# Using Coral Edge TPU with VOXL

Source: https://forum.modalai.com/topic/1856/using-coral-edge-tpu-with-voxl
Category: Flight Core and Legacy VOXL (https://forum.modalai.com/category/9/flight-core-and-legacy-voxl)
Tags: voxl
Posted: 2023-02-20 16:48:57 UTC by Alex Z
Replies: 5 · Views: 1449

## Alex Z · 2023-02-20 16:48:57 UTC

Hi,

I'm interested in using a Coral Edge TPU with the VOXL—has anyone tried this?
It looks like the [Coral USB Accelerator](https://coral.ai/static/files/Coral-USB-Accelerator-datasheet.pdf) draws on average 500mA (peak 900mA) @ 5V, so it seems like it should be fine on J8? 
I'm guessing using M.2 form factor (with a USB-M.2 adaptor), particularly the [Dual Edge TPU](https://coral.ai/static/files/Coral-M2-Dual-EdgeTPU-datasheet.pdf), is going to be a problem, though?

## Reply by Vinny (ModalAI staff) · 2023-02-21 16:45:19 UTC

Hi @Alex-Z 
I can't speak to the SW side of things, but it's not easy to support Host mode on J8. We stopped selling cables that allow it to function as host.
With that said, if you want to try it, you should use any one of our plug-in boards since that USB interface on J13 is certainly a host port and can provide 500mA on Voxl. The peaks of 900mA may trip it though due to a USB2 rated power switch, so it's likely that exceeding 500mA will trip these switches on any Voxl1 class product.

## Reply by Eric Katzfey (ModalAI staff) · 2023-02-21 17:03:25 UTC

@Alex-Z We tried to get the Coral drivers working on VOXL but weren't able to get them working.

## Reply by dlee · 2023-05-25 00:43:12 UTC (in reply to Eric Katzfey)

@Eric-Katzfey How about VOXL2 Board? Can I use Coral Edge TPU in VOXL2 board?

## Reply by Eric Katzfey (ModalAI staff) · 2023-05-25 00:49:09 UTC (in reply to dlee)

@dlee Not sure, we haven't tried. The VOXL2 does so well by itself that we haven't seen the need for an extra accelerator :slightly_smiling_face:

## Reply by Chad Sweet (ModalAI staff) · 2023-05-25 01:11:46 UTC

Coral advertises 4 TOPS with tensorflow lite and VOXL 2 should achieve 15 TOPS with tensorflow lite.