You might not know or have heard about Movidius, the Silicon Valley startup, but you’ve surely heard about Myriad VPU, or Myriad 2, its one of a kind vision processing chip. You’ll find the Myriad VPU right at the center of Google’s groundbreaking project Tango tablet and smartphone.
But there more to the story with the release of the new highly efficient version, the Myriad2 VPU. Movidius aims to provide a higher quality visual experience and more powerful computational imaging applications which have only been seen from outdated computer based GPU solutions for mobile devices which have historically been very power hungry.
Designed with the purpose of replacing, or augmenting at the very least, custom silicon and mobile equivalents, the Myriad 2 being hailed as the next big thing to make the scene.
The Myriad 2 is forecast to be 20 times better than the Project Tango chip
If you have seen the already astounding Project Tango in action, and what it is capable of doing in real time with 3D vision and mapping, the Myriad 2 is something else. When you compare the previous Project Tango on a tablet, being aided by the first Myriad chip coupled with the power of a Nvidia K1 GPU, it pales in comparison.
Movidius boasts that the Myriad 2 chip provides 20 times greater efficiency when compared to the first version. To begin with, that should improve the possibility of shrinking the Tango into a smaller and more powerful device.
With the capacity to execute over two trillion 16-bit operations each second, the Myriad 2 VPU will do this while demanding as few as 500 milliwatts. With numbers like that, mammoth data throughput are a cakewalk.
Offering real-time processing for video and vision applications, the Myriad 2 is leaps and bounds ahead of the pack. Coupling that with 12 vision specific programmable vector processors, the Myriad 2 is more like a GPU with a unique graphical development environment capable of supporting both Open CL and C++. Having the capability to work either as standalone or coprocessors in dedicated devices, the future looks sharper than it ever has.
The quest of computational imaging; Myriad 2 to revolutionize photography
Remi El-Ouazzane, Movidius CEO, identifies the third genesis of photography within computational imaging, leveraging computational power to drastically enhance the art of photography by including features like quicker autofocus, the inclusion of depth maps, more lifelike panoramas, and the integration of High Dynamic Range processing.
In this, Remi El-Ouazzane takes up rank with other visionaries like Pelican’s Kartik Venkataraman and Lytro’s Ren Ng. As a point of difference from Pelican and Lytro, Movidius won’t actually be building the imaging devices but will instead embark on a venture to embed their patented Systems on Chips into a variety of Original Equipment Manufacturer’s solutions.
Movidius has stated that by using its Myriad 2 that it adds little more than $10 to the production costs of a mobile device. A market comparison can be found in the camera module of a top of the range smartphone, which costs an estimated $20, so by including a Vision Processing Unit it would raise the production cost of the camera by as much as 50%. Not a big increase dollar-wise, but certainly a substantial jump in costs on the percentage split.
Very much in a similar fashion to a GPU taken off the shelf, the flexibility is given to anyone needing to custom their own vision or computational photography solution in a mobile device, a compelling value proposition can be found in this VPU.
We doubt that the few extra dollars will hinder the expansion in market share for the Myriad 2, as this is not a high price to pay for a VPU of this quality. At the present time, Movidius has the sandpit all to themselves, leaving them with the next stage in the project where there will be a need to act on increasing both market and mind share before any sizable competitors attempt to move into the space.