LogoLogo
Continuum WebsiteContinuum ApplicationsContinuum KnowledgeAxolotl Platform
Continuum Knowledge
Continuum Knowledge
  • Continuum
  • Data
    • Datasets
      • Pre Training Data
      • Types of Fine Tuning
      • Self Instruct Paper
      • Self-Alignment with Instruction Backtranslation
      • Systematic Evaluation of Instruction-Tuned Large Language Models on Open Datasets
      • Instruction Tuning
      • Instruction Fine Tuning - Alpagasus
      • Less is More For Alignment
      • Enhanced Supervised Fine Tuning
      • Visualising Data using t-SNE
      • UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
      • Training and Evaluation Datasets
      • What is perplexity?
  • MODELS
    • Foundation Models
      • The leaderboard
      • Foundation Models
      • LLama 2 - Analysis
      • Analysis of Llama 3
      • Llama 3.1 series
      • Google Gemini 1.5
      • Platypus: Quick, Cheap, and Powerful Refinement of LLMs
      • Mixtral of Experts
      • Mixture-of-Agents (MoA)
      • Phi 1.5
        • Refining the Art of AI Training: A Deep Dive into Phi 1.5's Innovative Approach
      • Phi 2.0
      • Phi-3 Technical Report
  • Training
    • The Fine Tuning Process
      • Why fine tune?
        • Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?
        • Explanations in Fine Tuning
      • Tokenization
        • Tokenization Is More Than Compression
        • Tokenization - SentencePiece
        • Tokenization explore
        • Tokenizer Choice For LLM Training: Negligible or Crucial?
        • Getting the most out of your tokenizer for pre-training and domain adaptation
        • TokenMonster
      • Parameter Efficient Fine Tuning
        • P-Tuning
          • The Power of Scale for Parameter-Efficient Prompt Tuning
        • Prefix-Tuning: Optimizing Continuous Prompts for Generation
        • Harnessing the Power of PEFT: A Smarter Approach to Fine-tuning Pre-trained Models
        • What is Low-Rank Adaptation (LoRA) - explained by the inventor
        • Low Rank Adaptation (Lora)
        • Practical Tips for Fine-tuning LMs Using LoRA (Low-Rank Adaptation)
        • QLORA: Efficient Finetuning of Quantized LLMs
        • Bits and Bytes
        • The Magic behind Qlora
        • Practical Guide to LoRA: Tips and Tricks for Effective Model Adaptation
        • The quantization constant
        • QLORA: Efficient Finetuning of Quantized Language Models
        • QLORA and Fine-Tuning of Quantized Language Models (LMs)
        • ReLoRA: High-Rank Training Through Low-Rank Updates
        • SLoRA: Federated Parameter Efficient Fine-Tuning of Language Models
        • GaLora: Memory-Efficient LLM Training by Gradient Low-Rank Projection
      • Hyperparameters
        • Batch Size
        • Padding Tokens
        • Mixed precision training
        • FP8 Formats for Deep Learning
        • Floating Point Numbers
        • Batch Size and Model loss
        • Batch Normalisation
        • Rethinking Learning Rate Tuning in the Era of Language Models
        • Sample Packing
        • Gradient accumulation
        • A process for choosing the learning rate
        • Learning Rate Scheduler
        • Checkpoints
        • A Survey on Efficient Training of Transformers
        • Sequence Length Warmup
        • Understanding Training vs. Evaluation Data Splits
        • Cross-entropy loss
        • Weight Decay
        • Optimiser
        • Caching
      • Training Processes
        • Extending the context window
        • PyTorch Fully Sharded Data Parallel (FSDP)
        • Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation
        • YaRN: Efficient Context Window Extension of Large Language Models
        • Sliding Window Attention
        • LongRoPE
        • Reinforcement Learning
        • An introduction to reinforcement learning
        • Reinforcement Learning from Human Feedback (RLHF)
        • Direct Preference Optimization: Your Language Model is Secretly a Reward Model
  • INFERENCE
    • Why is inference important?
      • Grouped Query Attention
      • Key Value Cache
      • Flash Attention
      • Flash Attention 2
      • StreamingLLM
      • Paged Attention and vLLM
      • TensorRT-LLM
      • Torchscript
      • NVIDIA L40S GPU
      • Triton Inference Server - Introduction
      • Triton Inference Server
      • FiDO: Fusion-in-Decoder optimised for stronger performance and faster inference
      • Is PUE a useful measure of data centre performance?
      • SLORA
  • KNOWLEDGE
    • Vector Databases
      • A Comprehensive Survey on Vector Databases
      • Vector database management systems: Fundamental concepts, use-cases, and current challenges
      • Using the Output Embedding to Improve Language Models
      • Decoding Sentence-BERT
      • ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
      • SimCSE: Simple Contrastive Learning of Sentence Embeddings
      • Questions Are All You Need to Train a Dense Passage Retriever
      • Improving Text Embeddings with Large Language Models
      • Massive Text Embedding Benchmark
      • RocketQAv2: A Joint Training Method for Dense Passage Retrieval and Passage Re-ranking
      • LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders
      • Embedding and Fine-Tuning in Neural Language Models
      • Embedding Model Construction
      • Demystifying Embedding Spaces using Large Language Models
      • Fine-Tuning Llama for Multi-Stage Text Retrieval
      • Large Language Model Based Text Augmentation Enhanced Personality Detection Model
      • One Embedder, Any Task: Instruction-Finetuned Text Embeddings
      • Vector Databases are not the only solution
      • Knowledge Graphs
        • Harnessing Knowledge Graphs to Elevate AI: A Technical Exploration
        • Unifying Large Language Models and Knowledge Graphs: A Roadmap
      • Approximate Nearest Neighbor (ANN)
      • High Dimensional Data
      • Principal Component Analysis (PCA)
      • Vector Similarity Search - HNSW
      • FAISS (Facebook AI Similarity Search)
      • Unsupervised Dense Retrievers
    • Retrieval Augmented Generation
      • Retrieval-Augmented Generation for Large Language Models: A Survey
      • Fine-Tuning or Retrieval?
      • Revolutionising Information Retrieval: The Power of RAG in Language Models
      • A Survey on Retrieval-Augmented Text Generation
      • REALM: Retrieval-Augmented Language Model Pre-Training
      • Retrieve Anything To Augment Large Language Models
      • Generate Rather Than Retrieve: Large Language Models Are Strong Context Generators
      • Active Retrieval Augmented Generation
      • DSPy: LM Assertions: Enhancing Language Model Pipelines with Computational Constraints
      • DSPy: Compiling Declarative Language Model Calls
      • DSPy: In-Context Learning for Extreme Multi-Label Classification
      • Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs
      • HYDE: Revolutionising Search with Hypothetical Document Embeddings
      • Enhancing Recommender Systems with Large Language Model Reasoning Graphs
      • Retrieval Augmented Generation (RAG) versus fine tuning
      • RAFT: Adapting Language Model to Domain Specific RAG
      • Summarisation Methods and RAG
      • Lessons Learned on LLM RAG Solutions
      • Stanford: Retrieval Augmented Language Models
      • Overview of RAG Approaches with Vector Databases
      • Mastering Chunking in Retrieval-Augmented Generation (RAG) Systems
    • Semantic Routing
    • Resource Description Framework (RDF)
  • AGENTS
    • What is agency?
      • Rephrase and Respond: Let Large Language Models Ask Better Questions for Themselves
      • Types of Agents
      • The risk of AI agency
      • Understanding Personality in Large Language Models: A New Frontier in AI Psychology
      • AI Agents - Reasoning, Planning, and Tool Calling
      • Personality and Brand
      • Agent Interaction via APIs
      • Bridging Minds and Machines: The Legacy of Newell, Shaw, and Simon
      • A Survey on Language Model based Autonomous Agents
      • Large Language Models as Agents
      • AI Reasoning: A Deep Dive into Chain-of-Thought Prompting
      • Enhancing AI Reasoning with Self-Taught Reasoner (STaR)
      • Exploring the Frontier of AI: The "Tree of Thoughts" Framework
      • Toolformer: Revolutionising Language Models with API Integration - An Analysis
      • TaskMatrix.AI: Bridging Foundational AI Models with Specialised Systems for Enhanced Task Completion
      • Unleashing the Power of LLMs in API Integration: The Rise of Gorilla
      • Andrew Ng's presentation on AI agents
      • Making AI accessible with Andrej Karpathy and Stephanie Zhan
  • Regulation and Ethics
    • Regulation and Ethics
      • Privacy
      • Detecting AI Generated content
      • Navigating the IP Maze in AI: The Convergence of Blockchain, Web 3.0, and LLMs
      • Adverse Reactions to generative AI
      • Navigating the Ethical Minefield: The Challenge of Security in Large Language Models
      • Navigating the Uncharted Waters: The Risks of Autonomous AI in Military Decision-Making
  • DISRUPTION
    • Data Architecture
      • What is a data pipeline?
      • What is Reverse ETL?
      • Unstructured Data and Generatve AI
      • Resource Description Framework (RDF)
      • Integrating generative AI with the Semantic Web
    • Search
      • BM25 - Search Engine Ranking Function
      • BERT as a reranking engine
      • BERT and Google
      • Generative Engine Optimisation (GEO)
      • Billion-scale similarity search with GPUs
      • FOLLOWIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions
      • Neural Collaborative Filtering
      • Federated Neural Collaborative Filtering
      • Latent Space versus Embedding Space
      • Improving Text Embeddings with Large Language Models
    • Recommendation Engines
      • On Interpretation and Measurement of Soft Attributes for Recommendation
      • A Survey on Large Language Models for Recommendation
      • Model driven recommendation systems
      • Recommender AI Agent: Integrating Large Language Models for Interactive Recommendations
      • Foundation Models for Recommender Systems
      • Exploring the Impact of Large Language Models on Recommender Systems: An Extensive Review
      • AI driven recommendations - harming autonomy?
    • Logging
      • A Taxonomy of Anomalies in Log Data
      • Deeplog
      • LogBERT: Log Anomaly Detection via BERT
      • Experience Report: Deep Learning-based System Log Analysis for Anomaly Detection
      • Log-based Anomaly Detection with Deep Learning: How Far Are We?
      • Deep Learning for Anomaly Detection in Log Data: A Survey
      • LogGPT
      • Adaptive Semantic Gate Networks (ASGNet) for log-based anomaly diagnosis
  • Infrastructure
    • The modern data centre
      • Enhancing Data Centre Efficiency: Strategies to Improve PUE
      • TCO of NVIDIA GPUs and falling barriers to entry
      • Maximising GPU Utilisation with Kubernetes and NVIDIA GPU Operator
      • Data Centres
      • Liquid Cooling
    • Servers and Chips
      • The NVIDIA H100 GPU
      • NVIDIA H100 NVL
      • Lambda Hyperplane 8-H100
      • NVIDIA DGX Servers
      • NVIDIA DGX-2
      • NVIDIA DGX H-100 System
      • NVLink Switch
      • Tensor Cores
      • NVIDIA Grace Hopper Superchip
      • NVIDIA Grace CPU Superchip
      • NVIDIA GB200 NVL72
      • Hopper versus Blackwell
      • HGX: High-Performance GPU Platforms
      • ARM Chips
      • ARM versus x86
      • RISC versus CISC
      • Introduction to RISC-V
    • Networking and Connectivity
      • Infiniband versus Ethernet
      • NVIDIA Quantum InfiniBand
      • PCIe (Peripheral Component Interconnect Express)
      • NVIDIA ConnectX InfiniBand adapters
      • NVMe (Non-Volatile Memory Express)
      • NVMe over Fabrics (NVMe-oF)
      • NVIDIA Spectrum-X
      • NVIDIA GPUDirect
      • Evaluating Modern GPU Interconnect
      • Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)
      • Next-generation networking in AI environments
      • NVIDIA Collective Communications Library (NCCL)
    • Data and Memory
      • NVIDIA BlueField Data Processing Units (DPUs)
      • Remote Direct Memory Access (RDMA)
      • High Bandwidth Memory (HBM3)
      • Flash Memory
      • Model Requirements
      • Calculating GPU memory for serving LLMs
      • Transformer training costs
      • GPU Performance Optimisation
    • Libraries and Complements
      • NVIDIA Base Command
      • NVIDIA AI Enterprise
      • CUDA - NVIDIA GTC 2024 presentation
      • RAPIDs
      • RAFT
    • Vast Data Platform
      • Vast Datastore
      • Vast Database
      • Vast Data Engine
      • DASE (Disaggregated and Shared Everything)
      • Dremio and VAST Data
    • Storage
      • WEKA: A High-Performance Storage Solution for AI Workloads
      • Introduction to NVIDIA GPUDirect Storage (GDS)
        • GDS cuFile API
      • NVIDIA Magnum IO GPUDirect Storage (GDS)
      • Vectors in Memory
Powered by GitBook
LogoLogo

Continuum - Accelerated Artificial Intelligence

  • Continuum Website
  • Axolotl Platform

Copyright Continuum Labs - 2023

On this page
  • Architecture
  • Performance and Specifications

Was this helpful?

  1. Infrastructure
  2. Servers and Chips

NVIDIA DGX H-100 System

An absolute beast

PreviousNVIDIA DGX-2NextNVLink Switch

Last updated 11 months ago

Was this helpful?

DGX refers to a family of purpose-built AI servers.

DGX systems simplify the adoption and deployment of AI infrastructure by providing an integrated hardware and software platform.

GPU Server Options

They come with NVIDIA Base Command, a software suite that includes cluster management, job scheduling, and monitoring tools. This allows organisations to quickly set up and manage their AI infrastructure without the complexity of building and integrating individual components.

Architecture

The NVIDIA DGX H100 is built around eight powerful NVIDIA H100 Tensor Core GPUs, enabling it to deliver unparalleled performance for AI training, inference, and high-performance computing (HPC) workloads.

NVIDIA H100 Tensor Core GPUs

At the heart of the DGX H100 system are the eight H100 GPUs, each providing 80GB of high-bandwidth GPU memory, totalling 640GB across the system.

The H100 GPUs are based on the NVIDIA Hopper architecture, which introduces significant advancements over the previous generation.

With fourth-generation Tensor Cores, the H100 GPUs achieve an astonishing 32 petaFLOPS of FP8 performance, enabling breakthrough speed and efficiency for AI workloads.

Connecting the GPUs - NVLink and NVSwitch Interconnects

The eight H100 GPUs in the DGX H100 system are connected using NVIDIA's high-speed NVLink and NVSwitch interconnect technologies.

NVLink provides high-bandwidth, low-latency communication, allowing the GPUs to work together efficiently on large-scale tasks.

The system also features four NVIDIA NVSwitch interconnects, enabling flexible and scalable GPU-to-GPU communication, optimising performance, and supporting larger models and datasets.

High-Performance Networking

The DGX H100 system includes an impressive networking infrastructure, with eight single-port NVIDIA ConnectX-7 VPI adapters offering up to 400Gb/s of InfiniBand or Ethernet connectivity per port, and two dual-port NVIDIA ConnectX-7 VPI adapters for additional high-speed networking capabilities.

This networking setup allows DGX H100 systems to be interconnected to form larger, scalable AI clusters, such as NVIDIA DGX SuperPOD, to tackle the most demanding AI and HPC challenges.

CPU and System Memory

Complementing the GPU and networking capabilities, the DGX H100 system is equipped with dual Intel Xeon Platinum 8480C processors, providing a total of 112 CPU cores. These processors are high-performance server-grade CPUs.

The high core count allows for efficient parallel processing and multitasking, enabling the system to handle complex computational tasks alongside the GPUs.

The Xeon Platinum 8480C processors support Intel Deep Learning Boost (Intel DL Boost) technology, which includes vectorized neural network instructions (VNNI) that accelerate AI inferencing performance.

CPU Frequency

  • The Intel Xeon Platinum 8480C processors have a base frequency of 2.00 GHz and a maximum boost frequency of 3.80 GHz.

  • The base frequency refers to the default clock speed at which the CPU cores operate under normal conditions.

  • The maximum boost frequency is the highest clock speed that the CPU cores can reach when performing demanding tasks or when thermal conditions allow.

  • The high boost frequency of 3.80 GHz enables the CPUs to deliver strong single-thread performance, which is beneficial for certain workloads that rely on single-core speed.

System Memory

  • The DGX H100 system is equipped with 2TB of system memory, which is a substantial amount for handling large datasets and memory-intensive workloads.

  • The ample system memory allows for efficient data caching, reducing the need for frequent data transfers between storage and memory.

  • With 2TB of memory, the DGX H100 system can accommodate large AI models, datasets, and intermediate results during training and inference processes.

  • The high memory capacity also enables data sharing and collaboration between the CPUs and GPUs, minimising data transfer bottlenecks.

Storage and Management

  • The DGX H100 system includes two 1.92TB NVMe M.2 drives for the operating system and eight 3.84TB NVMe U.2 drives for internal storage.

  • The system also features a baseboard management controller (BMC) for remote management and monitoring.

Performance and Specifications

AI Performance

  • With eight H100 GPUs, the DGX H100 system achieves 32 petaFLOPS of FP8 performance, enabling faster training and inference of large-scale AI models.

  • The system's high-speed NVLink and NVSwitch interconnects ensure efficient communication and collaboration between the GPUs, maximising AI throughput.

Networking Performance

  • The DGX H100 system's high-performance networking capabilities, with up to 400Gb/s InfiniBand or Ethernet connectivity per port, enable fast data transfer and communication between systems.

  • This high-speed networking infrastructure allows DGX H100 systems to be scaled up to form larger AI clusters, such as NVIDIA DGX SuperPOD, for tackling the most demanding AI and HPC workloads.

System Specifications

  • The DGX H100 system has a maximum power usage of 10.2kW, ensuring ample power delivery for the high-performance components.

  • The system weighs 287.6lbs (130.45kgs) and has dimensions of 14.0in (356mm) height, 19.0in (482.2mm) width, and 35.3in (897.1mm) length.

  • The operating temperature range for the DGX H100 system is 5–30°C (41–86°F).

Keeping a rack of 5 DGX systems cool

Power Density and Cooling Challenges

  • power density in data centres has significantly increased over the past decade, with racks consuming up to 10 times more power than before.

  • Cooling such high-density racks with traditional air cooling methods is becoming increasingly challenging.

  • With 5 DGX H100 systems in a rack, each consuming up to 19.8 kW, the total power consumption could reach around 100 kW per rack. This high power density necessitates advanced cooling solutions like liquid cooling.

Liquid Cooling as a Solution

  • liquid cooling is emerging as a promising solution to address the cooling challenges posed by high-density racks.

  • various liquid cooling technologies, such as immersion cooling and cold plate cooling, can effectively remove heat from high-power components.

  • Liquid cooling allows for higher power densities compared to traditional air cooling methods, making it suitable for racks with multiple DGX H100 systems.

Infrastructure Considerations

  • Implementing liquid cooling in existing data centres requires careful planning and infrastructure modifications.

  • challenges such as integrating liquid cooling with existing pipework, ensuring proper leak detection and containment, and monitoring the cooling system through the building management system (BMS).

  • Retrofitting liquid cooling into legacy data centres may involve operational challenges and additional costs.

Standardization and Collaboration

  • The participants emphasize the need for standardization and collaboration among industry players to drive the adoption of liquid cooling.

  • They suggest that having common standards and best practices for liquid cooling implementation would facilitate its deployment and ensure compatibility across different systems.

  • Collaboration with cooling solution providers, such as Stulz, is seen as crucial in developing tailored and efficient liquid cooling solutions for high-density racks.

Future-Proofing and Scalability

  • importance of designing data centres with future requirements in mind, considering the rapidly increasing power densities driven by AI and other advanced workloads.

  • Adopting liquid cooling technology not only addresses the immediate cooling needs of high-density racks but also future-proofs the data centre infrastructure for potential expansions and technology advancements.

  • Modular and scalable liquid cooling solutions are discussed as potential approaches to accommodate growing power densities and enable efficient cooling in both new and existing data centres.

implementing liquid cooling for a rack with 5 DGX H100 systems appears to be a viable and necessary solution. The high power density of these systems requires advanced cooling methods to ensure optimal performance and reliability.

However, the decision to adopt liquid cooling should be made after careful evaluation of the existing data centre infrastructure, considering factors such as space constraints, piping layout, and compatibility with existing systems.

Engaging with liquid cooling solution providers and industry experts can help in designing a tailored and efficient liquid cooling system for your specific requirements.

Additionally, it's crucial to consider the long-term scalability and future-proofing aspects of the liquid cooling implementation. As power densities continue to rise, the chosen liquid cooling solution should be flexible enough to accommodate potential expansions and technological advancements in the future.

Microsoft Azure NVIDIA DGX H100 Installation
Page cover image