone reported similar symptom on Android, may or may not be related
Reply
#78
I have one of the new Jetsons (Xavier NX) coming in about a week so I will try your image on that. IIRC it is essentially the same architecture and L4T install.
Reply
#79
(2020-08-28, 00:12)FredB451 Wrote: I have one of the new Jetsons (Xavier NX) coming in about a week so I will try your image on that. IIRC it is essentially the same architecture and L4T install.
Speaking of new versions of the Jetson:
https://www.tomshardware.com/news/nvidia...b-launched

New 2GB version - same cpu / gpu it appears for $60 US.

I've skimmed through this thread and did not see any mentions of memory usage on the 4GB model. Any concern from those who have the original unit that the 2GB of memory is going to be too thin for an enjoyable experience?
Current: Nvidia Shield TV (2nd Gen)
Retired: Asus M004u Chromebox - LibreELEC 7.0.1 | Raspberry Pi 2 - OpenELEC 6.0.3 | Apple TV w/ BCM70015 - CrystalBuntu 2.0 | Raspberry Pi - Model B - OpenElec Kodi 15
Reply
#80
(2020-10-05, 18:13)xi_Slick_ix Wrote:
(2020-08-28, 00:12)FredB451 Wrote: I have one of the new Jetsons (Xavier NX) coming in about a week so I will try your image on that. IIRC it is essentially the same architecture and L4T install.
Speaking of new versions of the Jetson:
https://www.tomshardware.com/news/nvidia...b-launched

New 2GB version - same cpu / gpu it appears for $60 US.

I've skimmed through this thread and did not see any mentions of memory usage on the 4GB model. Any concern from those who have the original unit that the 2GB of memory is going to be too thin for an enjoyable experience?

Yep - 3.0A 5V USB Type-C for power rather than Micro B or Barrel jack (which may mean a Pi 4B PSU is a convenient power source?), no DisplayPort alongside the HDMI, only one CSI connector for cameras (rather than 2), no M2 slot (so WiFi has to be added using a USB dongle) and only one (rather than 4) USB 3.0 Type A port (with 2 x USB 2.0 Type A and 1 x Micro USB Type B). No PoE option.  Ubuntu version optimised for the 2GB model uses LXDE.

Most of the hardware changes (apart possibly from the M2 slot going for WiFi) will make few differences to those using the device for Kodi - I guess the 2GB vs 4GB memory reduction will be the one we need to know more about.
Reply
#81
Discussion about the new $54 budget NVIDIA Jetson Nano 2GB Developer Kit including detailed specifications and reference links -> https://forum.kodi.tv/showthread.php?tid=357522


A new budget NVIDIA Jetson Nano Developer Kit consisting of a 2GB single board computer has now been launched for only $54 US.

https://developer.nvidia.com/embedded/je...eloper-kit

It is primarily meant to be an affordable AI + robotics development, prototyping and education platform for developers but can it be used as a powerful Kodi media player on Linux?

Image

Specifications copied from https://www.cnx-software.com/2020/10/06/...loper-kit/
  • Jetson Nano 2GB CPU Module (which is similar but not exactly the same as slimmed down NVIDIA Tegra X1)
    • ARM Cortex-A57 MPCore Quad-Core 64-bit (AArch64 / ARM64) processor @ 1.43 GHz per core
      • Nvidia Maxwell GPU with 128 CUDA cores (with NVENC/NVDEC and CUDA capabilities)
        • System Memory  – 2 GB 64-bit LPDDR4 25.6 GB/s
          • Storage  – MicroSD card slot  (SD Card not included)
            • Video Decode 500MP/sec
              • 1x 4K @ 60 (HEVC)
                • 2x 4K @ 30 (HEVC)
                  • 4x 1080p @ 60 (HEVC)
                    • 8x 1080p @ 30 (HEVC)
                      • 9x 720p @ 60 (HEVC)
               
              • Video Encode @ 250MP/sec
                • 1x 4K @ 30 (HEVC)
                  • 2x 1080p @ 60 (HEVC)
                    • 4x 1080p @ 30 (HEVC)
                      • 4x 720p @ 60 (HEVC)
                        • 9x 720p @ 30 (HEVC)
                 
                •  Dimensions of only the Jetson Nano 2GB CPU Module – 70 x 45 mm
     
    • Jetson Nano Baseboard for the 2GB GPU module
      • Mechanical 260-pin edge SO-DIMM connector for Jetson Nano module.
        • Video Output – HDMI 2.0
          • Connectivity
            • Gigabit Ethernet (RJ-45 port)
              • Optional 802.11ac wireless USB adapter and extension cable sold by Nvidia
             
            • USB – 1x USB 3.0 ports, 2x USB 2.0 ports, 1x USB 2.0 Micro-B port for power or device mode
              • Camera I/F – 1x MIPI CSI-2 DPHY lanes compatible with Raspberry Pi Camera Module V2 and HQ camera
       
      • Expansion
        • 40-pin expansion header with GPIO, I2C, I2S, SPI, UART signals
          • 12-pin header for power and related signals and UART
            • Misc – Power LED, 4-pin fan header
              • 4-pin Fan header (Not initially available in all regions)
                • Power Supply – 5V/3A via USB-C
         
        • Operating-systems
          • Supports Ubuntu 18.04 LTS by default but all NVIDIA Jetson Nano devkits are already well supported throughout the industry and open source communities
            • Android and Android TV is not officially supported but inofficial third-party builds are available
           
          • Dimensions of baseboard – 100 x 80 x 29 mm (with heatsink)



Could this budget Nvidia SBC with its Nvidia Tegra type SoC and GPU be a Raspberry Pi 4 killer Kodi media player users and RetroPlayer gamers will want?

Basically a mini single board computer version of an Nvidia Shield TV but meant for Linux instead of Android TV operating-system.


Worldwide pre-orders are open on sites like Amazon, Seeed Studio and others, with shipping scheduled to start at the end of the month.

https://www.nvidia.com/en-us/autonomous-.../#shop-all

https://www.seeedstudio.com/NVIDIA-Jetso...-4706.html

https://www.amazon.com/dp/B08J157LHH/





A similar but beefier and slightly more costly NVIDIA Jetson Nano Developer kit with a 4GB RAM SBC that also has DisplayPort, more USB 3.0 ports, an M.2 Key E socket, PoE support, plus one additional CSI camera connector was introduced in March 2019 for $99 US:

https://forum.kodi.tv/showthread.php?tid=342261

https://www.cnx-software.com/2019/03/19/...loper-kit/

Official getting started guide:

https://developer.nvidia.com/embedded/le...2gb-devkit

https://www.nvidia.com/en-us/autonomous-...-projects/

https://developer.nvidia.com/embedded/jetson-benchmarks



CNXSoft Getting Started with NVIDIA Jetson Nano Devkit:

https://www.cnx-software.com/2019/12/07/...eo-stream/
Reply
#82
(2019-03-21, 02:28)davilla Wrote: Wow, this thing is smoking Smile They are going to sell millions of them. Maybe time to put my porting hat back on.

@davilla Did you look into porting Kodi (or a fork of Kodi like MrMC) with Kodi's VideoPlayer and NVDEC support for V4L2 to the Nvidia Jetson Nano hardware?

More of an interesting device now that they have a $54/$59 budget-version in the new NVIDIA Jetson Nano 2GB Developer Kit ->  https://forum.kodi.tv/showthread.php?tid=357522

See that there in an unmaintained fork here -> https://github.com/aliubimov/xbmc/tree/f...ra-support

PS: Kodi 19 Matrix have removed VDPAU on Linux support and there is not yet support for its replacement NVDEC/NVDECODE -> https://forum.kodi.tv/showthread.php?tid=325329
Reply
#83
(2020-07-23, 16:46)hashtag Wrote:
Code:

-DCORE_PLATFORM_NAME=x11 -DX11_RENDER_SYSTEM=gles -DENABLE_INTERNAL_FFMPEG=ON -DENABLE_INTERNAL_FLATBUFFERS=ON -DENABLE_INTERNAL_FMT=ON -DENABLE_V4L2=ON -DENABLE_NVBUFFER=ON 
Really glad I found this thread.
Do these build instructions still work after your recent works on kodi19?
I've tried your docker and 4k content played from my nas was silky smooth (as I mentioned in the recently closed GitHub issue that I commented)
Can't believe I completely missed link to this thread in another issue. Doh.
I'm happy to be another tester too, should you require (although I think I've stated shady that I'm not experienced with coding\compiling massively. But I'm happy to try)
Reply
#84
(2020-12-09, 00:45)wodgey Wrote:
(2020-07-23, 16:46)hashtag Wrote:
Code:

-DCORE_PLATFORM_NAME=x11 -DX11_RENDER_SYSTEM=gles -DENABLE_INTERNAL_FFMPEG=ON -DENABLE_INTERNAL_FLATBUFFERS=ON -DENABLE_INTERNAL_FMT=ON -DENABLE_V4L2=ON -DENABLE_NVBUFFER=ON 
Really glad I found this thread.
Do these build instructions still work after your recent works on kodi19?
I've tried your docker and 4k content played from my nas was silky smooth (as I mentioned in the recently closed GitHub issue that I commented)
Can't believe I completely missed link to this thread in another issue. Doh.
I'm happy to be another tester too, should you require (although I think I've stated shady that I'm not experienced with coding\compiling massively. But I'm happy to try)

There is build options, I use for docker container

Code:
-DCORE_PLATFORM_NAME=x11 -DAPP_RENDER_SYSTEM=gles -DENABLE_INTERNAL_FFMPEG=ON -DENABLE_INTERNAL_FLATBUFFERS=ON -DENABLE_INTERNAL_FMT=ON -DENABLE_V4L2=ON -DENABLE_NVBUFFER=ON -DENABLE_INTERNAL_SPDLOG=ON -DENABLE_VAAPI=OFF

If you want to skip compilation, you can just use pre-build docker container. I updated all of them to the latest version of code-base.
Instructions are on docker hub

There is two versions, that are currently maintained
* latest - V4L2 with NVDEC acceletation
* snapshot - same as latest, but with some additional experiments with DSD Audio support.
Reply
#85
Yay! It's compiling. Taking a long time though
Reply
#86
(2020-12-09, 14:24)wodgey Wrote: Yay! It's compiling. Taking a long time though
I forgot to do the -J(number fo cores on the nano) thing

Also .....

I have 12 x86_64 cores available to me and I did read that blurb within the build guide regarding parallel-computing/workload-distribution and would like to give that a try. would it be easy to setup?

I have 4 x i7 2620m cores in my w520 thinkpad, (running full-fat Linux Mint)
4 x i5 10300H cores in my HP laptop (currently only WSL2 install of Ubuntu 20.04. - I have only had it for 1 day)
and 4 intel atom cores in my little Acer Revo net-top box that I use for my mail/webserver
Reply
#87
I have tried to do cross-compile, but that takes even more time. My build server I use to compile source code is located on AWS ARM instance.
Reply
#88
(2020-12-09, 14:45)wodgey Wrote: I have 12 x86_64 cores available to me and I did read that blurb within the build guide regarding parallel-computing/workload-distribution and would like to give that a try. would it be easy to setup?
so I've found a little guide for "distcc" as posted below


60-second instructions:
1 For each machine, download distcc unpack, and do
./configure && make && sudo make install
2 On each of the servers, run distccd --daemon, with --allow options to restrict access.
3 Put the names of the servers in your environment:
export DISTCC_POTENTIAL_HOSTS='localhost red green blue'

4 Build!
Wrap your build command in the "pump" script and use "distcc" as your C compiler:
cd ~/work/myproject; pump make -j8 CC=distcc

I think this means (counter-intuitively) that the machines called "servers"  would actually be the slaved x86_64 machines listed above? Am I correct in that assumption?

Would I just then compile with Huh
Code:
cmake ../xbmc -DCORE_PLATFORM_NAME=x11 -DAPP_RENDER_SYSTEM=gles -DENABLE_INTERNAL_FFMPEG=ON -DENABLE_INTERNAL_FLATBUFFERS=ON -DENABLE_INTERNAL_FMT=ON -DENABLE_V4L2=ON -DENABLE_NVBUFFER=ON -DENABLE_INTERNAL_SPDLOG=ON -DENABLE_VAAPI=OFF

pump make -j16 CC=distcc
assuming all 4 arm cores and all 12 x86_64 cores

Now, as for the cross-compiling stuff. Do I just need to install these packages to my x86_64 machines?
Code:
sudo apt-get install gcc-aarch64-linux-gnu g++-aarch64-linux-gnu
and I'm good to go?
I will be initiating the compiling within the nano environment so I assume that I don;t need to install the toolchain/headers for x86 on the jetson itself.

thanks in advance for pointers/corrections/advice
Reply
#89
(2020-12-09, 17:31)hashtag Wrote: I have tried to do cross-compile, but that takes even more time. My build server I use to compile source code is located on AWS ARM instance.

Ah ok. That's useful to know.

Thanks.
Reply
#90
(2020-12-09, 17:40)wodgey Wrote:
(2020-12-09, 17:31)hashtag Wrote: I have tried to do cross-compile, but that takes even more time. My build server I use to compile source code is located on AWS ARM instance.

Ah ok. That's useful to know.

Thanks.

it's currently at 81% because I am only using 1 core. I hope I learn my lesson and remember to use more cores next time.

I will also install the ccache packages I saw in the guide, to help avoid it taking so much time, if I need to do this again
Reply
  • 1
  • 4
  • 5
  • 6(current)
  • 7
  • 8
  • 11

Logout Mark Read Team Forum Stats Members Help
Nvidia Jetson Nano, superfast Kodi device?0