Kioxia和SanDisk在QLC NAND技术方面的新进展使1PB SSD的可能性更接近现实，尽管仍存在重大的工程挑战。另外，由代码执行和工具调用等任务驱动的AI代理日益增长的CPU需求，正促使人们重新评估数据中心传统的CPU与GPU比例，这意味着基础设施的规模应根据实际的计算、内存和延迟需求来确定，而不是固定的比例。 AI
影响 AI代理正在推动对更灵活的数据中心基础设施的需求，超越固定的CPU-GPU比例以适应可变的计算需求。
排序理由 该集群讨论了存储技术的进步和AI基础设施的需求，这些是支持性技术，而非核心AI发布或重大的行业转变。
在 Mastodon — fosstodon.org 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
The world’s first 1PB SSD could become a reality sooner than many people expect. Kioxia and Sandisk have demonstrated next-generation QLC NAND with 2Tb per die and extremely high areal density. With enough NAND dies, the technology could theoretically enable a single SSD with aro…
The world’s first 1PB SSD may be closer than it sounds. Kioxia and Sandisk’s latest QLC NAND reaches 2Tb per die and 37.6 Gb/mm² density, making a roughly 1PB SSD technically achievable with about 4,096 dies. But the NAND is only part of the challenge. Packaging, controllers, the…
AI agents can create significant CPU workloads through code execution, tool calls, database searches, and orchestration. But that doesn’t necessarily mean data centers need a 1:1 CPU-to-GPU ratio. The key issue is workload variability. Instead of counting CPUs versus GPUs, infras…