phonlamaiphoto - stock.adobe.com
GPU-buffed servers advance Cisco's AI agenda
Cisco's latest UCS servers with Nvidia's GPUs aim to help larger IT shops more quickly run AI and machine learning workloads -- and help Cisco keep pace with Dell and HPE.
Cisco Systems is the latest hardware vendor to offer gear tuned for AI and machine learning-based workloads.
Competition to support AI and machine workloads continues to heat up. Earlier this year archrivals Dell Technologies Inc., Hewlett Packard Enterprise and IBM rolled out servers designed to optimize performance of AI and machine learning workloads. Many smaller vendors are chasing this market as well.
"This is going to be a highly competitive field going forward with everyone having their own solution," said Jean Bozman, vice president and principal analyst at Hurwitz & Associates. "IT organizations will have to figure out, with the help of third-party organizations, how to best take advantage of these new technologies."
Cisco AI plan taps Nvidia GPUs
The Cisco UCS C480 ML M5 rack server, the company's first tuned to run AI workloads, contains Nvidia Tesla V100 Tensor Core GPUs and NVLink to boost performance, and works with neural networks and large data sets to train computers to carry out complex tasks, according to the company. It works with Cisco Intersight, introduced last year, which allows IT professionals to automate policies and operations across their infrastructure from the cloud.
This Cisco AI server will ship sometime during this year's fourth quarter. Cisco Services will offer technical support for a range of AI and machine learning capabilities.
Cisco intends to target several different industries with the new system. Financial services companies can use it to detect fraud and algorithmic trading, while healthcare companies can enlist it to deliver insights and diagnostics, improve medical image classification and speed drug discovery and research.
Server hardware makers place bets on AI
The market for AI and machine learning, particularly the former, represents a rich opportunity for systems vendors over the next year or two. Only 4% of CIOs said they have implemented AI projects, according to a Gartner study earlier this year. However, some 46% have blueprints in place to implement such projects, and many of them have kicked off pilot programs.
Jean Bozmanvice president and principal analyst, Hurwitz & Associates
AI and machine learning offers IT shops more efficient ways to address complex issues, but will significantly affect their underlying infrastructure and processes. Larger IT shops must heavily invest in training and the education of existing employees in how to use the technologies, the Gartner report stated. They also must upgrade existing infrastructure before they deploy production-ready AI and machine learning workloads. Enterprises will need to retool infrastructure to find ways to more efficiently handle data.
"All vendors will have the same story about data being your most valuable asset and how they can handle it efficiently," Bozman said. "But to get at [the data] you first have to break down the data silos, label the data to get at it efficiently, and add data protection."
Only after this prep work can IT shops take full advantage of AI-powered hardware-software tools.
"No matter how easy some of these vendors say it is to implement their integrated solutions, IT [shops] have more than a little homework to do to make it all work," one industry analyst said. "Then you are ready to get the best results from any AI-based data analytics."