As NVIDIA continues to grow in the data center market, the technology heavyweight has laid out a multi-year plan to create a new kind of chip for data centers.
The plan is said to be aimed at siphoning off more functions from its main rival, Intel.
Typically, the chips sit alongside Intel’s central processors to offload some of the computing work, but NVIDIA is now looking to take more of the IT load with its series of data processing unit chips.
“A new kind of processor is needed. Modern data centers are software-defined, making them more flexible and adaptable. That creates an enormous load. Running a data center’s infrastructure can consume 20-30% of its CPU cores,” said NVIDIA CEO Jensen Huang.
The data center chips will combine networking technology from Mellanox with artificial intelligence and computing power from Arm, both of which were acquired by NVIDIA for billions of dollars.
The new NVIDIA BlueField 2 DPU is a programmable processor with powerful Arm cores and acceleration engines for at-line-speed processing for networking, storage and security. As part of their partnership, VMware systems will also be ported onto BlueField.
“We are going to bring a ton of technology to networking. In just a couple of years, we’ll span nearly 1,000 times in compute throughput [on the DPU],” said Mr. Huang.
If used successfully, the chips could detect hackers trying to break into a data center or review network traffic for unusual patterns by using artificial intelligence, said Manuvir Das, NVIDIA’s Head of Enterprise Computing. In the past, these functions would have required a combination of different data center chips.
The data center chip models will have limited features in its debut in the coming months, while full-fledged versions are planned to be released in the next two years.
Dell Technologies and Lenovo are expected to integrate the chips into their systems.
NVIDIA’s chips have long been used for improving video game graphic procession, but the company is pivoting towards helping speed up artificial intelligence tasks.
NVIDIA also announced DOCA, a programmable data-center-infrastructure-on-a-chip architecture as well as new NVIDIA RTX A6000 and NVIDIA A40 GPUs to accelerate rendering and compute workloads.
“DOCA SDKs let developers write infrastructure apps for software-defined networking, software-defined storage, cybersecurity, telemetry and in-network computing applications yet to be invented,” Mr. Huang said.
NVIDIA has also seen successful widespread adoption of the NVIDIA EGX edge AI platform by the world’s leading tech companies, which is said to bring a new wave of secure, GPU-accelerated software, services and servers to enterprise and edge data centers.
The EGX platform will expand to combine NVIDIA’s Ampere architecture GPU and BlueField 2 DPU capabilities on a single PCIe card, giving enterprises a common platform to build secure, accelerated data centers.
Image credit: DICT