The rapid growth of artificial intelligence (AI) applications such as ChatGPT, Stable Diffusion, and DALL-E, is largely due to the emergence of SmartNICs (smart network interface cards) and DPUs (data processing units) used in AI servers.

Originally, SmartNICs were introduced to reduce CPU load on servers and to process network functions efficiently with high speed. The DPU is an evolution of the SmartNIC. The DPU has become an indispensable part of the AI server with more features and a customizable processor.

I-PEX offers CABLINE^®-CA IIP PLUS low-profile connecter with jumper cables employing our paddle card technology for card-to-card and card-to-board connection of DPU supporting PCIe Gen5.0.

Click here for information on paddle card technology.

Click here for information on CABLINE^®-CA IIP PLUS.

Product Specifications

GPUs (SmartNICs) and DPUs are used for other processing to reduce loads on CPUs and server network functions. 

In order to support the latest DPUs, transmission loss is large and PCIe traffic is excessive only through conventional boards and CEM (PCIe) connectors. It is necessary to prepare expensive substrates as a countermeasure, but in some cases it is difficult to meet the requirements.

By using cables to jumper signals that require high-speed transmission for differential transmission, it is not only possible to reduce PCIe bus traffic and achieve high-speed transmission, but also to flexibly arrange the placement of connectors within a limited space, the position of cards inside the server, and the design of the board within the form factor.

I-PEX is  currently developing a jumper cable product between "CABLINE^®-CA IIP PLUS" and 0.6mm pitch general standard connectors that are mainly used in server equipment. By using 0.6mm pitch connectors, existing patterns on motherboards can be used.

Product samples are currently available. Please contact us for further details or to request samples.