Can't install SDK Manager

I’m sorry if this issue has already been answered i’m new to the Jetson world and i can’t find any answer that works for me.

I’m trying to install the SDK Manager on my Jetson TX1, but my screen was too small and after some research, i stumbled upon this Thread and i downgraded my SDK to the 0.9.12 version. But i can’t valid Step 1 because my Host PC (Ubuntu 18.04) can’t detect my Jetson, even though i can ping and use it with ssh -X. So i am not sure if the issue is from the connection or the SDK version.

I don’t know if this is related to my issue, but my Jetson has 2 red lights (One is “CVM Powered”! and other one is “PCIE/SATA Powered up”) and my fan doesn’t start at all, very similar to this Thread. But my screen works fine.

If anyone has any advice or knowledge about this, please let me know.
Thanks

I’m not sure if the older releases will be able to authenticate, but if it works, use it. There were some recent changes in SDK Manager authentication.

The LEDs are correct for a working system. The fan does not normally run unless the system is under enough load to generate heat.

If the TX1 is not in recovery mode with the micro-B USB connected, then the TX1 will not be detected and you won’t be able to flash. When it comes to installing options, e.g., CUDA, then the TX1 must be fully booted (in which case the ethernet ability to connect via ssh indicates SDKM should be able to add CUDA or other options). Often the reason for a recovery mode Jetson not being detected is due to using a VM. Is the host PC a native install?

I have a 10" laptop with Ubuntu for when I need to use SDKM, and this too is too small. What I do is running with forwarding via ssh from my desktop Fedora. The SDKM is running on the laptop, but displaying to Fedora, where I have a full sized monitor.

If it were QT I know how to set scaling, but I’m not sure on SDKM. I know other people here figured out scaling of SDKM, but I lack notes on that. Anyone able to suggest scaling of the SDK Manager GUI?

Okay i finally succeeded to install the SDKM, that’s totally on me, i was expecting the recovery mode to display some information, but it only show black screen.
I did it with the xrandr command to reduce the window on the SDKM v1.3.
Thanks for your answer @linuxdev

Okay, but it never runs even if i work with the Jetson the whole day.

This is still expected unless the system is under a significant load. If you want to see it with max performance mode forced, then the fan will likely run:

sudo nvpmodel -m 0
sudo jetson_clocks

Normally the fan avoids a power drain unless cooling is needed. Jetsons just don’t generate much waste heat.

1 Like

I don’t know if i need to open a new topic, but my issue is similar to this topic.

I want to have an IDE to build projects inspired by the Nvidia demos. From the NsightEE i can’t have access to any demos, is it because my PC can’t have CUDA Toolchain/Toolkit ? If so, what can i do to workaround this problem ? Do i need to use a VM ? If there is a easier way to develop directly onto my TX1 i can get along with it.

My GPU :

VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Seymour [Radeon HD 6400M/7400M Series] (prog-if 00 [VGA controller])
Subsystem: ASUSTeK Computer Inc. Radeon HD 6470M

if i’m correct, i also can’t cross-compile into my Jetson because of the CUDA Toolchain.

Thanks.

I am not very good in this area, but from what I know, you could still compile the code for the Jetson even if you cannot compile or run the code for your PC (assuming you are cross compiling from the host PC). By default I believe it is set up to assume you have Ubuntu 18.04 and a semi-recent NVIDIA GPU on your host PC. Some of the steps for installing tools on the host PC could fail without the NVIDIA GPU, not sure, and this would in turn break compile steps which follow the failure. If done manually, then there might be some way around this, but I do not have an answer for that.

CUDA has basically two parts: First, user space libraries and headers for development; second, actual runtime support. Installing runtime support would be impossible since you cannot add a driver to run on the AMD GPU. Other content could be added so long as the packages used don’t have a dependency on the runtime content.

Anyone else, can you suggest how to cross compile and install CUDA development code when the host PC GPU is not NVIDIA, but the only intent is to build on host and execute on a Jetson?