Intel has now launched a USB device designed for students and enthusiasts to tinker with deep machine learning technology. This device is named the Movidius Neural Compute Stick and it is only $79. This is an upgrade from their previous Fathom NCS.
In 2016, Intel had acquired both the company’s Movidius and Fathom NCS to get involved with the trendy deep machine learning technology train. The new device is powered by the Myriad 2 vision processing unit, which can perform at 100 GFLOPs. The machine learning mechanism is based on a Caffe-based neural software.
Upgrades From The Fathom
In comparison to the Fathom, the Movidius has multiple features that make it so different including an acceleration mode. With the multi-stick mode, you can essentially have a cluster of Movidius devices to run the same network. Although there is no maximum for devices supported in the multi-stick mode, it has been officially tested to be stable at 8 sticks in conjunction.
One of the biggest hardware upgrades is that the RAM has been quadrupled to 4 GB of LPDDR3. This will allow for larger loads within the neural network as well as in increase in bandwidth with the upgrade in RAM technology.
The Movidius is also designed to be ultra-low power at only 1W of energy consumption. This gives it a very high performance to power consumption ratio. The low power consumption also allows it to run from just the USB port without needing an external power supply. This stick is compatible with both USB 1.1 and USB 2 ports but it has a huge decrease in power when using version 1.1.
It should also be noted that the Fathom was released in limited supply to only beta testers. Essentially, this makes Movidius the first product release and it is available to the general public.