#AMD

Tested local Wan video gen on a Radeon 8060S (Strix Halo, 48GB UMA, Windows). ZLUDA can't run stock PyTorch; AMD's TheRock gfx1151 wheel gives native ROCm. FastWan 1.3B in 4min, Wan 14B I2V in 13.6min — VAE decode and 16GB-RAM Segfaults are the real limits.

Based on EE Times' interview with AMD AI Software VP Anush Elangovan, we assess the ROCm vs CUDA ecosystem gap. Includes hands-on experience with ROCm breaking four times on Strix Halo, plus practical guidance on choosing between NVIDIA, AMD, and Apple Silicon.

Benchmarking NII's LLM-jp-4-32B-A3B-thinking on EVO-X2 (Ryzen AI Max+ 395) with ROCm. 62.9 t/s vs Qwen3.5-35B-A3B's 44.7 t/s. Covers thinking control issues, KV cache trade-offs, knowledge cutoff, Japanese quality comparisons, code generation tests, and training data composition.

Real-world testing of AMD Lemonade v10.0.1 on Ryzen AI Max+ 395. LLM, image generation, speech recognition, and TTS running simultaneously, NPU Hybrid execution, Vulkan vs ROCm benchmarks, and discovering shared memory leaks.

Lemonade is AMD's open-source local AI server that manages multiple backends like llama.cpp and FastFlowLM across GPU/NPU/CPU, serving text, image, and audio generation through an OpenAI-compatible API.

Qwen3.5-35B-A3B is an SSM+Attention hybrid where only 10 of 40 layers consume KV cache. Going from ctx-size 4096 to 65536 on llama-server + Vulkan added just 800MB VRAM with zero throughput loss. Tested on Strix Halo (Ryzen AI Max+ 395), with q8_0 KV quant benchmarks.

After updating to AMD Software 26.3.1 on a GMKtec EVO-X2 (Ryzen AI Max+ 395), Vulkan backend fails to allocate device memory properly and falls back to CPU. Investigation and workaround by changing BIOS VRAM allocation from 48GB/16GB to 32GB/32GB.

Isolating the cause of Qwen 3.5 failing on ROCm/Vulkan via CPU inference, llama-server, and LM Studio — an AMD driver update resolved everything.

Hands-on test of huihui-ai Qwen 3.5 abliterated models in Ollama: garbage-token failures, GLM-4.7-Flash chat-template breakage, and why the official model with thinking disabled worked better.

How to configure VRAM/main memory split on the GMKtec EVO-X2 (Strix Halo) for local LLM inference. A 29.6GB model ran fine with just 8GB of dedicated VRAM.

Building an NSFW-capable local LLM on the GMKtec EVO-X2 (Strix Halo). Getting GPU inference at ~11 tokens/s with LM Studio and MS3.2-24B-Magnum-Diamond.