Covering Scientific & Technical AI | Thursday, November 28, 2024

Quanta Cloud Raises Server Ante with GPUs 

Quanta Cloud Technology, the server and storage unit of the Taiwanese manufacturer Quanta Computer, has over the last year been targeting the high end of the enterprise datacenter market. The cloud vendor showed up at this week's Nvidia GPU technology conference with a trio of new server and storage products accelerated with graphics processors.

QCT claims to be among the first to develop x86-based servers that leverage Nvidia's NVLink high-speed interconnect technology. In terms of server performance, the partner said NVLink allows GPUs and CPUs to share data up to 12 times faster than conventional server designs.

The San Jose-based Quanta cloud unit also considers itself a pioneer in virtual desktop infrastructure on 3-D graphics. Hence, it said this week it is working with Nvidia on a hyper-converged "VDIBlock 3D Advance" approach that integrates compute and storage nodes with Nvidia GPUs. The design also leverages VMware's Horizon virtualized desktop and application suite.

QCT unveiled two GPU-accelerated servers at the Nvidia event: the QuantaGrid D51BV, a 2U server based on Intel's Xeon E5-2600 processor and DDR4 memory technology that can support up to a pair of GPU accelerators; and the Stratos S210, high-density server designed to support up to four GPUs in a 2U chassis.

Meanwhile, QCT said it would work with Nvidia on developing next-generation x86 servers around Nvidia's Tesla accelerated computing platform. Nvidia is targeting Tesla at data analytics, "deep learning" and supercomputing applications.

In a statement, QCT general manager Mike Yang said collaboration on bringing souped-up computing to enterprise applications and cloud datacenters via GPU-accelerated servers would be extended to supercomputing and analytics applications.

Nvidia's Sumit Gupta added that QCT has helped extend the benefits of faster data transfer rates between GPUs and CPUs to the x86-based server market.

QCT also unveiled a new QauntaPlex storage product at the Nvidia event. The 4U, two-node storage server includes two dual-socket server nodes and up to 70 top-loaded hot-swappable 3.5-inch HDD bays.

Quanta has been a leading proponent of the Open Compute Project, which was started by Facebook in April 2011 to open source the social network’s own server, storage, rack and datacenter designs. In October 2013, Quanta QCT started selling OCP-compliant machinery to customers in the United States and Canada.

Quanta continues to take on server leaders like Dell, Hewlett-Packard, IBM and others as it targets enterprise datacenters with its GPU-accelerated servers. Meanwhile, marker leader HP also rolled out new storage and other datacenter offerings designed to bring scalable, high-performance infrastructure to mid-size enterprises. Earlier this week, HP unveiled a hybrid flash array incorporating solid-state drives and adaptive tiering into a virtual storage system along with a SSD read-caching system.

As HP targets medium-sized companied, Quanta remains focused on the high end. "As for enterprises, we will be focusing on forward-thinking enterprise customers that are adopting private or hybrid clouds. This is something that is new to us, but we would like to create more in this space," QCT's Yang told EnterpriseTech last year.

About the author: George Leopold

George Leopold has written about science and technology for more than 30 years, focusing on electronics and aerospace technology. He previously served as executive editor of Electronic Engineering Times. Leopold is the author of "Calculated Risk: The Supersonic Life and Times of Gus Grissom" (Purdue University Press, 2016).

AIwire