Qeravio
Canonical AI event

AMD broadens MoE ncols_opt tile heuristic for RDNA3.5 architecture

The AMD Radeon 8060S GPU was used to verify the changes. The performance of the LFM2.5-8B-A1B-UD-Q4_K_M model improved by 16.198%.

16 Sept 20261 verified claims1 sources1 observations
What happened

The official source reports this update: b10997: HIP: broaden MoE ncols_opt tile heuristic on RDNA3.5 architecture. HIP: broaden MoE ncols_opt tile heuristic on RDNA3.5 architecture ( #28935 ) It's found the MoE ncols_opt tile heuristic needs to be broadened to include the RDNA3.5 architecture. The code change is implemented in ggml/src/ggml-cuda/mmq.cu and just change the GGML_CUDA_CC_IS_RDNA3_0 to GGML_CUDA_CC_IS_RDNA3 in the condition. The dense dispatch logic remains unchanged.

Why it matters

This official update documents a development concerning b10997: HIP: broaden MoE ncols_opt tile heuristic on RDNA3.5 architecture. Its practical significance depends on the scope and evidence stated by the source.

What to watch next

Read the official source update and verify its stated scope, evidence and timing before acting on it.

Connected knowledge

Entities affected by this event

Continue this topic
Evidence trail

Sources behind the event