Nvidia on Tuesday announced a new server platform, the HGX-2, designed to meet the needs of the growing number of applications that seek to leverage both high-performance computing (HPC) and artificial intelligence. The platform, Nvidia says, is the first to offer high-precision computing capabilities to handle both HPC and AI workloads. It uses FP64 and FP32 for scientific computing and simulations, while enabling FP16 and Int8 for AI training and inference. The HGX-2 server platform consists of a pair of baseboards. The 16 GPUs are fully connected through 12 NVSwitches to collectively deliver two petaflops of AI performance.
Seeking to embrace the modern developer building applications for the cloud, Oracle on Monday is announcing early adopter availability of the Oracle Container Native Application Development Platform. Three new services comprise the platform: Oracle Container Engine, a managed Kubernetes service to create and manage Kubernetes clusters; Oracle Container Registry Service, a private container registry service for storing and sharing container images across multiple deployments; and Oracle Container Pipelines, a full container lifecycle management service. The services can be consumed together or separately. Unveiled at the annual OpenWorld conference, the platform takes advantage of the extensive cloud infrastructure that Oracle introduced at last year's conference. "Containers and container native applications are one of the killer apps to use on top of bare metal," Bob Quillin, VP of the Oracle Container Group, said to ZDNet.
Huawei has released its Atlas computing platform, which is powered by its new Ascend series of artificial intelligence chips. It announced the new devices during Huawei Connect 2018, its annual conference in Shanghai. "The Atlas intelligent computing platform integrates various forms of products, such as modules, cards, boards, edge stations, and appliances, to build an all-scenario AI infrastructure covering the end, edge, and cloud," Huawei said. Huawei also said that the new Atlas platform is an evolution of the cloud hardware platform it released last year, and consists of several devices. These include the Atlas 200 AI accelerator module aimed at terminals, the Atlas 300 AI accelerator card designed for data centre use, the Atlas 500 AI edge station, and the Atlas 800 AI appliance, positioned for enterprises.
While infrastructure-as-a-service has existed for more than a decade, the market for cloud computing still has substantial room for growth: Worldwide spending on public cloud services will reach $186.4 billion this year, according to Gartner, yet remains just a fraction of overall IT spending. Recognizing this, Google on Tuesday unveiled the Cloud Services Platform, an integrated family of cloud services designed for organizations with workloads that remain on premise. The platform made its debut on the first day of the Google Next cloud conference in San Francisco. "CIOs... tell me they now realize they're going to be shutting down their data centers," Google Cloud CEO Diane Greene said in the Next Day One keynote address. However, "looking at their workloads, there's just a tiny, tiny fraction that are in the cloud.
With Baidu's existing capacity in internet data scraping and artificial intelligence, the platform says that, by comparing images that are circulated over the internet with data stored in a traceable blockchain, it would be able to substantiate allegations of intellectual property infringement. Currently, according to the new site, several traditional stock photo services have also moved onto the platform, including notable services such as Visual China Group, a local partner of the stock photo giant Getty Images. While it remains unclear whether the platform is being built on a public blockchain or a private one, it marks yet another effort by Baidu, often considered the "Chinese Google," in its push for blockchain adoption. The new launch comes just months after the firm introduced its blockchain-as-a-service platform in January, following which it launched a CryptoKitties knock-off dubbed Laici Go, as reported by CoinDesk. This isn't the first time that an established technology firm has looked to blockchain to solve the intellectual property issue for digital media assets.