What is NVIDIA DPU and how does it influence GPUs for gaming?

NVIDIA recently introduced a new type of card designed for Data Centers that they called DPU, but what is this new piece of hardware and how will it affect graphics cards in our PCs in the future? Does it mean such an important change in these as has been the implementation of Ray Tracing?

At the GPU Technology Conference, the CEO of NVIDIA presented a new family of processors under the BlueField-2 brand that has been baptized as DPU, to which they have given access to developers through an SDK called DOCA, which is the equivalent for the DPU to what is CUDA for the GPUs.

The NVIDIA DPU

Actually the NVIDIA DPU is nothing more than a new name for Mellanox SmartNIC, remember that NVIDIA bought this company for almost 7000 million dollars.

The BlueField-2 DPUs are nothing more than programmable network and storage interfaces that take the workload off the CPU when processing these functions.

This means that applications have higher performance and do not have to worry about repetitive tasks based on the distribution and storage of data on a network.

What is NVIDIA's interest in entering the data center market?

In its simplest definition, a DPU is nothing more than what we colloquially call SmartNIC, that is, an intelligent network controller in charge of managing the movement of data and access to them in a system.

In cloud computing, virtualization means that it is the processor that has to completely manage the infrastructure by software, there is no supporting hardware for this. This means that the processor, most of the time an X86 has to take care of everything, not only running the operating system and applications but also network traffic, managing storage, security and more.

In other words, everything is software and the applications compete with the software in charge of data movement in order to obtain the resources of the processor, hence the need for a SmartNIC that frees the processor from these tasks and allows it to focus its resources, not on the processor. data movement but to the execution of applications.

How does this affect the future of graphics cards in our PCs?

Current GPUs have an SoC-type infrastructure, but with the implementation of chiplet-based configurations, these will evolve to NoC-type configuration, which are based on a network infrastructure to communicate the different elements where a SmartNIC is in charge of managing communication. between the different elements.

It is at this point where a DPU will be necessary to be in charge of managing data movements efficiently under the new paradigm. That is, the instructions for data movements within the GPU would not run as shader programs within the Compute Units but would be done by invoking the DPU, as well as the DPU will be used to manage both the accesses to the RAM and the SSD.

The CEO of NVIDIA, Jensen Huang, showed on the roadmap for his DPUs, where the BlueField-4 stands out, which will be launched in 2023 and includes a GPU inside, or would it be the other way around? In any case, taking into account the launch timing of the different GeForce architectures, it is possible that BlueField-4 is NVIDIA Hopper.

The reality is that 2023 is a very important year for NVIDIA, as it is the date in which they have set aside the traditional raster-oriented GPU. Is DPU and GPU integration part of that roadmap? It is very likely, since for ray tracing you work with extremely complex databases.