Articles tagged with "power-efficiency"
Spectra supercomputer tests adaptive chips for nuclear security
Sandia National Laboratories has introduced Spectra, a prototype supercomputer developed in partnership with NextSilicon, designed to revolutionize national security simulations through adaptive, efficiency-focused computing. Unlike traditional CPU and GPU-based systems, Spectra employs 128 Maverick-2 dual-die accelerators—experimental chips that dynamically analyze and prioritize code tasks in real time to enhance performance while reducing power consumption. This approach aims to improve the speed and efficiency of complex simulations critical to maintaining the safety and reliability of the U.S. nuclear deterrent without underground testing. Spectra is the second system under Sandia’s Vanguard program, which tests cutting-edge technologies for potential large-scale deployment. Following the success of Astra, the first Vanguard machine that validated Arm processors for scientific workloads, Spectra seeks to demonstrate the viability of intelligence-driven, adaptive computing architectures. Early benchmarks, including HPCG, LAMMPS, and SPARTA, have shown promising results without requiring users to rewrite applications, potentially lowering the cost and complexity of
energysupercomputingadaptive-chipsnuclear-securityhigh-performance-computingpower-efficiencynational-laboratoriesPowerLattice attracts investment from ex-Intel CEO Pat Gelsinger for its power saving chiplet
PowerLattice, a startup founded in 2023 by veteran engineers from Qualcomm, NUVIA, and Intel, has developed a novel power delivery chiplet that reduces computer chip power consumption by over 50%. This innovation addresses the growing demand for energy-efficient semiconductor solutions amid the increasing compute capacity needs driven by AI workloads and large language models. The company recently emerged from stealth with a $25 million Series A funding round led by Playground Global and Celesta Capital, bringing total funding to $31 million. Pat Gelsinger, former Intel CEO and general partner at Playground Global, endorsed PowerLattice’s technology, highlighting the team's expertise and the significance of their power delivery approach. PowerLattice’s chiplet works by bringing power closer to the processor, thereby minimizing energy loss. The startup has reached a key milestone with its first batch of chiplets produced by TSMC and is currently undergoing testing with an unnamed manufacturer. The company plans to expand testing to other potential customers, including major chipmakers
energysemiconductorpower-efficiencychipletAI-hardwarePowerLatticePat-GelsingerQualcomm’s new AI systems promise 10x bandwidth, lower power use
Qualcomm has unveiled its next-generation AI inference accelerators, the AI200 and AI250, designed to significantly enhance data center generative AI performance with improved efficiency and scalability. The AI200 card supports 768 GB of LPDDR memory, enabling large-scale language and multimodal model inference with a focus on lowering total cost of ownership (TCO). Building on this, the AI250 introduces a near-memory computing architecture that delivers over 10 times higher effective memory bandwidth and substantially reduces power consumption, facilitating more efficient disaggregated AI inferencing. Both solutions feature direct liquid cooling, PCIe for scale-up, Ethernet for scale-out, and a rack-level power consumption of 160 kW, targeting hyperscaler-grade performance with sustainability in mind. Qualcomm emphasizes seamless integration through a rich software stack and open ecosystem, supporting major AI frameworks and enabling one-click deployment of pre-trained models. Their software tools, including the Efficient Transformers Library and AI Inference Suite, allow developers to operationalize AI models easily
energyAI-acceleratorsdata-centerspower-efficiencymemory-bandwidthQualcomm-AIgenerative-AISam Altman-backed Oklo to cool AI data centers with new nuclear tech
Oklo, a nuclear technology company backed by Sam Altman, has partnered with Vertiv, a leader in digital infrastructure, to develop an integrated power and cooling system for hyperscale and colocation data centers. This system will leverage Oklo’s small modular reactors (SMRs) to generate steam and electricity, combined with Vertiv’s thermal management technology, aiming to optimize both power and cooling efficiently and sustainably. The collaboration seeks to address common data center challenges such as high energy demand, reliance on power grids, and environmental impact by providing a reliable, carbon-free energy source that can be located near data centers for improved performance and scalability. The partnership comes amid the rapid growth of AI and high-performance computing, which significantly increases power consumption in data centers. Oklo’s SMRs are designed for flexibility and quick adaptation to changing energy needs, enabling continuous, stable power supply critical for data center operations. By integrating power generation and cooling solutions from the outset, Oklo and Vertiv aim to enhance energy efficiency
energynuclear-energydata-centerscooling-technologysmall-modular-reactorsAI-infrastructurepower-efficiencyChina’s dual-faced solar cells hit 23.4% efficiency record, cut power loss dramatically
Researchers in China have developed an innovative approach to significantly improve bifacial perovskite solar cells (Bi-PSCs), achieving a record power conversion efficiency (PCE) of 23.4% while dramatically reducing photon loss—a key factor limiting these cells’ performance. By identifying photon loss as the primary cause of performance degradation, the team introduced a high-quality thick-film deposition method that regulates perovskite crystallization dynamics. This method employs a multifunctional additive, 1-ethyl-3-guanidinium thiourea hydrochloride (EGTHCl), to control nucleation and crystallization in highly concentrated precursor solutions, resulting in dense, uniform, and defect-free perovskite films. The enhanced Bi-PSCs not only reached a record efficiency but also demonstrated remarkable stability, retaining over 80% of their initial performance after more than 2,000 hours of continuous light exposure. This advancement addresses the inherent design challenge of bifacial cells, which use semi-transparent rear
energysolar-cellsperovskitephotovoltaicphoton-lossbifacial-solar-cellspower-efficiencyUltra-efficient amplifier slashes quantum computer's power use by 90%
Researchers at Chalmers University of Technology have developed a novel amplifier for quantum computers that reduces power consumption by 90% compared to existing models. This breakthrough addresses a critical challenge in quantum computing—decoherence, which occurs when heat and electromagnetic interference from amplifiers disrupt qubit states during measurement. The new amplifier activates only when needed, significantly cutting heat generation and thus minimizing errors in qubit readout. This advancement could enable the construction of larger, more stable quantum computers with increased numbers of qubits and improved computational performance. The key innovation lies in the amplifier’s pulse-operated functionality, which allows it to switch on briefly and precisely to read qubit signals without continuous power use. The team overcame the challenge of rapid activation by implementing a smart control system using genetic programming, enabling the amplifier to respond within 35 nanoseconds. This highly sensitive, low-noise semiconductor amplifier represents the most efficient transistor-based design currently achievable and is expected to help overcome a major technical bottleneck in scaling quantum computers.
energyquantum-computingamplifier-technologypower-efficiencydecoherence-reductionmicrowave-electronicssemiconductor-amplifiersMIT builds new superconducting chip to power future quantum computers
Researchers at MIT’s Plasma Science and Fusion Center have developed a superconducting diode (SD)-based rectifier chip that converts alternating current (AC) to direct current (DC) at cryogenic temperatures, aiming to streamline power delivery in superconducting classical and quantum computers. This innovation addresses a critical challenge in quantum computing: reducing thermal and electromagnetic noise caused by numerous wires connecting ultra-cold components to ambient temperature systems. By integrating four superconducting diodes on a single chip, the team achieved efficient AC to DC conversion, potentially enhancing qubit stability and reducing interference, which is vital for the practical realization of quantum computers. Beyond quantum computing, the superconducting diode technology has broader applications, including serving as isolators or circulators to protect qubit signals and playing a role in dark matter detection circuits used in experiments at CERN and Berkeley National Laboratory. This advancement promises to make superconducting electronics more energy-efficient and practical, potentially revolutionizing computing power in the era of increasing demands from technologies like artificial intelligence. The
energysuperconducting-electronicsquantum-computingsuperconducting-diodepower-efficiencycryogenic-technologyMIT-research