• /
    • Blog

Blog

  1. Compressing KV cache 2026-06-14 Compressing KV cache And expanding the context window (2 minutes read)
  2. Switching LLamacpp to tensor split mode 2026-06-12 Switching LLamacpp to tensor split mode And almost doubling the performance (3 minutes read)
  3. Switching from Gemma 4 to Gemma 4 QAT 2026-06-06 Switching from Gemma 4 to Gemma 4 QAT While improving perplexity (4 minutes read)
  4. Ollama vs Llama.cpp 2026-06-05 Ollama vs Llama.cpp How I improved tokens per second performance by 10% (7 minutes read)
  5. Building Llama.cpp from source 2026-05-20 Building Llama.cpp from source So we can enable NVIDIA CUDA support (4 minutes read)
  6. LTO Host bus adapters - LTO part 2 2025-09-14 LTO Host bus adapters - LTO part 2 Backups are costing me only 1.1 euro per 1TB (5 minutes read)
  7. Starting with LTO tapes - LTO part 1 2025-09-12 Starting with LTO tapes - LTO part 1 Backups are costing me only 1.1 euro per 1TB (9 minutes read)
  8. BLE WCH chips 2025-08-02 BLE WCH chips Wireless options from Nanjing Qinheng Microelectronics (4 minutes read)
  9. Improving performance of branching part 1 2025-06-07 Improving performance of branching part 1 By annotating the code (5 minutes read)
  10. 2025-06-01 WCH single cycle FLASH is a lie It really is an illusion (3 minutes read)