inference/research
verifiedhigh · 0.90updated 2026-10-03

llama.cpp issue #11559: Vulkan Q4_K_M inference speed degradation

forum

Fields

id
sources/gotcha-llamacpp-issue-11559-vulkan-q4km-regression
URL
https://github.com/ggml-org/llama.cpp/issues/11559
Publisher
llama.cpp GitHub issues
Kind
forum
Published
2025-01-31
Accessed
2026-10-03
notes
Bisected to commit adc5dd92 on Windows/Vulkan with dual Radeon PRO W7800; pp unchanged, tg regressed 42.1 -> 36.4 tok/s.

Notes

Bisected to commit adc5dd92 on Windows/Vulkan with dual Radeon PRO W7800; pp unchanged, tg regressed 42.1 -> 36.4 tok/s.

Cross-references

This record references no other records.