Despite recent smartphones packing 7‑10 kWh batteries thanks to silicon‑carbon anodes, they still require daily charging because high‑resolution screens, power‑hungry chips, 5G radios, and background apps have dramatically increased overall power consumption.
By combining low‑overhead eBPF data collection with AI‑driven diagnosis and an agent‑based execution layer, the authors present a three‑tier system that shifts mobile optimization from peak performance to sustained energy efficiency, achieving sub‑1% monitoring overhead and up to 20% power savings in real‑world video workloads.
The scaleX640 supernode, unveiled at the Wuzhen World Internet Conference, packs 640 AI accelerators into a single rack, delivering unprecedented compute density, energy efficiency, open ecosystem compatibility, and reliability features that enable massive AI model training and inference at scale.
Google reveals that a single Gemini query now consumes only 0.24 Wh of electricity, emits 0.03 g CO₂e and uses about five drops of water, thanks to a comprehensive measurement framework and aggressive optimizations across model architecture, quantization, hardware design, and data‑center operations.
Starting June 20, 2025, the EU will mandate that all smartphones and tablets sold within its market carry a new energy‑efficiency label, detailing EEI grades, battery endurance, durability, repairability, and other performance metrics, aiming to guide consumers and push manufacturers toward more sustainable product designs.
The article analyzes the rapid growth of AI model memory and compute demands, the slow increase of chip memory capacity, and argues that memory bandwidth and energy consumption, rather than raw compute, will dominate AI chip design, emphasizing multi‑tenancy, DSA flexibility, and data‑flow optimization.
This article provides a technical breakdown of AI chip key performance indicators—accuracy, throughput, latency, and energy consumption—explains how MAC and processing‑element design affect these metrics, and outlines design strategies for maximizing throughput while minimizing latency and power use.
This article explains how a mixed‑integer programming tool developed by ByteDance's SYS‑DCD team integrates cost, reliability, delivery speed, and environmental metrics to optimize data‑center planning, reduce power waste, and accelerate deployment across multiple regional scenarios.
As AI accelerates, large‑model compute infrastructure becomes a cornerstone of digital transformation, with specialized accelerators, heterogeneous architectures, massive distributed clusters, intelligent scheduling, soaring costs, energy concerns, software‑hardware co‑design challenges, and data‑privacy issues shaping its future development.
This article provides a comprehensive overview of server CPUs, covering their core role, major architectures such as x86, ARM, POWER, and SPARC, key performance indicators, energy‑saving features, and emerging trends like quantum, photonic computing and AI acceleration.
This article analyzes the technical roadmap for upgrading AI super‑large GPU clusters to support trillion‑parameter multimodal models, covering single‑chip performance, super‑node scaling, DPU‑based compute fusion, energy‑efficient designs, converged storage, high‑throughput networking, and fault‑tolerant checkpoint strategies.
A 2017 study by six Portuguese researchers compared the energy consumption, execution time, and memory usage of 27 programming languages across ten benchmark problems, revealing that faster languages aren't always more energy‑efficient and that compiled languages generally outperform interpreted ones in both speed and power usage.
A research team from Beijing University of Science and Technology tackled the high energy consumption of aluminum electrolysis by developing an AI‑driven prediction model and a cloud‑edge monitoring platform on Huawei Cloud, achieving significant power savings and winning a national competition.
Researchers analyzed 27 programming languages across ten benchmark problems, measuring power consumption, execution time, and memory usage, revealing that faster languages aren't always greener, compiled languages generally consume less energy, and the optimal choice depends on specific performance and resource priorities.
The National Energy Center’s recent green computing evaluation highlighted Ant Group’s innovative cloud‑native scheduling, AI‑driven prediction, and offline mixed‑branch technologies that significantly improve data‑center energy efficiency, reduce carbon emissions, and demonstrate strong promotion value across major enterprises.
The white paper released at the ODCC winter plenary outlines the definition, development stages, key technologies, application scenarios, challenges, and policy recommendations for green computing in China, emphasizing its role in sustainable digital transformation and low‑carbon economic growth.
Although both ARM and x86 processors aim to balance transistor count, program complexity, speed, power consumption, and cost, their fundamentally different architectures—RISC versus CISC—make them non‑interchangeable, leading to distinct trade‑offs in performance, energy efficiency, and application suitability across devices from smartphones to supercomputers.
The article explains the fundamental differences between ARM and x86 CPUs, compares RISC and CISC instruction set philosophies, discusses how each architecture balances transistor count, performance, power consumption and cost, and highlights why ARM dominates low‑power devices while x86 remains prevalent in high‑performance computers.
The 2023 Inclusion·Bund Conference examined the rapid growth of large‑model AI, the resulting GPU shortage, and presented multi‑sector strategies—including policy guidance, ESG standards, AI‑driven optimization, and collaborative standards—to achieve sustainable, energy‑efficient computing across the entire ecosystem.
The 2023 Inclusion·Bund Conference examined the surge in large‑model demand, GPU scarcity, and the resulting energy challenges, presenting policy initiatives, industry collaborations, and AI‑driven techniques aimed at building an end‑to‑end green computing ecosystem across the cloud and data‑center stack.
The 2022 Computing Strengthening Action product catalog presents a comprehensive evaluation of seven server models, detailing their energy‑saving technologies, performance under CPU, memory and storage loads, and classifying them as excellent, recommended, or outstanding products to guide data‑center hardware selection.
The article explains how DPUs, as low‑power, high‑efficiency data‑processing units, complement CPUs and GPUs in modern data centers, reducing total cost of ownership while handling data movement, security, and analytics tasks more effectively than traditional processors.
Bilibili’s next‑generation data center replaces traditional air‑cooling with a hybrid liquid‑cooling system—combining water‑cooled chillers, indirect evaporative cooling, magnetic‑levitation pumps and cold‑plate modules—to raise inlet temperatures, cut fan power, achieve PUE below 1.15, and demonstrate greener, cost‑effective operation while shaping industry standards.
The article examines the growing bottlenecks of large‑scale AI model training caused by the separation of storage and compute, analyzes why conventional GPU architectures cannot keep pace with exponential model growth, and presents in‑memory and near‑memory computing, as well as storage‑compute integration, as promising solutions to boost performance, energy efficiency, and scalability for cloud and edge deployments.
Using a lumped-parameter approach, researchers built a dynamic mathematical model of an office air‑conditioned space and simulated it in MATLAB/Simulink, revealing how solar radiation, occupant count, and supply airflow rate each impact indoor temperature and can guide energy‑efficient HVAC control.
The article examines Armv9’s A710, A715, and A510 cores, detailing their microarchitectural tweaks—branch predictor expansion, dispatch width changes, SVE2 support, and 64‑bit‑only design—while highlighting modest performance and energy gains, code‑compatibility shifts, and evolving core‑count configurations in modern Snapdragon and MediaTek SoCs.
The article examines China's push for green computing as a strategic response to rapid technological change, high energy consumption, and carbon‑neutral goals, highlighting industry challenges, the need for efficient hardware and software solutions, and the role of competitions and talent cultivation in fostering sustainable digital transformation.
This article examines how a 5G cloud‑native RAN built on Intel FlexRAN, Xeon D CPUs and the Chronos AI framework can integrate time‑series ML models to predict traffic and dynamically adjust processor states, achieving 15‑30% overall power savings while maintaining service quality.
This article presents a comprehensive overview of TencentOS’s evolution and its data‑center resource and energy management techniques, covering the OS’s mission for economic and green operation, resource utilization metrics (UTE, QOS, ROI), current management technologies, energy‑aware scheduling, performance results, and future development directions.
This article reviews a decade of Google TPU development, highlighting ten technical and architectural lessons, the hardware's impact on the AI industry, performance and energy‑efficiency improvements, and strategies for reducing machine‑learning carbon footprints.
This article explains how to calculate data‑center floor area, select appropriate UPS capacity, determine cooling load, estimate total power consumption, and plan space requirements, providing practical formulas and guidelines for efficient and reliable data‑center operation.
Bilibili’s SYS team is advancing green data‑center technology by aligning with national energy‑efficiency mandates, co‑authoring cold‑plate liquid‑cooling standards, and piloting the custom R2‑AZ2 Innovation Room Phase‑1, which combines indirect evaporative and magnetic‑levitation hybrid cooling to achieve PUEs as low as 1.13 while testing AI‑driven O&M tools for future large‑scale, low‑carbon deployment.
A recent paper examines the energy consumption, runtime, and memory usage of software across 27 programming languages, revealing that compiled languages like C and C++ are the most energy‑efficient, while highlighting the growing importance of energy awareness in software development and its implications for embedded, mobile, and cryptocurrency applications.
The article explains how liquid cooling—including immersion, cold‑plate, and hybrid air‑liquid solutions—offers higher heat‑transfer efficiency and lower PUE compared to traditional air cooling, addressing energy‑saving goals, regional regulations, and water‑scarcity challenges in modern data centers.
This guide explains how to calculate data‑center floor area, size UPS units and cooling systems, estimate total power demand, and implement effective electricity monitoring to improve energy efficiency in large‑scale computer rooms.
The Chinese government has officially launched the “East Data West Computing” program, establishing national compute‑hub nodes across multiple regions, building ten data‑center clusters, and outlining energy‑efficiency standards to shift massive computing workloads from the east to the resource‑rich west while supporting AI and digital transformation.
Alibaba Cloud’s 10 kV AC‑input DC UPS system and high‑elasticity cooling technology were officially listed in the Ministry of Industry and Information Technology’s 2021 National Telecom Energy‑Saving Product Catalog, highlighting significant efficiency gains and supporting China’s green data‑center initiatives.
A recent analysis comparing 27 programming languages shows that despite Python's popularity, it and Perl consume far more energy than languages like C, C++, and Rust, challenging common assumptions about performance and environmental impact.
This article provides a comprehensive overview of data centers, covering definitions, classifications, cloud‑compatible designs, scale and availability tiers, lifecycle stages, energy consumption metrics, power and cooling technologies, network architectures, and emerging industry trends.
This article explains three power‑saving techniques—clock gating, power gating, and protocol‑level power management—for PCI Express IP in system‑on‑chip designs, detailing their impact on dynamic and static power, implementation challenges, and how designers can achieve high energy efficiency while meeting fast recovery requirements.
On September 16, Alibaba inaugurated the globally largest fully immersed liquid‑cooled data center in Zhejiang, featuring single‑phase immersion cooling that achieves a PUE of 1.09, supports the new‑infrastructure drive, and marks a major step toward greener, high‑performance cloud computing in China.
Alibaba's Zhejiang liquid‑cooled data center became China’s first 5A‑rated green data center, showcasing advanced immersion cooling, HVDC power, distributed redundancy and AI‑optimized management that deliver a design‑year PUE of 1.12 and set a new benchmark for energy‑efficient cloud infrastructure.
The article examines Microsoft's bold Natick project of placing servers on the ocean floor, highlighting how underwater data centers can cut cooling costs, reduce latency for coastal users, and address the growing energy and space challenges of traditional land‑based data centers.
This article explains how UCloud reduced its data center Power Usage Effectiveness to around 1.3 through advanced cooling technologies, energy‑efficient UPS, modeling, and other green measures, while also reviewing international benchmarks and China’s regulatory targets for sustainable data centers.
The article examines the rapid growth of information system energy consumption, defines green computing, outlines its research focus—especially on data centers, cloud computing and servers—and analyzes a range of energy‑saving technologies such as DVFS, heterogeneous computing, liquid cooling, rack‑level servers, advanced power supply, backplane cooling and fluorocarbon pump AC systems.
The article explores how the superior efficiency of electric vehicles, the inevitable limits of fossil‑fuel energy, and the need for high‑quality, closed‑loop data feedback together shape a sustainable path for humanity and business alike.
Beijing’s 2018 prohibition catalogue tightens rules on new data centers, limiting construction to ultra‑efficient facilities, which reshapes the local IDC market, spurs green‑focused opportunities in surrounding regions, and underscores the growing importance of energy‑saving cloud infrastructure.
Google’s DeepMind AI system autonomously optimizes data‑center cooling, reducing energy consumption by up to 40%, and its success is prompting plans to extend the technology to national power grids and other large‑scale infrastructure, showcasing AI’s potential for massive environmental impact.
At the ICT China High‑Level Forum, JD's self‑built Suqian data center was honored with the 2017 Data Center Annual Infrastructure Design Innovation Award, highlighting its advanced power, cooling, and energy‑efficiency technologies that set a new benchmark for high‑performance data centers in China.
This article examines why electricity dominates data‑center operating costs, outlines practical green‑IT measures—including enclosure design, airflow management, lighting, and renewable power—and presents three detailed case studies that illustrate how modular design, cold‑aisle containment, and innovative cooling can dramatically reduce PUE and overall energy consumption.
JD’s self‑constructed data center in Suqian, Jiangsu, combines innovative free‑cooling, high‑efficiency chillers, redundant power architecture, and advanced monitoring to deliver over 180 days of natural cooling, reduce PUE to ≤1.3, and ensure continuous operation with robust backup systems.
The 2017 TOP500 list shows China's Sunway TaihuLight leading with 125,435.9 TFlop/s, highlights the role of Linpack‑based rankings, details vendor participation, and contrasts performance‑focused TOP500 with energy‑focused Green500, underscoring China's growing high‑performance computing capabilities and the importance of power efficiency.
The U.S. Department of Energy’s annual Connected Lighting workshop in Santa Clara emphasized industry collaboration on standards and testbeds, showcasing how LED‑based, data‑driven lighting systems can improve energy efficiency, enable new services, and integrate with IoT platforms for smarter building management.
In just one year, Alibaba built two world‑class green data centers in Zhangbei, harnessing abundant wind and solar power to achieve record‑low PUE, modular construction, and advanced cooling, thereby boosting cloud, big data and AI services for millions of enterprises in northern China.
The article examines various unconventional data‑center cooling methods—including lake‑water, seawater, waste‑water, mineral‑oil, and hot‑water systems—highlighting their impact on Power Usage Effectiveness (PUE) and overall energy efficiency across major tech companies.
This article explains China's green data‑center policies, the importance of PUE, and demonstrates through calculations and real‑world Dalian case studies how natural cooling can halve cooling energy use, lower PUE from 2.5 to 2.0, and save millions in electricity bills.
On September 8, Alibaba launched its QianDao Lake data center in Zhejiang, employing sealed deep‑lake water cooling that eliminates electric refrigeration for 90% of operation, achieving a PUE as low as 1.17, a water‑use efficiency of 0.197, and showcasing custom hardware and renewable energy integration to support its cloud and big‑data services.
Facebook’s Cold Storage system redesigns hardware and software to store rarely accessed data with up to 75% lower power consumption, modular 2U racks holding 30 disks, Reed‑Solomon erasure coding for cheap redundancy, and a self‑healing “anti‑entropy” process that improves performance as the system scales.
This article examines Facebook's cold storage system, detailing how the company redesigned hardware and software to slash power consumption, improve reliability with Reed‑Solomon coding, mitigate bit‑rot, and balance loads while supporting massive photo archives in energy‑constrained data centers.
