能源盲点:NVIDIA 旗舰边缘 AI 硬件无法支持进程级能源归因
The Energy Blind Spot: NVIDIA's Flagship Edge AI Hardware Cannot Support Process-Level Energy Attribution
AI总结:
本文审计了 ASUS Ascent GX10 (GB10 SoC) 平台的能源可观测性,发现其缺乏 CPU 能源计数器等关键接口,导致无法像 x86 的 RAPL 那样进行进程级能源归因,并提出通过外部直流计量和 GPU 减法进行校准的临时方案,呼吁将能源可观测性作为硬件的一等要求。
摘要:
代理型 AI 工作负载——其中单个用户目标触发多步编排、工具调用、重试和故障恢复——正被瞄准用于边缘部署,NVIDIA、戴尔、惠普、华硕、微星、宏碁和技嘉都将在 2026 年出货基于 GB10 的桌面 AI 系统。我们最近证明,编排结构主导了代理型能源成本,工作流每个成功目标消耗的能源是线性基线的 4.33 倍,而多步推理任务的 OOI 达到 7.63 倍。另外,Rajat 等人表明,在代理型工作负载中,CPU 端处理占总延迟的 90.6%,占总动态能源的 44%。我们报告了对 ASUS Ascent GX10 (GB10 SoC) 的系统性能源可观测性审计,发现该平台通过任何支持的软件接口都不暴露 CPU 能源计数器、INA 电源轨监视器、IPMI/BMC 和 SCMI powercap 协议。唯一的设备上能源遥测是通过 NVML 的瞬时 GPU 功率。我们进一步发现,联发科固件已经通过未记录的 ACPI 接口 (SPBM) 在内部计算每轨能源,但 NVIDIA 表示“没有计划暴露 CPU 轨信息”。因此,通过支持的接口,无法在此平台上重现像 x86 通过 RAPL 执行的设备上每进程能源归因。我们形式化了能源归因 AI 的硬件需求规范,提出了使用外部直流计量结合 GPU 减法的临时校准桥接,并确定了通过 SCMI powercap 的标准轨道路径。我们的发现激励低碳计算社区将能源可观测性作为硬件的头等要求。
英文摘要:
Agentic AI workloads - where a single user goal triggers multi-step orchestration, tool calls, retries, and failure recovery - are being targeted for edge deployment, with NVIDIA, Dell, HP, ASUS, MSI, Acer, and Gigabyte all shipping GB10-based desktop AI systems in 2026. Prior work shows orchestration structure dominates agentic energy cost and CPU-side processing accounts for up to 44% of total dynamic energy. We report a systematic energy-observability audit of the ASUS Ascent GX10 (GB10 SoC) and find that the platform exposes no CPU energy counter, no INA power-rail monitor, no IPMI/BMC, and no SCMI powercap protocol through any supported software interface. The only on-device energy telemetry is instantaneous GPU power via NVML. We further discover that the MediaTek firmware already computes per-rail energy internally via an undocumented ACPI interface (SPBM), but NVIDIA states there are "no plans to expose CPU rail information." On-device per-process energy attribution - as performed on x86 via RAPL - is therefore not reproducible on this platform through supported interfaces. We formalize a hardware requirements specification for energy-attributed AI, propose an interim calibration bridge for per-domain energy decomposition - confirmed on the Acer Veriton GN100 where CPU energy accumulators are live - and identify a standards-track path via SCMI powercap. Our findings motivate the low-carbon computing community to demand energy observability as a first-class hardware requirement.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。