News Categories

Intel shadows NVIDIA; unveils Infrastructure Processing Unit (IPU) to drive up data center efficiency

By Vijay Anand - on 21 Jun 2021, 12:22pm

Intel shadows NVIDIA; unveils Infrastructure Processing Unit (IPU) to drive up data center efficiency

With the IPU, a cloud provider can securely manage infrastructure functions while enabling its customer to entirely control the functions of the CPU and system memory. (Image source: Intel)

Targeting hyperscale cloud service providers, Intel's new Infrastructure Processing Unit (IPU) is expressly designed to free up CPU cycles from managing often used microservices and allow the data center CPUs to tackle more more critical and valuable applications and improve inefficiencies.

According to a Facebook research, 22 to 80% of CPU cycles could be consumed on managing microservice overheads within the data center such as security, storage and network virtualisation workloads. In the digital and ever-more connected world that we live in today, data center needs are only rising and with that, data center traffic, security and storage management needs are only going to increase. 

As a programmable device that's designed to enable cloud and communication service providers to reduce processing overheads, and free up CPU resources, the IPU allows the data center customers to better utilize resources and with more certainty to balance, plan and forecast their needs.

(Image source: Intel)

All this might sound quite familiar and that's because NVIDIA's Data Processing Unt (DPU) is also designed to accelerate data center infrastructure and services in the same three pillars. However, the difference is that Intel's solution is the first to be built in collaboration with hyperscale cloud partners while NVIDIA's solution was born out of wanting to evolve the data center infrastructure with their Mellonox network fabric acquisition. Now, while NVIDIA already has BluField-2 and 3 DPUs, DOCA SDK to build applications to take advantage of DPU-accelerated servers/services, Intel has yet to unveil more details on their IPU's capability or its architecture.

The first of Intel’s FPGA-based IPU platforms are deployed at multiple cloud service providers and our first ASIC IPU is under test. - Patty Kummrow, VP of Data Platforms Group and GM of Ethernet Products Group at Intel.

Intel's upcoming plans are to roll out more FPGA-based IPU platforms and dedicated  ASICs, and perhaps at the next hop, we will get to hear more details on their IPU specific add-on solutions.

The intelligent data center architecture will have three categories of compute — CPU for general-purpose compute, XPU for application-specific or workload-specific acceleration (such as GPUs), and IPU for infrastructure acceleration — that will be connected through programmable networks to efficiently utilize data center resources. (Image source: Intel)

Source: Intel

Join HWZ's Telegram channel here and catch all the latest tech news!
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.