ryan4yin

Qwen3.8-Flash-Next — llama.cpp on RTX 4090 24GB

Platform:

• CPU: Intel Core Ultra 7 270K Plus (24C/24T)
• RAM: 96GB DDR5-5600 (2x48GB dual channel, ≈89.6 GB/s theoretical bandwidth)
• GPU: RTX 4090 24GB

Model: unsloth/Qwen3.8-Flash-Next-GGUF, UD-IQ3_XXS (≈82GB, 76.3 GiB) + mmproj-F16 (≈900MB, vision enabled). llama.cpp mainline (ghcr.io/ggml-org/llama.cpp:full-cuda13). Final config as of 2026-08-29.

ryan4yin

Best llama.cpp config for Qwen3.8-27B (RTX 4090 24GB)

Quality-first setup (current 2026-08-28, matches docker-compose.qwen3.8-27b-xl.yml), measured on:

• CPU: Intel Core Ultra 7 270K Plus (24C/24T, no HT)
• RAM: 96GB DDR5-5600 (2x48GB dual channel, ≈89.6 GB/s theoretical bandwidth)
• GPU: RTX 4090 24GB Use the BeeLlama fork image (server-cuda13-v0.4.3) — mainline llama.cpp silently falls back to CPU for non-q4 KV caches on Qwen3.x hybrid architecture (no error is reported).

ryan4yin

# https://wiki.metacubex.one/en/example/conf/
######### 锚点 start #######
# 策略组相关
pr: &pr
  type: select
  proxies: [默认, 香港, 香港自动选择, 台湾, 台湾自动选择, 日本, 日本自动选择, 新加坡, 新加坡自动选择, 美国, 美国自动选择, 其它地区, 全部节点, 自动选择, 直连]
#这里是订阅更新和延迟测试相关的
p: &p
  type: http
  interval: 3600

ryan4yin

/*# 1- Identify the UART interface, GPIO pins and Alternate Function #########*/
/* For example:
 * B-L475E-IOT01A:
 * PB6     ------> USART1_TX
 * PB7     ------> USART1_RX
 * 
 * NUCLEO-L476RG:
 * PA2     ------> USART2_TX
 * PA3     ------> USART2_RX
 *