Unverified50% confidenceFactExact time
GKE's GPU metrics come from Google's in-house nvidia-metrics-collector container, which runs inside the device-plugin DaemonSet rather than the community-standard dcgm-exporter.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/3/2026
First Seen
Valid until: 12/2/2026
Sources
Related Entities
Related Claims
UnverifiedNVIDIA提供官方dcgm-exporter容器，可将DCGM指标暴露给Prometheus进行抓取，通常与Prometheus和Grafana配合构建可视化监控大盘63% similarUnverifiedNVIDIA的k8s-device-plugin会将每块物理GPU注册为不可分割的资源单元（nvidia.com/gpu: 1），调度器以此为原子单位分配62% similarVerifiedGGUF是llama.cpp生态广泛使用的格式，支持2bit至8bit灵活量化粒度，对CPU推理友好62% similarUnverifiedGPU显存内容存储在设备侧DRAM（如GDDR6X或HBM3），驱动以私有数据结构管理上下文，内核模块nvidia.ko维护独立状态机，导致CRIU无法捕获GPU端计算状态61% similarUnverifiedML训练的完整GPU计算栈包括GPU驱动、CUDA Runtime、cuDNN、cuBLAS以及NCCL多GPU集合通信库61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/851091API
curl https://kongchang.com/api/v1/knowledge/claims/851091MCP
get_claim(id=851091)