Intel GPU D-Series 'Crescent Island' gains thermal sensors, RAS logging for Linux 7.5

Intel is expanding Linux kernel support for its GPU D-Series 'Crescent Island' AI accelerator in the upcoming Linux 7.5 kernel, which will become the first stable release of 2027. The latest drm-xe-next pull request adds hardware monitoring capabilities including temperature sensors for PCIe, memory controllers, and VRAM thermal channels, alongside reliability features like Common Platform Error Record logging and double-bit ECC error handling.

This hardware, based on the Xe3P architecture, was announced in October 2025 and has seen continuous driver development since. The new additions build on a year of upstream enablement work by Intel engineers, targeting data center and AI inference deployments where thermal management and error reporting are critical operational requirements.

Xe driver additions for Crescent Island in Linux 7.5

• HWMON thermal sensors: The Xe driver now handles temperature reporting for the GPU die, PCIe interface, memory controller, and platform-reported VRAM thermal channels through the hardware monitoring subsystem.
• CPER error logging: Common Platform Error Record support enables standardized RAS (Reliability, Availability, Serviceability) error logging compatible with UEFI firmware error reporting infrastructure.
• OOBMSM telemetry: Out-of-band management subsystem telemetry gains 'OCode' late binding support, allowing dynamic configuration of telemetry collection without requiring driver reload.
• Memory ECC handling: The driver now specifically manages double-bit ECC error detection and reporting, a critical feature for data integrity in AI accelerator workloads.

Why it matters

These additions bring Crescent Island closer to production readiness for Linux-based AI infrastructure, giving data center operators standard thermal and error monitoring through existing kernel interfaces rather than proprietary tools. For distributions and system administrators, the upstream HWMON and CPER integration means the GPU D-Series will be manageable through familiar tooling like lm-sensors and rasdaemon when hardware becomes available.