Computer processing units come in many different flavors, including CPU, GPU, EPU and the FPGA. However, Google claims to have created what they are calling the Tensor Processing Unit, which is supposedly seven years more advanced than anything else on the market.
The processing unit is being called such because it supposedly underpins what is called the TensorFlow, which is the software engine that allows for deep learning services. Google claims that the machine that defeated world Go champion Lee Sedol at his own game was powered by the TPU.
In terms of applications, the TPU is considerably more capable of allowing for deep learning which can be used for everything from photo recognition to sending out automatic email responses. This makes the technology incredibly important for social media sites and search engines in upgrading their capabilities to serve users.
Google CEO Sundar Pichai didn't provide that much insight into the technology during the presentation at Google I/O conference. In a recent blog by Google, though, the tech giant went into a few more details as to what exactly their TPU is capable of.
"We've been running TPUs inside our data centers for more than a year, and have found them to deliver an order of magnitude better-optimized performance per watt for machine learning," the blog said. "This is roughly equivalent to fast-forwarding technology about seven years into the future (three generations of Moore's Law)."
Among the benefits of using the TPU is that it does not require as many transistors when performing computations. As a result, it can provide significantly better results even when using hardware that is not that powerful.
For those with vastly superior equipment, however, the TPU can allow for substantially more processing power. This can be used to have more sophisticated machine learning models and to get more precise results.
