Adreno Driver Update Slows Snapdragon X2 LLM Speeds
Users report Snapdragon X2 Elite laptops lose most local LLM generation speed after Adreno driver 32.0.181.0 is installed.
Updated on Nov. 1, 2026 in Computers
Local LLM text generation using the llama.cpp OpenCL backend dropped 69 to 87 percent on Snapdragon X2 Elite systems after users moved from Adreno driver 32.0.172.2 to 32.0.181.0. Users also reported Premiere Pro render times rising sixfold after the same update.
Why it matters
The slowdown affects users running AI models locally on Windows laptops powered by the Snapdragon X2 Elite and Adreno GPU, where performance gains from the driver update are now in question. The reports point to a GPU driver regression rather than a hardware limit, since CPU and NPU results stayed largely unchanged.
Benchmarks show Qwen3-0.6B generation falling from 98.1 to 12.8 tokens per second and GPU prompt processing dropping 73 percent on Qwen3-Coder-30B-A3B-Instruct after the driver update. The reports do not yet establish whether the regression is limited to the OpenCL path, though the Redditor who first raised it suspects so.
The players
Qualcomm
Qualcomm is the semiconductor company that designs the Snapdragon X2 Elite processor and the Adreno GPU whose driver is at the center of the reported slowdown.
The details
Users found the slowdown disappeared when the driver was rolled back to 32.0.172.2 and returned when 32.0.181.0 was reapplied. Qwen3-0.6B running on the GPU through WebGPU held steady at around 90 tokens per second on both drivers, pointing to the OpenCL backend as the likely source.
Timeline
Article reporting the Adreno driver 32.0.181.0 slowdown reports was published on 2026-11-01.
The problem has been posted to the Qualcomm Developers Discord server, where a hotfix may be issued as soon as possible.
Under the Hood
The slowdown sits in the OpenCL path that llama.cpp uses on Adreno GPUs under Windows, while WebGPU on the same hardware kept its throughput, which narrows the fault to a specific driver backend. This makes the Snapdragon X2 Elite's GPU compute stack a more fragile foundation for local AI than its CPU and NPU results suggest.
Anyone running local LLMs on a Snapdragon X2 Elite laptop may see text generation fall to a fraction of its previous speed after installing Adreno driver 32.0.181.0. Rolling back to 32.0.172.2 restored the earlier performance in testing, while WebGPU-based inference was unaffected.
The takeaway
Check your Adreno driver version before running local models on a Snapdragon X2 Elite laptop. If the slowdown appears, rolling back to the prior driver or using the WebGPU path may keep generation speeds usable until a fix ships.
Further reading
Explore more reporting in the Computers section for context on how driver updates affect local AI performance.
