Insights
Technical analyses on storage acceleration, AI datacenter construction and efficiency, and the domestic-GPU ecosystem. Articles are generated by Xinyuanyi's content engine and pass automated QC — factual numbers must come from signed test reports or whitelisted sources. Translated from the Chinese originals.
多轮对话KV Cache应用案例
多轮对话场景下 KV Cache 复用技术的实际应用案例
多轮对话KV Cache复用效果评估
多轮对话场景下KV Cache复用策略的实际效果评估
KV Cache外置存储技术实现
大模型推理KV Cache外置存储方案的技术实现路径有哪些
KV Cache池化共享资源利用率
大型模型推理中，KV Cache池化共享如何提升资源利用率
国产KV Cache发展趋势
国产 KV Cache 的发展趋势与市场需求分析
国产KV Cache多数据中心性能损耗
国产 KV Cache 产品在多数据中心部署中的性能损耗分析
KV Cache显存推理
显存除了待机都是满载跑吗：KV Cache 如何撑起推理的显存水位
KV Cache预填充解码存储配置
KV Cache 预填充与解码阶段的存储配置差异
SSDKV Cache性能优化
SSD-Based KV Cache Optimization: How to Balance Capacity and Performance
大模型本地部署KV Cache内存显存
Local LLM Deployment: KV Cache Offloading to Memory and the Boundaries of VRAM Residency
昇腾910B平台NVIDIAGPU能耗
A Measured Perspective on Inference Storage Energy Consumption: Ascend 910B vs. NVIDIA GPU
GPU空闲碎片化
How GPU Idle Time and Fragmentation Drag Down Inference Throughput
NVIDIAGPU能耗
Power Composition of Inference Storage and Controllable Optimization Paths
NVMe-oFRDMA时延对比
NVMe-oF vs. RDMA: A Latency Analysis for Inference Storage
KV Cache分层存储吞吐提升
How Tiered KV Cache Storage Improves LLM Inference Throughput
KV CacheDRAM效率
KV Cache and DRAM Co-Optimization Improves Cache Efficiency
KV Cache数据预取延迟
How KV Cache Prefetching Reduces Storage Latency
推理一体机成本效益KV Cache
Cost-Benefit Breakdown of Inference Appliances and KV Cache Accelerated Storage
算力租赁服务绿色节能降低PUE
How Compute Rental Services Can Lower PUE
KV Cache空间局部性优化
KV Cache Spatial Locality Optimization: From Paging to Tiered Storage
KV Cache外置TTFT性能提升
How KV Cache Externalization Reduces LLM Inference TTFT
算力租赁成本结构按需包年
Is "Compute Freedom" a False Proposition? A Cost Structure Breakdown of On-Demand vs. Annual Subscription
算力租赁成本优化动态伸缩
Optimizing Compute Rental Costs: A Comparison of On-Demand Allocation vs. Dynamic Scaling
算力租赁数据传输效率优化方法