llama.cpp issue #28515: SYCL device memory query aborts on Intel Arc Pro B60 (Battlemage)

Fields

id

sources/noncuda-llamacpp-28515-sycl-b60-memquery

URL

https://github.com/ggml-org/llama.cpp/issues/28515

Publisher

ggml-org/llama.cpp

Kind

forum

Published

2026-09-06

Accessed

2026-10-03

notes

Sparkle Arc Pro B60 24GB (Battlemage G21, 8086:e211) on a fresh Ubuntu 26.04 with the xe driver, GuC/HuC firmware loaded, libze-intel-gpu1 26.31.39395.13 and oneAPI 2026.1: llama-server aborts at model load in ggml_backend_sycl_device_get_memory even though sycl-ls lists the device correctly over both Level Zero and OpenCL. Same abort signature as the Lunar Lake iGPU case.

Notes

Sparkle Arc Pro B60 24GB (Battlemage G21, 8086:e211) on a fresh Ubuntu 26.04 with the xe driver, GuC/HuC firmware loaded, libze-intel-gpu1 26.31.39395.13 and oneAPI 2026.1: llama-server aborts at model load in ggml_backend_sycl_device_get_memory even though sycl-ls lists the device correctly over both Level Zero and OpenCL. Same abort signature as the Lunar Lake iGPU case.

Cross-references

This record references no other records.