Repository navigation
[SYCL]Sysman free-memory query may be unavailable #28239
Description
Name and Version
version: 0.3.0-dev (build 10759, commit b81c99b)
built with Clang 20.1.8 for Windows x86_64
Operating systems
Windows
Which llama.cpp modules do you know to be affected?
llama-server
Command line
llama\llama-server.exe -m llama\Huihui-Qwen3.8-27B-abliterated-Q6_K.gguf --host 0.0.0.0 --n-gpu-layers -1 --reasoning off --reasoning-effort minimal --reasoning-budget 0 --temp 0.6 --top-k 20 --top-p 0.95 -c 40960 -n 32768 --min-p 0.0 --alias ALIAS --timeout 99999999 --cpu-moe --log-file log.txt --spec-type draft-mtp,ngram-map-k4v,ngram-mod -sm layer --load-mode auto -ctk q8_0 -ctv q8_0 --jinjaProblem description & steps to reproduce
In the console, the message "Sysman free-memory query may be unavailable" keeps recurring（Not appearing in the log）.
And after some time, error
Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable.
Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable.
Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable.
Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable.
Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable.
Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable.
Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable.
3.50.265.387 W ggml_sycl_flash_attn_ext_onednn: oneDNN SDPA failed (level_zero backend failed with error: 20 (UR_RESULT_ERROR_DEVICE_LOST)); falling back to TILE kernel
Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable.
3.50.267.586 E srv update_slots: decode() failed: level_zero backend failed with error: 2147483646 (UR_RESULT_ERROR_UNKNOWN)
3.50.267.590 E srv send_error: task id = 10, error: decode() failed: level_zero backend failed with error: 2147483646 (UR_RESULT_ERROR_UNKNOWN)
3.50.267.597 I slot release: id 3 | task 10 | stop processing: n_tokens = 958, truncated = 0
3.50.267.609 W srv stop: cancel task, id_task = 10
First Bad Commit
Relevant log output
Logs
[log.txt](https://github.com/user-attachments/files/31734482/log.txt)
Activity
[-]Misc. bug:[/-][+]The Sysman free-memory query may not be available[/+][-]The Sysman free-memory query may not be available[/-][+][SYCL]Sysman free-memory query may be unavailable[/+]arthw commentedon Sep 2, 2026
@DDXDB
#28227 fix it.
Please try it!
Thank you!
DDXDB commentedon Sep 3, 2026
I tried, and the result was:
Use
microcai commentedon Sep 4, 2026
same error
arthw commentedon Sep 4, 2026
@DDXDB @microcai
I think
F:\llama.cpp\ggml\src\ggml.c:3953: GGML_ASSERT(a->ne[0] == b->ne[0]) failedis another issue.Have you met the same issue with old version?
Could you help confirm if
Sysman free-memory query may be unavailableissue is fixed by #28227?Thank you!
DDXDB commentedon Sep 4, 2026
This is the result obtained by using Build #28227.
arthw commentedon Sep 5, 2026
@DDXDB
The new issue of
F:\llama.cpp\ggml\src\ggml.c:3953: GGML_ASSERT(a->ne[0] == b->ne[0])is due to the functionggml_set_rows()error.The tensor shapes of a, b are different which is not supported by ggml.
The GGUF is new one.
Need ggml to check and support it.
Thank you!
handanikr commentedon Sep 9, 2026
@arthw Excuse me. I am experiencing a similar issue, and It seems like a duplicate issue, I am commenting this here because it seems more appropriate to continue the discussion in the comments rather than opening a new issue.
OS: Windows 11 IoT Enterprise LTSC 24H2
CPU: AMD Ryzen 7600
RAM: 32GB (16 * 2)
GPU: Intel Arc A770 16GB
Backend: SYCL
Model: Unsloth's Qwen3.8-27B-UD-IQ4_XS.gguf
Build: This issue first occurred in 0.3.0-dev (build 10717, commit a32af33), I am currently using llama.cpp version 0.4.0-dev (build 10878, commit 4850c77) but it doesn't fixed. 0.3.0-dev (build 10715, commit 662a0b0) has no problem and I can't find build 10716.
Parameters: .\llama-server.exe -m "D:\model\Qwen\Qwen3.8-27B-UD-IQ4_XS.gguf" -c 12288 -n 8192 -ngl 99 -fa on --temp 1 --top-k 20 --top-p 0.95 --min-p 0 --presence-penalty 0 --repeat-penalty 1.
In my case, The message Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable. is flooding the console. Even when running
llama-cli.exe --versionto check the version, this message appeared first, followed by the version information.This message appears repeatedly while the model is loading and the server is starting up; furthermore, when the chatbot is actually generating responses, console become so flooded with this message that it is nearly impossible to read any other console entries.
The interesting thing is that the output is still produced somehow, though the tok/s value drops significantly compared to when using llama.cpp or koboldcpp 1.120 (Vulkan). However, it is not certain whether it succeeds every time, so I think I need to test it further.
arthw commentedon Sep 10, 2026
@handanikr
It's fixed by PR #28227.
The PR is still not merged.
Please try and feedback your result!
The PR has been created for 8 days.
We are waiting for the feedback of it from the users.
Thank you!
MartinEmrich commentedon Sep 19, 2026
This apparently got merged a few days ago, and with b11052, I no longer get these Warning: zesInit failed [ggml_check_sycl] with code 2013265921. Sysman free-memory query may be unavailable. warnings. (Windows 11, Arc Pro 140T)
arthw commentedon Sep 19, 2026
@MartinEmrich
Good news!
Thank you for your feedback!