Ingram Micro Adds Xilinx FPGA Cards to Line Card

Ingram Micro this week announced it has added Xilinx, a provider of field programmable gate array (FPGA) cards, to its line card in the U.S.

FPGA cards are gaining traction in high-performance computing (HPC) environments that employ them to more efficiently process machine learning algorithms.

Ingram Micro is specifically interested in making Alveo data center cards from Xilinx available to system builders looking to include FPGA cards within their white box offerings, says Ryan Grant, vice president of the components business unit at Ingram Micro.

Via the alliance with Xilinx, Grant says Ingram Micro sees an  opportunity for systems builders to construct platforms incorporating FPGAs, graphical processor units (GPUs) and traditional x86 processors optimized to run artificial intelligence models based on machine learning algorithms. Those AI models are typically trained using GPUs, but the inference engine on which the AI model is deployed typically run on an x86 processor. Increasingly, however, a case is being made to deploy those inference engines on either a lower cost GPU or an FPGA card. Eventually, those inference engines will be distributed everywhere from the cloud to the network edge.

“We want to help make this idea of AI and ML real for partners,” says Grant.

Other classes of application workloads that lend themselves especially well to FPGAs include video processing and database searches.

Less clear right now is to what degree these advanced classes of applications will be deployed on white boxes. Commercial system vendors such as Dell EMC and Hewlett-Packard Enterprise have already begun to target these applications. The opportunity system builders uniquely have is that many of the developers of these advanced applications are going to want to have a lot more influence over what classes of processors are made available in a system than commercial system vendors have been inclined to historically abide.