Intel is now using the new graphics processor architecture Xe on the PCIe card H3C XG 265 for servers. This card as well as many other Intel computing accelerators with x 75, FPGA and KI processors can be controlled via the standardized oneAPI programming interface . OneAPI should appear as a gold release in December and make it easier for developers to work with many tools, including porting CUDA software.

Video streaming accelerator On Intel’s H3C-XG 310 there are four Xe-LP graphics processors and a PCIe switch.

(Image: Intel)

An estimated 75 percent of the total Internet data traffic is accounted for by video streaming; Cloud servers and the machines in the hyperscale data centers of the social networks are busy decoding, encoding and transcoding video data streams.

For this purpose, Intel has announced the server “graphics card” (Server GPU) H3C XG 310, which is equipped with four Xe-GPUs of the type Low Power (Xe-LP). It has hardware units for encoding and decoding MPEG-2, AVC (H. 264), HEVC (H. 265), VP9 and can decode AV1. A card should process around 60 video streams in parallel.

The Xe-LP-GPU is plugged in also in Intel’s Iris Xe MAX for notebooks alias DG1.

oneAPI and better Linux drivers Intel wants to better support programmers of AI and video streaming applications for servers, among other things with the mentioned oneAPI. In Intel’s DevCloud, developers can gain experience with hardware and programming examples. In addition, Intel has significantly improved the Linux drivers for the Xe GPU compared to older drivers according to their own information.

(ciw)