Excuse me, maybe I am misinterpreting something but are you saying that this stick is twice as costeffective as the 1080 but the stick may still be useful for embedded?
"Like other players in the edge inference market, Movidius is looking to promote and capitalize on the need for low-power but capable inference processors for stand-alone devices."
Sounds like this is targeted at inference only, given the specs, that seems like a reasonable conclusion, right?
4 comments
[ 1.7 ms ] story [ 24.8 ms ] threadFor comparison:
Tesla P40 = 11,758 GFLOPS at 250 W ($5,600).
GeForce GTX 1080 = 8,873 GFLOPS at 180W ($550).
The stick may be still useful for embedded/in the field/robotics apps?
"Like other players in the edge inference market, Movidius is looking to promote and capitalize on the need for low-power but capable inference processors for stand-alone devices."
Sounds like this is targeted at inference only, given the specs, that seems like a reasonable conclusion, right?