Qwen 27B (IQ4_XS, Q8/Q4 KV) at full 262k context on a 16GB RX 9070 XT: 80+ t/s decode with Adaptive KV Streaming
英文标题：Qwen 27B (IQ4_XS, Q8/Q4 KV) at full 262k context on a 16GB RX 9070 XT: 80+ t/s decode with Adaptive KV Streaming
I have been working on implementing adaptive KV streaming and RDNA 4 / ROCm optimizations for llama.cpp to enable running Qwen3.8-27B (IQ4_XS weights, Q8_0/Q4_0 KV cache) with its full 262,144-token context window on a single 16 GB RX 9070 XT GPU at 80+ t/s decode. Qwen 27B IQ4_XS is ~14.3 GB,
打开原文相关 2 · 机会 0
中文
I have been working on implementing adaptive KV streaming and RDNA 4 / ROCm optimizations for llama.cpp to enable running Qwen3.8-27B (IQ4_XS weights, Q8_0/Q4_0 KV cache) with its full 262,144-token context window on a single 16 GB RX 9070 XT GPU at 80+ t/s decode. Qwen 27B IQ4_XS is ~14.3 GB,
中文全文翻译正在补齐，可先查看下方英文原文。
English / 原文
I have been working on implementing adaptive KV streaming and RDNA 4 / ROCm optimizations for llama.cpp to enable running Qwen3.8-27B (IQ4_XS weights, Q8_0/Q4_0 KV cache) with its full 262,144-token context window on a single 16 GB RX 9070 XT GPU at 80+ t/s decode. Qwen 27B IQ4_XS is ~14.3 GB,