NVIDIA has revealed that Chinese internet giant Tencent will deploy Tesla GPUs in its public cloud infrastructure. The GPUs are being deployed to help enterprise customers develop machine learning services.
Based on revenue, Tencent is China’s largest internet company, offering a variety of social media, e-commerce, and gaming services for the world’s fastest growing online market. The company took in $15 billion in revenue in 2015, trailing only Amazon ($107B), Google ($75B), and Facebook ($18B). Tencent’s market value that year was $206 billion.
The company’s public cloud is aimed at online businesses and other enterprise customers that have turned to the utility model for their computing infrastructure. The Tesla GPUs – including both the P100 and P40 – will be made available to these customers via the Tencent cloud, who are building machine learning services, either for their own customer base or for internal use. These include such things as natural language processing, facial recognition, automated customer service, and supply chain logistics.
"Tencent Cloud GPU offerings with NVIDIA's deep learning platform will help companies in China rapidly integrate AI capabilities into their products and services," said Sam Xie, vice president of Tencent Cloud, in prepared statement. "Our customers will gain greater computing flexibility and power, giving them a powerful competitive advantage.
Tesla GPUs have become the de facto standard for machine learning workloads for hyperscale companies. The P100, in particular, is NVIDIA’s most powerful neural network training processor to date, and is being deployed in both AI-oriented supercomputers and public cloud infrastructure. It’s general-purpose enough to handle both machine learning workload and more traditional HPC and data analytics applications.
The P40 is more precisely targeted to the machine learning space, and is specifically aimed at the inferencing side of those applications. In December 2016, Tencent had launched servers with M40 GPUs, the Maxwell-generation version of the P40.
Tencent, by the way, has also deployed IBM Power servers in its hyperscale datacenters for “big data” workloads. Theoretically, Tencent could deploy the NVIDIA Tesla GPUs in such servers, which would leverage the built-in NVLink support included in the latest Power processors. This would accelerate data communications between the host CPU (Power8) and the GPU (the P100), and offer Tencent’s enterprise customers a premier platform for machine learning.
More likely though, the GPUs are being hooked up to Intel Xeon processors in a more traditional server setup. According to NVIDIA, the Tencent cloud servers will incorporate up to eight GPUs in order to maximize performance for these machine learning workloads. Deployment is scheduled for the first half of this year.