Google Coral Expansion Card

Introduction

The moment I saw the Framework laptop and the expansion card system I knew I have to make something. First I was thinking of making a raspberry pico card, ethernet, LTE or magnet charger, but since all of them were being done by the amazing community I thought I could not provide anything to the table.

However, while I was training an ML model on my laptop I was frustrated by how slow it was going. (I have HP Folio 9480m) A friend of mine saw me having a bad time and gave me a Google Coral USB Accelerator which sped up the training process substantially. After some time I went online to buy one for myself and saw on the website that the Accelerator Module was being sold.

Accelerator Module

I cannot put an image of the module (I am new user) so trust me when I say it is tiny but powerful

The module is a 120-pin LGA chip which scared me at first. 120 small lanes and all the supporting components can’t fit into an expansion card, can they? It turns out this chip has 104 pins that are either 3.3V or GND. From the other 16 pins, you can run the module using only 7 pins. So yeah… it’s not only doable but with the example schematics provided by Coral and Framework, it is 15 min work.

The PCB

I present you the Framework Coral Accelerator Card:

The board is using USB2.0 since it is the easiest to make into a functioning circuit but it has PCIe and USB3.0 possibilities for that:

I am a new user and cant show the picture but basically, I have to contact Google’s sales team.

Limitations

  • The Chip Shortage == No modules available
  • Applied Computer Science Degree == I have basic circuit design skills

EasyEDA Project on Github

NOTE: I do not have a Framework YET, but being a broke student will not stop me from developing cards for this amazing laptop.

21 Likes

Love to see something developed from this!

1 Like

This is awesome. A TPU Expansion Card has been on our wishlist of interesting projects, so it is great to see one!

7 Likes

I didn’t even know these accessories existed before now, but that’s amazing. Also, welcome to the forum. :slight_smile:

2 Likes

This is amazing! Keep us updated if you ever get to test it.

1 Like

@Cheese if I ever manage to get the compute module I will definitely show the progress.

2 Likes

Oh my god, you are a genius. I was just thinking of this idea because I ordered a new Framework and had doubts about Neural Networks but seeing this is already created is great.
Can’t wait to try this as soon as my Framework arrives and Google Coral cards are available again

1 Like

This is probably the coolest idea I’ve seen. Good luck with the project.

1 Like