# Infrastructure - [The modern data centre](https://training.continuumlabs.ai/infrastructure/the-modern-data-centre.md) - [Enhancing Data Centre Efficiency: Strategies to Improve PUE](https://training.continuumlabs.ai/infrastructure/the-modern-data-centre/enhancing-data-centre-efficiency-strategies-to-improve-pue.md) - [TCO of NVIDIA GPUs and falling barriers to entry](https://training.continuumlabs.ai/infrastructure/the-modern-data-centre/tco-of-nvidia-gpus-and-falling-barriers-to-entry.md) - [Maximising GPU Utilisation with Kubernetes and NVIDIA GPU Operator](https://training.continuumlabs.ai/infrastructure/the-modern-data-centre/maximising-gpu-utilisation-with-kubernetes-and-nvidia-gpu-operator.md) - [Data Centres](https://training.continuumlabs.ai/infrastructure/the-modern-data-centre/data-centres.md) - [Liquid Cooling](https://training.continuumlabs.ai/infrastructure/the-modern-data-centre/liquid-cooling.md) - [Servers and Chips](https://training.continuumlabs.ai/infrastructure/servers-and-chips.md) - [The NVIDIA H100 GPU](https://training.continuumlabs.ai/infrastructure/servers-and-chips/the-nvidia-h100-gpu.md) - [NVIDIA H100 NVL](https://training.continuumlabs.ai/infrastructure/servers-and-chips/nvidia-h100-nvl.md): For large language model inference workloads - [Lambda Hyperplane 8-H100](https://training.continuumlabs.ai/infrastructure/servers-and-chips/lambda-hyperplane-8-h100.md) - [NVIDIA DGX Servers](https://training.continuumlabs.ai/infrastructure/servers-and-chips/nvidia-dgx-servers.md) - [NVIDIA DGX-2](https://training.continuumlabs.ai/infrastructure/servers-and-chips/nvidia-dgx-2.md) - [NVIDIA DGX H-100 System](https://training.continuumlabs.ai/infrastructure/servers-and-chips/nvidia-dgx-h-100-system.md): An absolute beast - [NVLink Switch](https://training.continuumlabs.ai/infrastructure/servers-and-chips/nvlink-switch.md): Rapid Communication between GPUs - [Tensor Cores](https://training.continuumlabs.ai/infrastructure/servers-and-chips/tensor-cores.md) - [NVIDIA Grace Hopper Superchip](https://training.continuumlabs.ai/infrastructure/servers-and-chips/nvidia-grace-hopper-superchip.md) - [NVIDIA Grace CPU Superchip](https://training.continuumlabs.ai/infrastructure/servers-and-chips/nvidia-grace-cpu-superchip.md) - [NVIDIA GB200 NVL72](https://training.continuumlabs.ai/infrastructure/servers-and-chips/nvidia-gb200-nvl72.md) - [Hopper versus Blackwell](https://training.continuumlabs.ai/infrastructure/servers-and-chips/hopper-versus-blackwell.md): A comparison between the two latest GPU servers - [HGX: High-Performance GPU Platforms](https://training.continuumlabs.ai/infrastructure/servers-and-chips/hgx-high-performance-gpu-platforms.md) - [ARM Chips](https://training.continuumlabs.ai/infrastructure/servers-and-chips/arm-chips.md) - [ARM versus x86](https://training.continuumlabs.ai/infrastructure/servers-and-chips/arm-versus-x86.md) - [RISC versus CISC](https://training.continuumlabs.ai/infrastructure/servers-and-chips/risc-versus-cisc.md) - [Introduction to RISC-V](https://training.continuumlabs.ai/infrastructure/servers-and-chips/introduction-to-risc-v.md) - [Networking and Connectivity](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity.md) - [Infiniband versus Ethernet](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/infiniband-versus-ethernet.md): Networking Technologies - [NVIDIA Quantum InfiniBand](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/nvidia-quantum-infiniband.md): Networking Solution - [PCIe (Peripheral Component Interconnect Express)](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/pcie-peripheral-component-interconnect-express.md) - [NVIDIA ConnectX InfiniBand adapters](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/nvidia-connectx-infiniband-adapters.md) - [NVMe (Non-Volatile Memory Express)](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/nvme-non-volatile-memory-express.md) - [NVMe over Fabrics (NVMe-oF)](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/nvme-over-fabrics-nvme-of.md): A protocol that enables high-performance, low-latency access to shared storage resources over various network fabrics - [NVIDIA Spectrum-X](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/nvidia-spectrum-x.md) - [NVIDIA GPUDirect](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/nvidia-gpudirect.md) - [Evaluating Modern GPU Interconnect](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/evaluating-modern-gpu-interconnect.md): Ang Li et al - [Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/scalable-hierarchical-aggregation-and-reduction-protocol-sharp.md) - [Next-generation networking in AI environments](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/next-generation-networking-in-ai-environments.md) - [NVIDIA Collective Communications Library (NCCL)](https://training.continuumlabs.ai/infrastructure/networking-and-connectivity/nvidia-collective-communications-library-nccl.md) - [Data and Memory](https://training.continuumlabs.ai/infrastructure/data-and-memory.md) - [NVIDIA BlueField Data Processing Units (DPUs)](https://training.continuumlabs.ai/infrastructure/data-and-memory/nvidia-bluefield-data-processing-units-dpus.md) - [Remote Direct Memory Access (RDMA)](https://training.continuumlabs.ai/infrastructure/data-and-memory/remote-direct-memory-access-rdma.md) - [High Bandwidth Memory (HBM3)](https://training.continuumlabs.ai/infrastructure/data-and-memory/high-bandwidth-memory-hbm3.md): SK Hynix Inc - [Flash Memory](https://training.continuumlabs.ai/infrastructure/data-and-memory/flash-memory.md) - [Model Requirements](https://training.continuumlabs.ai/infrastructure/data-and-memory/model-requirements.md) - [Calculating GPU memory for serving LLMs](https://training.continuumlabs.ai/infrastructure/data-and-memory/calculating-gpu-memory-for-serving-llms.md) - [Transformer training costs](https://training.continuumlabs.ai/infrastructure/data-and-memory/transformer-training-costs.md) - [GPU Performance Optimisation](https://training.continuumlabs.ai/infrastructure/data-and-memory/gpu-performance-optimisation.md) - [Libraries and Complements](https://training.continuumlabs.ai/infrastructure/libraries-and-complements.md) - [NVIDIA Base Command](https://training.continuumlabs.ai/infrastructure/libraries-and-complements/nvidia-base-command.md) - [NVIDIA AI Enterprise](https://training.continuumlabs.ai/infrastructure/libraries-and-complements/nvidia-ai-enterprise.md) - [CUDA - NVIDIA GTC 2024 presentation](https://training.continuumlabs.ai/infrastructure/libraries-and-complements/cuda-nvidia-gtc-2024-presentation.md): Steven Jones' presentation on CUDA at NVIDIA GTC 2024 - [RAPIDs](https://training.continuumlabs.ai/infrastructure/libraries-and-complements/rapids.md) - [RAFT](https://training.continuumlabs.ai/infrastructure/libraries-and-complements/raft.md) - [Vast Data Platform](https://training.continuumlabs.ai/infrastructure/vast-data-platform.md) - [Vast Datastore](https://training.continuumlabs.ai/infrastructure/vast-data-platform/vast-datastore.md) - [Vast Database](https://training.continuumlabs.ai/infrastructure/vast-data-platform/vast-database.md) - [Vast Data Engine](https://training.continuumlabs.ai/infrastructure/vast-data-platform/vast-data-engine.md) - [DASE (Disaggregated and Shared Everything)](https://training.continuumlabs.ai/infrastructure/vast-data-platform/dase-disaggregated-and-shared-everything.md) - [Dremio and VAST Data](https://training.continuumlabs.ai/infrastructure/vast-data-platform/dremio-and-vast-data.md) - [Storage](https://training.continuumlabs.ai/infrastructure/storage.md) - [WEKA: A High-Performance Storage Solution for AI Workloads](https://training.continuumlabs.ai/infrastructure/storage/weka-a-high-performance-storage-solution-for-ai-workloads.md) - [Introduction to NVIDIA GPUDirect Storage (GDS)](https://training.continuumlabs.ai/infrastructure/storage/introduction-to-nvidia-gpudirect-storage-gds.md) - [GDS cuFile API](https://training.continuumlabs.ai/infrastructure/storage/introduction-to-nvidia-gpudirect-storage-gds/gds-cufile-api.md) - [NVIDIA Magnum IO GPUDirect Storage (GDS)](https://training.continuumlabs.ai/infrastructure/storage/nvidia-magnum-io-gpudirect-storage-gds.md) - [Vectors in Memory](https://training.continuumlabs.ai/infrastructure/storage/vectors-in-memory.md) --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://training.continuumlabs.ai/infrastructure.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.