Thanks but I'm still not sure what they mean by tensor processing. Any ideas?
If you have a discreet graphics card it is referred to as a GPU (Graphics Processing Unit) and lives on the PCI-e bus of the motherboard. NVidia Tesla's and Intel XEON PHI's are considered co-processors. Basically, you slam a shitload of CPU cores on a gaming card form factor and attach it to the fastest bus on the motherboard. (The PCI Express slot.) It's a way to increase your computing power, but, they generate heat. Think of heat as inefficiency - wasted computing power.
Co-processors can do basically anything the CPU can.
Now, what Google did, was to take ASICs, (Application Specific Integrated Circuit) which differ from the standard CPU cores in that in a full custom ASIC, you have total control over the size of every transistor forming every logic gate, so you can "fine tune" each gate for optimum performance. A full custom ASIC performs electronic operations as fast as is possible, providing that the circuit design is efficiently architected. They are also extremely power efficient.
The new Tensor Processing Units, are specifically designed to do only tensor calculations, which means they can fit more transistors on the chip that do only one thing, and do it extremely well--achieving higher efficiency than other types of chips.
ASICs are very well suited for specific purposes. In fact, if you are going to set up a Crypto-Currency mining rig, ASICs are the way to go. But I digress...
Tensor calculations are the only thing the TSU does. It won't accelerate your graphics or give you more frames per second. It is a pure compute module. But, if you are using it in the context of what was described in the article, highly complicated tensor calculations (for deep machine learning), the calculations can be accomplished exponentially quicker and at a fraction of the power draw of a GPU, Field Programmable Gate Arrays, or Vision Processing Units.
Did that help, or did I just muddy the waters ?
Try this... http://www.techradar.com/news/computing-components/processors/google-s-tensor-processing-unit-explained-this-is-what-the-future-of-computing-looks-like-1326915