fix(xsysinfo): include AMD GTT in APU VRAM detection - #12094
Open
leilei3167 wants to merge 2 commits into
Open
leilei3167 wants to merge 2 commits into
leilei3167 wants to merge 2 commits into
Conversation
On shared-memory AMD APUs (e.g. Strix Halo), rocm-smi VRAM is only the BIOS UMA carve-out while usable HIP memory lives in GTT. Query --showmeminfo all and fold GTT when the VRAM/GTT ratio matches an APU, preserving dedicated dGPU accounting. Fixes mudler#12058 Signed-off-by: lei_lei <imleilei123@gmail.com>
leilei3167
force-pushed
the
fix/issue-12058-amd-gtt-vram
branch
from
September 16, 2026 22:29
a9eb813 to
903e0a4
Compare
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Fixes #12058
On AMD shared-memory APUs (Strix Halo / gfx1151 and similar),
rocm-smi --showmeminfo vramonly reports the BIOS UMA carve-out (e.g. 512 MiB). Usable HIP memory lives in the much larger GTT pool (often 100+ GiB). LocalAI then showed "512 MB GPU RAM", and auto-fit /MinPerGPUVRAMdecisions were sized from that undercount.This change:
rocm-smi --showmeminfo all --csv(falls back tovramon older builds)GTT >= 8×VRAM), folds GTT into reported Total/UsedNotes for Reviewers
LOCALAI_VRAM_BUDGETremains an allocation cap on the detected total; it still cannot raise a bad detection by itself.Evidence (old red → new green)
Signed commits