Majestic Labs has unveiled its Prometheus AI server, designed to overcome memory limitations in AI training. The system features custom accelerator and memory interface chips, employing a "memory-first" architecture that decouples processors from memory. This allows processors to access a shared, massive memory pool directly, with inter-processor communication routed through this high-speed memory. AI
影响 This new server architecture could significantly boost AI training efficiency by addressing memory bottlenecks, potentially lowering costs and accelerating development.
排序理由 Launch of a new AI server with a novel architecture and custom hardware.
在 Mastodon — mastodon.social 阅读 →
AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →
