CHINA TOPIX

11/24/2024 04:11:39 am

Make CT Your Homepage

Google Taps AMD GPUs to Help Build AI, Cloud Network

Google

(Photo : Getty Images) The new Google logo is displayed at the Google headquarters on September 2, 2015 in Mountain View, California.

Chip manufacturer AMD is finally throwing its hat into the deep learning platform as it announced its partnership with tech giant Google.

Under the agreement, the chipmaker will provide graphics processing units to Google's growing data center infrastructure. The goal is to use the immense computing power of GPUs to accelerate Google's deep learning technology.

Like Us on Facebook

GPUs are typically used to generate high-end gaming graphics. However, these specific hardware have been used lately in the booming deep learning platform. Experts believe that GPUs are more powerful than CPUs in terms of "parallel computing," a technique that requires the hardware to perform multiple calculations simultaneously. Because of this technology, GPUs are much faster in running deep learning neural networks compared with CPUs.

According to Extreme Tech, Google is planning to use the AMD Radeon FirePro S9300 X2 GPUs on its infrastructure. The search engine giant is planning to roll out AMD's hardware in 2017. Google's will be using AMD GPUs on its cloud platform, Google Cloud Machine Learning service, and the Google Compute Engine.

In a statement acquired by Forbes, AMD senior vice president and chief architect Raja Koduri said, "Google is building up its GPU-based infrastructure, and they want to ensure they offer AMD's architecture. Nobody has heard what AMD is doing in deep learning. This is a major first step for us."

The recent deal with Google will have a tremendous impact on AMD's business, which has started to dwindle in the face of its rivals, Intel and Nvidia. However, AMD was able to regain a portion of its lost market share recently with the release of some of the Radeon RX line of graphics card, which is intended towards the entry-level gaming market as oppose to the more lucrative yet very volatile enthusiast market.

Real Time Analytics