Finally, "real" GPUs on the Raspberry Pi 5

feiticeir0 said:

Why ? Why run an external GPU ?

I think if you're going at something "from scratch" it can be a valid question to ask "Why?", if you're going "but I already have these components lying around" or "I'm technically curious" then it can make sense. It also furthers support for hardware in the arm linux infrastructure that may not have been there before and that could open up possibilities.

feiticeir0 said:

I have the Jetson Nano developer KIT and it's not the most up to date stuff. Python is 3.6 and because of that I'm limited on the TensorFlow, Pytorch versions that I can install.

That's cool, I'd be interested in reading more about how people put it to use, I often see the end result or reference manuals, but not many blog examples of people actually using it on the element14 Community.

As soon as I'm done with it, I can post it here.It's nothing fancy really - just read my license plate and open my garage door - this will be used - at least for a proof of concept - on my work place .

First I'm just learning how to use the NVIDIA specific functions for deep learning and ML.

NVIDIA has quite a number of tutorials for one to learn. They even offer free certification after you submit a project and be graded by them. 

cstanton said:

That's cool, I'd be interested in reading more about how people put it to use, I often see the end result or reference manuals, but not many blog examples of people actually using it.

hackster.io has many projects.

Since it's a Linux ( It runs Ubuntu 18.04 LTS) you can install TensorFlow and Pytorch and run pretty much anything that don't requires newer versions.

Newer Jetsons (and more expensive  ) run more up to date versions. 

cstanton said:

I think if you're going at something "from scratch" it can be a valid question to ask "Why?", if you're going "but I already have these components lying around" or "I'm technically curious" then it can make sense. It also furthers support for hardware in the arm linux infrastructure that may not have been there before and that could open up possibilities.

That's an really good example (and I missed it on my original reply) of - "because I can" ! 

I remember one episode of the Big Bang Theory (really really excellent series) when they guys are turning the living room lights on and off (and the stereo) using Internet by bouncing the signal through several countries and Penny asks them why and they reply - "Because we can" .

Edit media

Dimensions x Small Medium Large Custom

Subject (required) Brief Description Tags (separated by comma) Video visibility in search results Visible Hidden

Parent content

RE: Finally, "real" GPUs on the Raspberry Pi 5

Poster

Upload Preview

Ok

Cancel

Yay! Hardware decode for h.264 and VP9

scottiebabe said:

Hardware decode for h.264 and VP9

And encode, that's a good point, depending on the graphics card you could even have hardware encode/decode for AV1.

These are valid points we are talking about the Pi here, there's going to be compromises in these non optimal conditions

feiticeir0 said:

you can install TensorFlow and Pytorch and run pretty much anything that don't requires newer versions.

I've been meaning to get my head around these myself, do you know of any beginner resources?

Gough Lui said:

I've seen working with them are mostly focusing on the software, rather than the electronics side. Everything to them is code ... and this isn't the "forum" they would naturally gravitate towards.

I feel like that needs to be improved on this Community but I may be in the minority thinking that.

I've sent you a message.

I wonder how Intel ARC support is on arm64 Linux ... *chuckle

- Gough