Intel announced Arctic Sound M, a 150W datacenter GPU targeting energy-efficient AI acceleration.

Intel has detailed its Arctic Sound M data center GPU, a PCIe Gen 4 graphics card based on the Intel Xe-HPG GPU architecture, announcing an availability target of Q3 2022. However, given Intel's history of delays with its Arc GPU line, including the recent announcement that Arc Alchemist desktop GPUs would be delayed and launched as a China-exclusive instead, expectations for the Arctic Sound M timeline remain cautious.

The Arctic Sound M comes in two distinct configurations. The flagship design features a single ACM-G10 GPU with a 150W power envelope, while the second variant houses two ACM-G11 GPUs and operates at a lower 75W TDP, positioning it for high-density multi-purpose workloads. Both are positioned as "super flexible data center GPUs aimed at cloud gaming, media processing and delivery, virtual desktop infrastructure, and inference."

The Arctic Sound M line offers substantial performance capabilities for its target applications. The GPUs can reportedly handle 30+ 1080p streams, 40+ game streams, support up to 62 Virtualized Functions, and deliver up to 150 AI TOPs. Intel has equipped the new Arctic Sound M GPUs with key technologies including AV1 hardware encode and decode functionality, along with XMX AI-Accelerators built-in.

Industry adoption appears to have generated early momentum, with Intel reporting 15+ system designs already in development from major manufacturers including Dell Technologies, HPE (Hewlett-Packard Enterprise), CISCO, and Supermicro. This existing ecosystem support may contribute to more predictable deployment compared to past Intel GPU launches, though the Q3 2022 launch timeline remains subject to the company's historical pattern of GPU delivery delays.