With the official launch of the Intel Core Ultra Series 3 (Panther Lake), I am looking to confirm the baseline compute support for the new Xe3 iGPU architecture within the BOINC environment.

Preliminary reports indicate the Arc B390 delivers roughly 77% more performance than the previous generation, putting it in the range of entry-level discrete desktop GPUs like the RX 6600 in some workloads. Given the transition to the Intel 18A process and the increased 12 Xe-core count on top SKUs like the Ultra X9 388H, these units appear to have significant potential for high-efficiency, 24/7 volunteer computing in small form factor systems.

I am specifically interested in:

Current reliability of project-specific OpenCL kernels on Xe3 (Einstein@Home, PrimeGrid, etc.).

Status of Level Zero support for reducing CPU overhead during concurrent GPU tasks.

Any observed stability issues with the 26.01 production compute runtime under sustained full load.

If any early adopters have stderr logs or completion times for GPU-heavy tasks on this hardware, I would appreciate seeing how they compare to the previous Meteor/Lunar Lake generations.