NVIDIA Introduces DSX MaxLPS to Increase AI Factory Performance per Watt

NVIDIA Introduces DSX MaxLPS to Increase AI Factory Performance per Watt NVIDIA has detailed DSX MaxLPS, a combination of dynamic power management software, workload-specific tuning, and 45°C liquid-cooling architectures intended to overcome static provisioning limits and maximize compute capacity inside fixed power envelopes. Replaces static worst-case rack power provisioning with Dynamic Power Software (DPS) that monitors telemetry and dynamically allocates unused power headroom across the fleet. Reduces provisioned rack power from 125 kW to 90 kW on GB200 NVL72 and from 136 kW to 101 kW on Vera Rubin NVL72, allowing 39% and 35% more racks respectively under the same power budget. Utilizes a 45°C liquid-cooling inlet design to reduce reliance on mechanical chillers, driving down PUE and converting saved cooling power into compute power. Projects that 100 MW AI factories can deploy up to 40% more Rubin GPU capacity, supporting up to 40,000 Rubin GPUs and delivering 2 zettaflops of NVFP4 inference. Dynamic Power Software and Stranded Capacity Traditional data center planning provisions racks for peak power draw, stranding unused headroom and leaving as little as 60% of site power for active AI workloads in a representative 100 MW facility. NVIDIA DSX MaxLPS counters this with Dynamic Power Software (DPS), currently in Developer Preview, which models data center topology down to the GPU level. DPS runs a continuous control loop that reclaims unused power from underutilized nodes and redistributes it across managed groups. This management layer can integrate with DSX Exchange, an open-source event bus connecting DPS to building management, grid interfaces, and compute schedulers. Workload Profiling and Performance Validation MaxLPS optimizes job execution via Workload Profile Power Solutions (WPPS) and the Application Performance and Power Manager (APPM), which align memory, frequency, and node behavior with specific compute phases. In evaluations using Kimi-K2.5 on GB200 NVL72 and DeepSeek-R1 on Vera Rubin NVL72, dynamic provisioning lowered rack power allocations while maintaining workload throughput. Performance per watt improved approximately 1.5x on GB200 NVL72 and 1.3x to 1.4x on Vera Rubin NVL72. Under a 100 MW facility plan, this architecture supports 40,000 Rubin GPUs, providing 11 petabytes of HBM4 memory capacity and 1.4 zettaflops of NVFP4 training performance. 45°C Thermal Architecture and Site Sizing At the hardware and site layer, Vera Rubin NVL72 systems operate with a 45°C liquid-cooling inlet. Warmer coolant enables greater reliance on free cooling methods, such as dry coolers, shrinking mechanical chiller usage and lowering annual power usage effectiveness (PUE). Heat transfer is coordinated between the external facility loop and the internal Technology Cooling System (TCS) loop. Facility sizing for MaxLPS is planned around the gross site power and total target rack positions up front, allowing operators to start with higher-power training workloads and gradually populate remaining rack positions as workloads migrate toward inference.

NVIDIA推出DSX MaxLPS架构：通过动态供电与45°C液冷提升AI工厂每瓦性能 面对AI数据中心的电力受限瓶颈，NVIDIA发布了DSX MaxLPS（最大用地电力外壳）技术套件。该方案涵盖芯片、热控、系统与软件层面，旨在打破传统静态机架供电造成的闲置电力冗余，将更多可用电能转化为计算吞吐量，使Vera Rubin NVL72在相同固定电力预算下的GPU部署容量最高提升40%。 在100 MW典型AI工厂中，仅约60%电力能转化为AI计算负载，其余被设施损耗及静态冗余占用 DSX MaxLPS利用动态电力软件（DPS）对数据中心拓扑进行实时监控，将未用电力空间重新分配至其他机架 针对Vera Rubin NVL72与GB200 NVL72测试显示，机架配置功率分别从136 kW/125 kW降至101 kW/90 kW，每瓦性能提升1.3至1.5倍 Rubin NVL72机架支持45°C进水温液冷，可大幅降低对冷水机的依赖，通过自然冷却节约散热功耗供算力使用 MaxLPS确立全生命周期容量规划模型，允许前期根据训练负载部署较少机架，后期随推理业务扩展平滑增容 静态机架配置的电力搁置难题 传统数据中心按最坏情况下的峰值功耗为每个机架分配固定电力，将机架作为孤立电力单元，导致未能使用的闲置电力无法转给邻近机架。以100 MW AI数据中心为例，输入电力中设施开销占20 MW、机架损耗占10 MW、故障与重启等运营损耗占10 MW，仅余60 MW供AI负载。由于训练与推理包含计算激增、内存受限、同步及空闲间隙等不同阶段，实际功耗常低于峰值，形成大量可回收的闲置冗余。 动态电力软件（DPS）的闭环协同 NVIDIA DSX MaxLPS通过处于开发者预览阶段的动态电力软件（DPS）替代静态分配。DPS对从电网、机架、节点到GPU的拓扑建模，实时对比分配额与实际功耗，在预设策略内将未充分利用的电力借给同组设备。配合开源事件总线DSX Exchange，DPS可接入楼宇自控与电网事件。在540 kW的站点测试中，动态配置成功回收170 kW闲置电力，在同一电力预算下实现额外多部署一个机架。 软件每瓦性能优化与基准验证 MaxLPS引入了工作负载配置文件电力方案（WPPS），为推理、训练、显存与计算受限等模式提供经过验证的参数，由APPM管理器及NVIDIA Dynamo进行调度。在DeepSeek-R1（Rubin NVL72）与Kimi-K2.5（GB200 NVL72）推理测试中，机架配置功率分别从136 kW降至101 kW、125 kW降至90 kW，机架部署数量增加35%与39%，每瓦性能分别提高约1.3–1.4倍和1.5倍。在100 MW规模下，结合前期电力规划，可支持4万颗Rubin GPU，提供2 ZFLOPS NVFP4推理算力与11 PB HBM4容量。 45°C温水液冷与生命周期站点规划 Vera Rubin NVL72支持45°C温水直接液体冷却（DLC），使设施能依靠干冷器等室外自然冷却方式散发热量，减少冷水机功耗以优化年均PUE，回收散热能耗供计算使用。此外，MaxLPS建议站点按远期极限容量一次性设计空间、供电管道与网络。第一天面对重载训练时可少部署机架，当硬件生命周期中工作负载逐步转向推理、单机架功耗下降时，运营方无需改造设施即可随时填补机架位。

AI factories are power-constrained industrial systems. The question is no longer how many GPUs fit in a data center, but how much AI output each available megawatt can deliver . For AI inference workloads, this makes application-level performance per watt the key metric for measuring AI factory efficiency. Not every megawatt translates to revenue-generating compute . Power distribution, cooling, networking, storage, backup, and facility overhead take a share of the power before it reaches a GPU. Static rack provisioning exacerbates this: an outdated approach to data center design allocates the maximum power draw per rack to meet worst-case peak demand, even though real workloads have different power needs and may leave some portion of that maximum power unused. Operators reserve additional capacity for failures, operational flexibility, and expansion. In one representative power-budget view examined by NVIDIA, about 60% of delivered site power is allocated to compute for AI output. NVIDIA DSX MaxLPS is a suite of chip, thermal, system, and software technologies that maximizes AI factory throughput within a fixed power budget. MaxLPS stands for Maximum Land Power Shell, the site-level constraints that define an AI factory: land, utility power, and the physical shell holding power, cooling, networking, and compute infrastructure. MaxLPS is designed to optimize these three layers: Dynamic power allocation: Continuously monitors and allocates unused power headroom to GPUs Advanced performance per watt techniques: Software power optimization techniques that improve job-level performance at a fixed power budget 45° C thermal efficiency and site design: Cuts cooling overhead through warm-water liquid cooling, improving power usage effectiveness (PUE) to convert directly into more compute within the same fixed envelope Why static rack provisioning strands power Traditional data center power planning reserves enough power as though every rack could draw its specified maximum power simultaneously. That protects the facility against peak demand, but it treats each rack as an isolated power island. An isolated rack provisioned with excess power cannot lend that unused power to a neighbor that could turn it into tokens. At AI factory scale, facility overhead, rack losses, and operational inefficiencies during failures, restarts, and checkpointing reduce the power available to the AI load. Separately, static rack provisioning can strand headroom within the power allocated to racks. Power reserved for one rack’s peak demand may sit unused while another rack could use it. Dynamic power allocation targets this reclaimable rack-level headroom. Figure 1 shows a power-budget waterfall for a 100 MW AI factory. Of the 100 MW grid input, 20 MW is allocated to facility overhead, 10 MW to rack losses, and 10 MW is unavailable for AI load because of operational inefficiency during failures, restarts, and checkpointing. This leaves 60 MW available for AI load. Each deduction is expressed as a share of the original grid input. Figure 1. Illustrative power-budget waterfall for a 100 MW AI factory Training, post-training, and inference move through compute bursts, memory-bound execution, synchronization, checkpointing, prefill, decode, idle gaps, and network-bound communication. Each phase draws power differently. For stability, a rack must be provisioned to handle the workload’s peak power draw, though actual energy consumption remains below that level for meaningful periods. How does NVIDIA DSX MaxLPS dynamic power allocation work? NVIDIA DSX MaxLPS replaces static power provisioning with dynamic power allocation, using Dynamic Power Software (DPS) (currently in Developer Preview) to monitor and optimize power utilization within the data center’s fixed power budget. DPS is a comprehensive power management system that models the data center topology from the utility level down to racks, nodes, and GPUs. Operators define resource groups, power budgets, and policies that govern how power can be allocated, constrained, and enforced. Within those boundaries, DPS continuously compares allocated power against actual consumption. When GPUs or racks operate below their reserved level, DPS makes that headroom available to others in the same managed group. The site power envelope remains unchanged: DPS extracts more productivity from the available power. DPS runs a continuous control loop, as shown in Figure 2. The system collects GPU-, rack-, and group-level power telemetry, identifies unused power capacity, reallocates power within policy, validates compliance with the approved group power budget, and responds to power events or emergency policies on a best-effort basis. Figure 2. Power-aware MaxLPS GPU allocation workflow The value comes from coordinating many power decisions across the fleet rather than treating every rack as a one-off configuration. When the site budget changes, or when a grid, maintenance, or emergency event alters available power, DPS adapts operating limits across the data center without forcing operators to re-plan every rack by hand. Figure 3 illustrates these efficiency gains by comparing the power utilization of static provisioning with that of MaxLPS dynamic provisioning. Static provisioning strands 170 kW of power, while MaxLPS dynamic provisioning reclaims this headroom, enabling an additional rack deployment within the same 540 kW site power budget. Figure 3. Static power provisioning versus MaxLPS dynamic power provisioning DSX Exchange is an open source event bus for AI factory operations (currently in Developer Preview). It adds an IT/OT event bus connecting DPS and other DSX services to building management systems, electrical power monitoring systems, cooling infrastructure, grid interfaces, and compute schedulers. MaxLPS does not require DSX Exchange to function, but the integration exposes signals such as seasonal cooling headroom and facility power events that DPS can act on. MaxLPS techniques for advanced performance per watt Beyond rack-level power steering, MaxLPS also includes software features that optimize how each GPU uses the power it already has. DSX MaxLPS includes optimized workload profile power solutions (WPPS) for common data center operating modes, including inference, training, memory-bound, and compute-bound configurations. Rather than tuning power, memory, frequency, and other node behavior for every job, operators apply validated profiles that align compute behavior with the workload. The Application Performance and Power Manager (APPM) applies the selected configuration to participating GPUs, while software such as NVIDIA Dynamo can further optimize inter-rack performance and power behavior for inference services. The principle is AI factory optimization: align GPU configuration, application behavior, and serving topology to increase fleet-wide performance and output per watt. Inference provides a clear example because maximizing tokens per second per watt translates the gain into measured workload throughput. The same profiles apply to training and post-training. For NVIDIA Vera Rubin NVL72 AI factories , NVIDIA projects that MaxLPS, combined with data center power planning, can enable up to 40% more Rubin GPU capacity within the same power budget. This means that within a 100 megawatt AI factory, operators can fit 40K Rubin GPUs, enabling the compute capability outlined in Table 1. Specification 100 MW AI factory NVFP4 inference 2 zettaflops NVFP4 training 1.4 zettaflops HBM4 memory capacity 11 petabytes HBM4 memory bandwidth 800 petabytes/second Table 1. Compute capability enabled within a 100 megawatt AI factory with DSX MaxLPS combined with data center power planning Together with measured results on NVIDIA GB200 NVL72 , these results demonstrate how dynamic power management can increase AI factory productivity per megawatt. Figure 4 shows MaxLPS results evaluated using representative inference workloads: Vera Rubin NVL72 was tested with DeepSeek-R1, while GB200 NVL72 was tested with Kimi-K2.5. MaxLPS reduces provisioned rack power from 125 kW to 90 kW on GB200 NVL72 and from 136 kW to 101 kW on Vera Rubin NVL72, enabling 39% and 35% more racks, respectively, within the same power envelope while preserving workload throughput. Performance per watt improves approximately 1.5x on GB200 NVL72 and 1.3–1.4x on Vera Rubin NVL72. Figure 4. Representative inference workload validation for NVIDIA GB200 NVL72 and GB300 NVL72 systems How to size an AI factory site for MaxLPS MaxLPS infrastructure design starts with a fixed gross facility power envelope and works inward to determine the maximum number of GPU rack positions the site can support at the MaxLPS operating point. It also establishes the long-term infrastructure target for power, cooling, space, and network capacity. Day 1 deployment can be lower than this infrastructure limit. A site that begins with a training- or post-training-heavy workload may operate at an average rack power above the MaxLPS average operating point. Because each populated rack consumes more power, fewer racks can initially be deployed within the fixed facility envelope. MaxLPS therefore defines the lifecycle capacity target, not a requirement to populate every rack from the start. As the workload mix shifts toward inference over the hardware lifecycle, average rack power can decline toward the MaxLPS average operating point, creating headroom to populate additional rack positions without increasing the facility power envelope. As Figure 5 shows, MaxLPS sizing anticipates this evolution from the start by selecting the GPU product family, setting the site PUE target at 45° C DLC inlet operation, sizing the east-west network for the MaxLPS GPU count, and deriving the total deployable GPU rack-position count. Figure 5. Sizing an AI factory for MaxLPS within a fixed power envelope. The flow runs from gross site power to PUE-adjusted IT power, then network deduction, GPU rack-position count, and day-one deployment decision The expected day-one workload mix then determines how many of those positions are initially populated. With the associated space, power distribution, cooling, and network capacity configured up front for the full MaxLPS capacity target, operators can add GPUs and racks incrementally as workloads evolve without a later facility retrofit. Why is 45° C liquid cooling important? Software power steering at the data center level is the largest part of the MaxLPS story, but it is not the whole system. MaxLPS also depends on chip-, thermal-, and system-level advances that allow more of the facility power budget to reach the compute infrastructure. Vera Rubin NVL72 racks are designed for 45° C liquid-cooling inlet operation . Warmer liquid may sound counterintuitive, but the goal is to remove heat efficiently while meeting performance, reliability, and lifetime requirements. Higher coolant temperatures can enable facilities to rely more often on “free cooling,” which uses outside air or water to reject heat with minimal mechanical chilling. Depending on the climate and site design, this method can reduce reliance on energy-intensive chillers or water-intensive adiabatic (evaporative) coolers. Chillers remain important during hot conditions and for resilience, but using them only when needed reduces overall cooling power and average annual power usage effectiveness (PUE). Recovering that cooling power for compute is a facilities design problem as much as an operating one. Heat rejection systems are sized for worst-case scenarios, yet many sites require far less chiller compressor power under typical operating conditions, especially when dry coolers can handle a larger portion of the load. If the electrical distribution, dry coolers, and controls are sized for that flexibility, power budgeted for cooling during the worst hour becomes available to compute during much of the year. Operating the technology cooling system loop efficiently The technology cooling system (TCS) loop is the technology-side liquid loop that moves heat between racks, cold plates, and the coolant distribution unit (CDU). It works alongside the broader facility cooling loop, which rejects heat through chillers, dry coolers, cooling towers, or other site equipment. A well-operated TCS loop balances three constraints: Adjusting the coolant flow rate to match GPU thermal demand, allowing the facility loop to absorb variation on the heat rejection side Staying within the designed inlet-temperature and thermal-reliability limits Responding to workload-driven thermal events without creating instability Many facilities use proportional-integral-derivative (PID) control to manage CDU behavior. PID control is common and useful, but it is reactive—it responds only after sensors report a deviation. Large AI factories create fast, synchronized changes in thermal load, so operators often run the loop colder than necessary to preserve margin. A conservative PID-controlled buffer, as described, wastes cooling power that could otherwise support compute. Agentic control systems, including work by Phaidra , use power telemetry and learned control policies to anticipate thermal behavior. A rack with 45 °C thermal efficiency does not require agentic control to operate: the hardware and facility design define the supported thermal envelope, and agentic control is an optimization on top of that. Figure 6 shows the technology cooling loop versus the facility cooling loop. The TCS loop carries heat from GPU cold plates to the coolant distribution unit (CDU), which then transfers it to the facility cooling loop. Facility pumps, chillers, dry coolers, cooling towers, or other site equipment then reject the heat outside the building. Optional monitoring and predictive controls can help optimize efficiency. Figure 6. Facility cooling loop versus technology cooling system loop Get started with NVIDIA DSX MaxLPS validation Physical infrastructure is difficult to change once built. Whether upgrading an existing facility or planning NVIDIA Vera Rubin NVL72 sites, operators should retrofit or design for NVIDIA DSX MaxLPS at the site level, even if they do not deploy every rack on Day 1. That means validating power topology, redundancy, cooling behavior, telemetry availability, policy requirements, network capacity, rack-position optionality, and space or utility constraints before infrastructure decisions are fixed. For teams evaluating the software path, the NVIDIA Dynamic Power Software documentation and NVL72 inference power pilot provide a starting point. A scoped validation compares an unmanaged static baseline against a MaxLPS-managed run using representative workloads, tracking throughput, latency, service error rate, power draw, utilization, and policy compliance. The results show operators when and how fast to scale as they prepare for Vera Rubin NVL72. Facilities teams should engage NVIDIA early to validate site thermal design for 45° C inlet operation, MaxLPS rack-position optionality, and power distribution flexibility before infrastructure is fixed. The site-level design decision should be made early and does not depend on software validation. Software testing can proceed in parallel or follow at a later stage. To turn fixed power into more AI output, power-limited AI factories need a new operating model. Static rack provisioning strands power that real workloads would otherwise turn into useful AI output. NVIDIA DSX MaxLPS addresses this on three fronts: dynamic power allocation software that unlocks stranded rack capacity, performance-per-watt techniques that raise output per watt, and 45° C thermal and facility design that reduces cooling overhead while preserving the physical optionality to land more compute. For operators planning Vera Rubin NVL72 deployments, MaxLPS is a path to prepare sites for up to 40% more Rubin GPU capacity within fixed power budgets. Prepare Vera Rubin NVL72 sites to provision up to 40% more Rubin GPU capacity within your fixed power budget. To get started, read the MaxLPS Overview , explore the NVIDIA Dynamic Power Software documentation , and the NVL72 inference power pilot . Engage NVIDIA early to validate the 45° C thermal design, rack-position optionality, and site-level MaxLPS readiness.