Models & Labs
OpenAI Reveals Custom Chip 'Jalapeño' with Broadcom
TechCrunch AIhigh confidence
Why it matters
- →Custom chips like Jalapeño can significantly reduce operational costs for AI companies.
- →This development allows OpenAI to optimize its entire tech stack, enhancing performance and reliability.
- →Reducing reliance on Nvidia GPUs could lead to more competitive AI infrastructure solutions.
OpenAI has announced its first custom-built inference processor, Jalapeño, created in partnership with Broadcom. This chip is designed to enhance the efficiency of OpenAI's AI models, offering improved performance-per-watt over current alternatives. The development marks a strategic move to reduce dependence on Nvidia GPUs, aligning with similar efforts by tech giants like Google and Amazon. Jalapeño is specifically optimized for inference tasks, which could significantly lower operational costs for OpenAI's real-time coding models. This development underscores OpenAI's commitment to optimizing its entire technology stack for better performance and affordability.
Read originalMore from TechCrunch AI
More in Models & Labs
Models & Labsmodels
Llama.cpp b9784 Release Enhances Hexagon Performance
The latest b9784 release of llama.cpp brings significant optimizations to Hexagon's matrix multiplication capabilities. By reworking the MUL_MAT and MUL_MAT_ID operations, the update introduces a 32x32 tiled weight repack and improved kernel parameters, enhancing performance and efficiency. These changes aim to optimize register usage and streamline activation processing, particularly benefiting users leveraging Hexagon's architecture. This release doesn't introduce new models but focuses on refining existing processes, making llama.cpp more robust for developers working with diverse hardware configurations.
llama.cpp Releases
Models & Labsmodels
llama.cpp b9788 release enhances dual-GPU support
The latest release of llama.cpp, b9788, introduces significant improvements for dual-GPU setups with SYCL support, particularly enhancing tensor parallelism. By implementing a degenerate ring all-reduce for dual-GPU configurations, the update optimizes performance for both small and large tensor operations, mirroring CUDA's NCCL allreduce pattern. This release notably boosts performance metrics, with Llama-3.3-70B and Qwen3-Coder-Next-80B-A3B models showing substantial speed improvements. The update positions llama.cpp as a more competitive option for multi-GPU environments, without adding new dependencies or altering build configurations.
