Here is the instruction link, It should work on a fresh Jetpack 6.1 install without anything special needed: ominousindustries.com/blogs/ominous-industries/jetson-box64-install
this was great i subscribed, im a total noob with this stuff but i was interested in this very idea and this was a great demonstration and the case for the jetson was pretty awesome
I can image this box with remote desktop access into Windows launching and running Windows programs all under LLM voice control. Add a time management function and you could say something like "Resume that file i was working on last Monday, pull up the test results from xyz. Start a new folder with linked files for project ABC and call it ABC master". You could create the computer interface from the Enterprise. LOL. "computer, locate..."
That would be very cool and is essentially describing a voice controlled computer agent which is probably something that is closer to existing on a large scale than it is further away hahah.
I'm curious to know where in the future this would even be applicable. Do you know what advantages this would offer in comparison with a regular graphics card? If, in the future, these things could run better than GPUs, what advantages would you consider? Could this potentially replace a lot of computing components?
I think the best future scenario is that models become more efficent in terms of how much power they need to run, perhaps specific tiny models for certain tasks that can run on smaller devices like this that use far less power. I also believe unified memory devices like this: www.nvidia.com/en-us/project-digits/
Hi Ominous, Great video. Can you try and run ComfyUI, or ComfyUI on Stability Matrix on an Orin or Xavier Jetson. I've been trying in vain through various Docker Container without success. Can you try with Box64 and/or Wine on Orin and post a video. It would inspire me to try on my AGX Xavier.
Thanks very much. I don't know that ComfyUI will work right now, though it seems like it is slated to eventually support comfyui and flux on the orin nano based on this: www.jetson-ai-lab.com/tutorial_comfyui_flux.html
I see you respond to your comment section! Been following you for the Jetson content because im on ancient hardware 1060ti and need a solution that wont break my pockets My question is how upgradeable do you see the Jetson? I need more vram for AI and if i cant get that upgrade potential out of this device it may influence my choice. Its between this and a mini/used pc with 8gb vram and of course nivida gpu or this Super Nano
I try my best haha! As for your query, I believe a general approach is that for the best upgrade potential, it is best to go with a PC and full sized gpu. On small sbc's like this it won't be easy (or even possible) to upgrade them unless the manufacturers specifically come out with an upgraded plug and play module, which I don't think historically has happened or will in the future. From your description, I believe your best avenue right now would be to get a pc (even just using the one you have the 1060ti in currently) and swap that out for a 3060 12gb or something else like that. The 3060 12gb can be had for $250 right now which is an absolute steal for a world where Vram is still king for AI applications. It's not a power hungry card and won't break the bank either. Hopefully this was helpful!
I am going to put the digital files on etsy for cheap along with in-depth instructions as well as a parts list for everything so that people can 3D print their own!
Very glad to hear, thanks for mentioning this, I had checked which dependencies were installed on my fresh os and could have sworn cmake was there but this is good to know and I will check again! P.s., assuming that's a vector robot in your pfp, I love those things!
Here is the instruction link, It should work on a fresh Jetpack 6.1 install without anything special needed: ominousindustries.com/blogs/ominous-industries/jetson-box64-install
this was great i subscribed, im a total noob with this stuff but i was interested in this very idea and this was a great demonstration and the case for the jetson was pretty awesome
Thanks very much, I am glad it was helpful!
I can image this box with remote desktop access into Windows launching and running Windows programs all under LLM voice control. Add a time management function and you could say something like "Resume that file i was working on last Monday, pull up the test results from xyz. Start a new folder with linked files for project ABC and call it ABC master". You could create the computer interface from the Enterprise. LOL. "computer, locate..."
That would be very cool and is essentially describing a voice controlled computer agent which is probably something that is closer to existing on a large scale than it is further away hahah.
Looking forward to your review of Nvidia Project Digits in a few months :D
Loved watching 🎉🎉🎉
Thanks very much!
easy copy paste idea, simple to follow for anyone, thanks
Thanks very much! I put the link to the site in the pinned comment now as well!
I'm curious to know where in the future this would even be applicable. Do you know what advantages this would offer in comparison with a regular graphics card? If, in the future, these things could run better than GPUs, what advantages would you consider? Could this potentially replace a lot of computing components?
I think the best future scenario is that models become more efficent in terms of how much power they need to run, perhaps specific tiny models for certain tasks that can run on smaller devices like this that use far less power. I also believe unified memory devices like this: www.nvidia.com/en-us/project-digits/
Hi Ominous, Great video. Can you try and run ComfyUI, or ComfyUI on Stability Matrix on an Orin or Xavier Jetson. I've been trying in vain through various Docker Container without success. Can you try with Box64 and/or Wine on Orin and post a video. It would inspire me to try on my AGX Xavier.
Thanks very much. I don't know that ComfyUI will work right now, though it seems like it is slated to eventually support comfyui and flux on the orin nano based on this: www.jetson-ai-lab.com/tutorial_comfyui_flux.html
nice!!!
Thanks, I really appreciate it!
@@OminousIndustries Pumping the algorithm my friend!!!
I see you respond to your comment section! Been following you for the Jetson content because im on ancient hardware 1060ti and need a solution that wont break my pockets
My question is how upgradeable do you see the Jetson? I need more vram for AI and if i cant get that upgrade potential out of this device it may influence my choice.
Its between this and a mini/used pc with 8gb vram and of course nivida gpu or this Super Nano
I try my best haha! As for your query, I believe a general approach is that for the best upgrade potential, it is best to go with a PC and full sized gpu. On small sbc's like this it won't be easy (or even possible) to upgrade them unless the manufacturers specifically come out with an upgraded plug and play module, which I don't think historically has happened or will in the future.
From your description, I believe your best avenue right now would be to get a pc (even just using the one you have the 1060ti in currently) and swap that out for a 3060 12gb or something else like that. The 3060 12gb can be had for $250 right now which is an absolute steal for a world where Vram is still king for AI applications. It's not a power hungry card and won't break the bank either. Hopefully this was helpful!
where can I buy a similar unit like what you made.
I am going to put the digital files on etsy for cheap along with in-depth instructions as well as a parts list for everything so that people can 3D print their own!
Nvidia CES event starts in 20 mins!
Hell yeah, it was cool to see those releases
The Retro Arena project by techtoytinker
This is it, thank you!
haha i thought that was the power supply
Hahaha it is around the size of a desktop psu now that you mention it LOL
bro... please learn ctrl+c and ctrl+v
(great content otherwise)
Hahaha yeah I should have, that would have been a lot faster! Thanks!
Thanks, working great, but also needed to install cmake: sudo apt install cmake
Very glad to hear, thanks for mentioning this, I had checked which dependencies were installed on my fresh os and could have sworn cmake was there but this is good to know and I will check again! P.s., assuming that's a vector robot in your pfp, I love those things!