-1

I'd like to buy a computer for deep learning experiments.

I found various articles, which proposed a PC system with an NVIDIA RTX 3060 card.

Recently, I discovered Jetson Orin NX Systems, which seem to consume much less power as ordinary PCs.

Especially, I found the JETSON ORIN 4012 System with 16 GByte of graphics RAM. A seller claims 100 TOP AI performance [for around 1000 EUR].

I found various pages in NVIDIA's web, which explain how to install ready-made models.

While inferencing Performance is certainly of interest, I'd like to do deep learning training.

On no NIVIDIA page, I found hints regarding training performance with Jetson Orin 4012.

At the other side, I read, that Coral's USB accelerator isn't suitable for training by design. As far as I understand, this means, that the USB accelerator doesn't speed up much versus conventional CPUs performance.

Q: Would a JETSON ORIN 4012 be useful for deep learning training [useful = generate significant speedup]?

SteAp
  • 99
  • 2

1 Answers1

1

Both are two class of computing devices, It depends on - usage, performance ,form factor and cost

USB Accelerator is co - processor,TPU - Tensor Processing Unit primarily used for inferencing . We can build a compact Inferencing device when combined with a raspberry but cannot be used independently.

And also , it may not have computations capabilities for training, not sure how much RAM it has got on it, to train a model you need more computing resources on the other had Jetson Orin is complete desktop with Powerful Model Training Capabilities and inferencing too.

Hope the above info helps in making decision between USB Accelerator vs Jetson Orin

Biku
  • 11
  • 2