51–60 of 218 posts
> 472 gigaflops Whenever I hear a number of gigaflops or terraflops, I like to look up the history of super computers [0]. This $99 computer is faster (on paper) than the world's fastest supercomputer in 1996, or a bit over 20 years ago. That's pretty cool. [0] https://en.wikipedia.org/wiki/History_of_supercomputing#Mass...
Cheaper than the $150 Google TPU Dev Board, and looks like it can do training as well as inference. Also, doesn't require you to send your model to their company. Nice!
Cheaper than the $150 Google TPU Dev Board, and looks like it can do training as well as inference. Training is a slight win (although it's going to be too slow for anything useful really). But it looks like the TPU will outperform this somewhat for inference. The 256 core Jetson (this has 128 cores) could run MobileNet-v2 at between 12 and 20 ms per image (depending on batch size)[1], while the USB TPU adapter takes…
Yes it does require you to send your model to Google: https://coral.withgoogle.com/web-compiler/
Cheaper than the $150 Google TPU Dev Board, and looks like it can do training as well as inference. Also, doesn't require you to send your model to their company. Nice!
Quad core arm, 4gb of ram and gigE! Seems like a really nice board
> 472 gigaflops Whenever I hear a number of gigaflops or terraflops, I like to look up the history of super computers [0]. This $99 computer is faster (on paper) than the world's fastest supercomputer in 1996, or a bit over 20 years ago. That's pretty cool. [0] https://en.wikipedia.org/wiki/History_of_supercomputing#Mass...
Earlier quoted context omitted.
Cheaper than the $150 Google TPU Dev Board, and looks like it can do training as well as inference. Training is a slight win (although it's going to be too slow for anything useful really). But it looks like the TPU will outperform this somewhat for inference. The 256 core Jetson (this has 128 cores) could run MobileNet-v2 at between 12 and 20 ms per image (depending on batch size)[1], while the USB TPU adapter takes…
Seems to lose in Mobilenet and win in other benches according to Nvidia's benchmarks. I'll wait for independent testing before I drop $100. https://devblogs.nvidia.com/jetson-nano-ai-computing/
OTOH, fp32 models are _much_ easier to work with, and this thing has more RAM so you can waste it on 32 bit weights, and NVIDIA's software toolkit is second to none. So the Jetson looks pretty tempting as well. I just wish they didn't try to insult my intelligence.
Earlier quoted context omitted.
This. And all of the third party camera solutions for TX2 cost $500+ when equivalent USB cameras with the same sensor cost $50. They really need to get their act together and sell some NVIDIA-sanctioned camera solutions at scale, and at price points similar to Raspberry Pi cameras. A lot of third party carrier boards also have a complete sh_tshow of connectors. Auvidea's boards, for example, ship with a Raspberry Pi…
They've actually fixed this hopefully this time around. According to their blog post it actually has driver support for the RPi CM2 8MP (IMX219) and they'll be releasing their own Nvidia-sanctioned cameras available from their partners. It should hopefully just work. No lowlight options at this time however, which means external CCTV is out of the question :(
Cheaper than the $150 Google TPU Dev Board, and looks like it can do training as well as inference. Also, doesn't require you to send your model to their company. Nice!
there is no reason to train on that thing (except for zero shot classification demos). any cheap/lower end Nvidia gpu would do a much better job, and you would then transfer the model to the embedded thing.
why would you want to train on the device itself?
Earlier quoted context omitted.
there is no reason to train on that thing (except for zero shot classification demos). any cheap/lower end Nvidia gpu would do a much better job, and you would then transfer the model to the embedded thing.
Would it do better than a CPU for training? I do my dev on a MacBook Air and use AWS for training, if this cheap gpu will be a few times faster than my Air CPU than I’d be willing to get it. I usually work with medium sized models/data, too big for CPU but don’t need a multi gpu cluster.
I have a Jetson TX2, at the $99, I'm super tempted to buy one to see how it compares. That being said, I would definitely buy this if there is a way to make Plex server work with transcoding.