VentureBeat presents: AI Unleashed - An exclusive executive event for enterprise data leaders. Network and learn with industry peers. Learn More
At GTC 2021, Nvidia this morning took the wraps off of the BlueField-3 data processing unit (DPU), the latest in its lineup of datacenter machines built for AI and analytics workloads. BlueField-3 packs software-defined networking, storage, and cybersecurity acceleration capabilities, offering what Nvidia claims is the equivalent of up to 300 CPU cores of horsepower — or 1.5 TOPs.
As of 2019, the adoption rate of big data analytics stood at 52.5% among organizations, with a further 38% intending to use the technology in the future, according to Statista. The advantages are obvious. A 2019 survey by Enterprenuer.com found that enterprises implementing big data analytics have seen a profit increase of 8% to 10%.
Nvidia’s BlueField-3 DPUs features 300GbE/NDR interconnects and can deliver up to 10 times the compute of the previous-generation BlueField-2 DPUs, with 22 billion transistors, while isolating apps from the control and management plane. The 16 ARM A78 cores inside can manage 4 times the cryptography performance, and BlueField-3 is the first DPU to support fifth-generation PCIe and time-synchronized datacenter acceleration.
Event
AI Unleashed
An exclusive invite-only evening of insights and networking, designed for senior enterprise executives overseeing data stacks and strategies.
BlueField-3 can additionally act as a monitoring agent for Morpheus, Nvidia’s AI-enabled cloud cybersecurity platform that was also announced today. Moreover, it takes advantage of DOCA, the company’s datacenter-on-a-chip architecture for building software-defined, hardware-accelerated networking, storage, security, and management apps running on BlueField DPUs.
BlueField-3 is expected to sample in the first quarter of 2022. It’s fully backward-compatible with BlueField-2, Nvidia says.
“Modern hyperscale clouds are driving a fundamental new architecture for data centers,” Nvidia founder and CEO Jensen Huang said in a press release. “A new type of processor, designed to process data center infrastructure software, is needed to offload and accelerate the tremendous compute load of virtualization, networking, storage, security and other cloud-native AI services. The time for BlueField DPU has come.”

Above: Nvidia’s DPU roadmap.
Nvidia’s datacenter business, which includes its DPU segment, is fast becoming a major revenue driver for the company. In February, it posted record quarterly revenue of $1.9 billion, up 97% from a year ago. Full-year datacenter revenue jumped 124%, to $6.7 billion.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.