NVIDIA Tesla P100 Comes To Google Compute Engine In The Cloud
Today the NVIDIA Tesla P100 was introduced for the public in the Google Cloud Platform for cloud computing. These NVIDIA P100 GPUs were launched first in beta mode, while at the same time the NVIDIA K80 GPU was made available on the Google Compute Engine. Both the K80 and the P100 are available in US West (Oregon), US East (South Carolina), Europe West (Belgium), and Asia East (Taiwan) immediately if not soon.
The Google Compute Engine lives inside the Google Cloud Platform. Inside the cloud, the Google Computer Engine provides scalable virtual machines for computing needs of all sorts. The Compute Engine is just one of a wide variety of cloud-based products and solutions.
The NVIDIA Tesla P100 rolls with Pascal GPU architecture and are employed for their flexibility, low cost, and top-notch performance. The difference between the P100 and the K80 is the performance boost and the cost – the P100 has approximately 10x faster performance than the K80 for a number of tasks. The P100 for PCIe "enables a single node to replace half a rack of commodity CPU nodes by delivering lightning-fast performance in a broad range of HPC applications.
"For certain tasks, [NVIDIA] GPUs are a cost-effective and high-performance alternative to traditional CPUs. They work great with Shazam's core music recognition workload, in which we match snippets of user-recorded audio fingerprints against our catalog of over 40 million songs," said Ben Belchak, Head of Site Reliability Engineering, Shazam. "We do that by taking the audio signatures of each and every song, compiling them into a custom database format and loading them into GPU memory. Whenever a user Shazams a song, our algorithm uses GPUs to search that database until it finds a match. This happens successfully over 20 million times per day."
Each of the four regions where these GPUs are available for cloud computing, Google's offering sustained use discounts. These discount rates can be found in the chart below. Not that these discounts work with all machine types, but do not apply to "preemptible instances."
In all cases, users only pay for the number of minutes that they use. Google's Compute Engine automatically gives the lowered price rate when the threshold is met. Google suggested today that "No lock-in or upfront minimum fee commitments are needed to take advantage of these discounts."