What Is A Data Processing Unit (DPU) And Why Is NVIDIA Betting On It?


At the GPU Technology Conference 2020, Jensen Huang, NVIDIA’s CEO, unveiled a new family of processors branded as the BlueField-2 Data Processing Unit (DPU). The DPU is accessible to the developers via the software platform, the DOCA SDK. The DPU and DOCA SDK are comparable to NVIDIA’s powerful combination of GPU hardware and CUDA software.  

Having dominated the AI accelerator market, NVIDIA is now aiming to expand it to the data center infrastructure acceleration and optimization. 

Why is Jensen Huang bullish about the DPU market and how it matters to the enterprise data center? Here is an attempt to explain the evolution of DPU in simple terms.

The Aggregation and Disaggregation of Enterprise Infrastructure

During the 90s, the combination of Intel x86 CPU and OS software offered unmatched power to enterprises. The rise of client/server computing, followed by n-tier computing, paved the way for distributed computing. Enterprises ran databases, application servers and custom line of business software on a fleet of x86 servers. 

During the early 2000s, VMware introduced ESX, a hypervisor that brought the ability to virtualize the x86 CPU. Enterprises could run multiple virtual machines on a single powerful server. CPU virtualization was the first step towards the aggregation of enterprise infrastructure. 

The hypervisor made the hardware programmable. Developers could write code to define and provision a virtual machine without manual intervention. This programmability aspect of infrastructure became the foundation of the modern cloud computing paradigm. 

Based on the success of ESX, VMware moved towards network and storage virtualization. Traditional infrastructure players such as Cisco and EMC have started to build virtualized network and storage services to compete with VMware. In 2012, VMware acquired Nicira, the software-defined networking startup, for $1.26 billion, which was branded as NSX. In March 2014, VMware announced vSAN, its

NVIDIA Wants To Change The Game Yet Again With The ‘DPU’


This week, I tuned into NVIDIA’s annual GPU Technology Conference, or GTC, albeit virtually due to the pandemic. The processor powerhouse has a long history of category-defining innovation, dating back to its launch of the first 3D GPU in 1999. This week’s keynote was chock full of news and innovation, including another all-new processor category designed to muscle NVIDIA into the datacenter market. Let’s take a look at the new processor and several other announcements from GTC 2020.

Look out, data center market

One of the most significant announcements was the unveiling of the DPU (data processing unit)—a whole new category of processors designed to offload networking, security, and storage tasks from CPUs in the data center. Essentially, you can think of these DPUs as smart NICs. Under the new family moniker BlueField, NVIDIA unveiled the first two of these accelerators: the BlueField-2 and the BlueField-2X.

BlueField-2 brings together an array of fully programmable Arm cores, a ConnectX-6 Dx network adapter (derived from NVIDIA’s acquisition of Mellanox) and an NVME-optimized Controller to deliver up to 200 GB/s performance for both traditional and modern workloads. The DPU includes hardware offloads for workloads such as software-defined storage, zero-trust security, networking and management. By offloading these functions from the CPU, NVIDIA says a single BlueField-2 can perform the same services that can require as many as 125 CPU cores. Just think about what those freed-up CPU cores could accomplish. 

This claim isn’t “crazy-talk.” I look at what the teams is doing at AWS with “Nitro” and offload and virtualization is very much a reality at the cloud giant.  NVIDIA can offer this capability to every cloud giant and enterprise OEM who does not have an in-house silicon team.

The second DPU, BlueField-2X, includes the same key

“DPU” Smart NIC And More


NVIDIA Co-founder and CEO Jensen Huang rarely disappoints his audience nor his investors. This week he once again delivered the goods at the European GPU Technology Conference. Announcing a broad range of hardware and software innovations, Jensen made it clear that he intends to reshape computation, from GPUs to CPUs to NICs and switches. So, let’s dive into Jensen’s goody bag and see how he plans to shape the market and NVIDIA’s future.


NVIDIA led off with a new NIC from the Mellanox division. Setting the stage for a reimagining of the server, NVIDIA added multi-core Arm CPUs and AI acceleration to the staid Network Interface Card (NIC). Smart NICs have been around for a while, but have yet to reach broad-scale adoption outside a few hyperscale cloud providers. However, as networking becomes the new bottleneck for moving massive data sets (and the lynchpin for data center security), NVIDIA sees an opportunity to augment the NIC with far more processing power. This could potentially alter the data center landscape significantly. By offloading much of the work usually done by server CPUs, the Data Processing Unit (DPU) can create a new platform of high-performance processing, and leverage the GPU’s AI capabilities into the networking market. If NVIDIA can gain traction here, this move could effectively capture some of the revenues currently going to CPU vendors, shifting that spend to more capable, presumably higher-margin Smart NICs. Said another way, NVIDIA is freeing up cycles on the server cores, which allows it to focus on the application that will process the data, instead of spending time unpacking TPC/IP headers and acting as a firewall.