Tag
OpenAI announces the first performance results of Jalapeño, its custom inference chip, which delivers better throughput and latency than commercial systems, as part of an integrated compute strategy.
OpenAI's Jalapeño custom inference chip shows industry-leading speed and efficiency, delivering higher throughput and lower latency across various AI models.
LongCat-2.0 is a 1.6 trillion parameter mixture-of-experts model trained entirely on custom AI ASICs.
Meta developed a custom CXL bridge chip (Vistara) to reuse older DIMMs in new servers, addressing memory shortages and reducing costs without significant performance loss.
Anthropic is in discussions with Samsung to develop custom AI chips, aiming for hardware independence and improved efficiency, following OpenAI's recent custom chip announcement.
This tweet discusses the convergence of ML research on attention-based, matmul-optimized algorithms due to hardware constraints, drawing on the 'hardware lottery' concept and noting OpenAI's 9-month chip tape-out as a potential sign of hardware-research co-design.
OpenAI unveiled its custom inference chip called Jalapeño, built with Broadcom, marking a significant move by Big Tech to reduce dependence on Nvidia.
OpenAI announces a custom chip designed in collaboration with Broadcom to enhance its AI infrastructure.
OpenAI has announced its first custom AI chip designed specifically for running ChatGPT and future AI agents, marking a major step in reducing reliance on external hardware providers.
OpenAI unveiled its first custom-built inference processor, named Jalapeño, developed with Broadcom to improve performance-per-watt and reduce reliance on Nvidia GPUs.
A custom FPGA implementation of a Transformer with KV cache achieves 56,000 tokens per second at 80 MHz, running microGPT on a tiny LCD.
A custom digital chip designed gate-by-gate achieves over 56,000 tokens/sec running a Transformer with KV cache at just 80 MHz, prototyped on an FPGA.