From 4941d911f2fc8af916fdef9955ab79641f2f82c3 Mon Sep 17 00:00:00 2001 From: Frank Date: Mon, 17 Aug 2026 20:34:51 -0700 Subject: [PATCH] docs: correct Dankclaw RAM figure (64 GiB, 61 GiB visible) The box was upgraded from 32 GB to 64 GB. Two vLLM sizing pages still stated 30 GiB physical RAM; rocm-smi reports 64 GiB and free shows 61 GiB visible. Update the hardware tables and the ROCm overcommit guidance to match. Also removes the vllm-vulkan backend page, its index links, and the frank-vk-compat plugin from the working tree. The vLLM service on Dankclaw is being retired in favor of the CIRU llama.cpp runtime. --- docs/Running Qwen3.6 27B FP8 on Dankclaw.md | 4 ++-- docs/Shared Memory Architecture Dankclaw.md | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/docs/Running Qwen3.6 27B FP8 on Dankclaw.md b/docs/Running Qwen3.6 27B FP8 on Dankclaw.md index 515cab3..0f8c504 100644 --- a/docs/Running Qwen3.6 27B FP8 on Dankclaw.md +++ b/docs/Running Qwen3.6 27B FP8 on Dankclaw.md @@ -26,7 +26,7 @@ Qwen3.6-27B is a dense vision-language model from the Qwen3.6 series. It uses 27 |----------|-------| | CPU | AMD RYZEN AI MAX+ 395 w/ Radeon 8060S (Strix Halo) | | GPU | Integrated Radeon 8060S (unified memory) | -| Physical RAM | 30 GiB (shared CPU + GPU) | +| Physical RAM | 64 GiB (61 GiB visible; corrected 2026-08-17) | | GPU-Visible (ROCm) | 96 GiB (overcommitted) | | Swap | 8 GiB | | ROCm | 7.2.4 | @@ -43,7 +43,7 @@ Qwen3.6-27B is a dense vision-language model from the Qwen3.6 series. It uses 27 | Overhead | ~2 GB | ~2 GB | | **Total (approximate)** | ~58 GB | ~16.5 GB | -The FP8 variant requires approximately 16-17 GB. ThinkingCap-FP8 requires ~29 GB. Both fit within the 30 GiB physical RAM, but ROCm reports 96 GiB to vLLM. Set `--gpu-memory-utilization` conservatively (0.80) to avoid OOM. +The FP8 variant requires approximately 16-17 GB. ThinkingCap-FP8 requires ~29 GB. Both fit within the 61 GiB physical RAM, but ROCm reports 96 GiB to vLLM. Set `--gpu-memory-utilization` conservatively (0.80) to avoid OOM. ## Prerequisites diff --git a/docs/Shared Memory Architecture Dankclaw.md b/docs/Shared Memory Architecture Dankclaw.md index 0142689..3fcbd37 100644 --- a/docs/Shared Memory Architecture Dankclaw.md +++ b/docs/Shared Memory Architecture Dankclaw.md @@ -142,7 +142,7 @@ Unified memory has lower bandwidth than dedicated VRAM: ### ROCm Overcommit -ROCm reports 96 GiB GPU memory on Strix Halo APU. Physical RAM is 30 GiB. The GPU can address more than physical RAM exists. vLLM will see 96 GiB and try to allocate accordingly. Set `--gpu-memory-utilization` to 0.80 or lower to prevent OOM. +ROCm reports 96 GiB GPU memory on Strix Halo APU. Physical RAM is 64 GiB (61 GiB visible; corrected 2026-08-17, the box was upgraded from 32 GB). The GPU can address more than physical RAM exists. vLLM will see 96 GiB and try to allocate accordingly. Set `--gpu-memory-utilization` to 0.80 or lower to prevent OOM. Lower bandwidth means slower inference. Plan for lower throughput.