H200 NVL 141GB

The H200 NVL 141GB is a Hopper-generation NVIDIA GPU built on the GH100 architecture, manufactured on a TSMC 4N process node with 80 billion transistors. Released in 2024, it delivers 1979 TFLOPS of FP16 throughput and 1979 TFLOPS BF16 — the two precision formats most commonly used for transformer model training and inference. FP8 precision is supported at 3958 TFLOPS — roughly 2.0× the FP16 rate — enabling near-doubled throughput for inference workloads that can tolerate reduced numerical precision with calibration. The GH100 architecture represents NVIDIA's approach to balancing compute throughput, memory bandwidth, and power efficiency for data center AI workloads. At 600W TDP, the H200 NVL 141GB sits in the high-power data center tier (600W), compatible with standard high-density GPU server configurations.

Memory capacity is 141GB of HBM3e with 4800 GB/s bandwidth. This determines which models can run without quantization: approximately 70B parameters at FP16 (2 bytes/param), 141B at INT8 (1 byte/param), or up to 282B parameters at INT4/GGUF quantization (0.5 bytes/param). These figures are theoretical maximums — actual capacity is reduced by KV cache, framework overhead, and activation memory, typically by 10–20% for inference and 30–40% for training. The arithmetic intensity ceiling — the ratio of compute (1979 TFLOPS) to memory bandwidth (4800 GB/s) — is approximately 412 FLOP/byte. Transformer inference is typically memory-bound below this threshold, meaning the 4800 GB/s bandwidth figure is the primary determinant of tokens-per-second for most LLM serving workloads.

Multi-GPU configurations are a first-class use case for the H200 NVL 141GB. NVLink 4.0 / PCIe 5.0 provides 900 GB/s of bidirectional NVLink bandwidth between cards, enabling tensor-parallel inference across multiple GPUs with near-linear VRAM scaling. A two-card configuration provides 282GB of pooled VRAM — sufficient for 141B parameter models at FP16 — while a four-card setup reaches 564GB. NVLink's low-latency, high-bandwidth fabric makes all-reduce operations in data-parallel training significantly faster than PCIe-based alternatives, which top out at ~64 GB/s bidirectional for PCIe 5.0 x16.

The primary workloads for the H200 NVL 141GB are Large model inference, Multi-GPU NVLink clusters, Memory-bound serving. NVLink PCIe form factor H200. 141GB HBM3e with NVLink connectivity. Lower TDP than SXM variant, suitable for standard PCIe servers. Key limitations to factor into your evaluation: Premium pricing over H100 — ~40–60% higher hourly cost; Limited availability vs H100 — fewer providers offer it; Same compute as H100 — only memory bandwidth improvement. When comparing this GPU against alternatives at similar price points, the most important metrics are memory bandwidth (for inference throughput), VRAM capacity (for model size), and FP16/BF16 TFLOPS (for training speed). Raw TFLOPS figures can be misleading for inference — a GPU with lower TFLOPS but higher memory bandwidth will often outperform a higher-TFLOPS card on tokens-per-second for memory-bound autoregressive generation.

In the broader GPU market, the H200 NVL 141GB is a 2-year-old architecture that is still widely deployed in cloud data centers. CUDA compatibility is a significant advantage: the H200 NVL 141GB benefits from the full NVIDIA software ecosystem including cuDNN, TensorRT, FlashAttention, and all major inference frameworks (vLLM, TGI, TensorRT-LLM). CUDA's maturity means optimized kernels are available for virtually every model architecture. For cloud rental, availability varies significantly by provider — some specialize in this GPU tier while others may have limited stock. Compare on-demand and spot pricing across providers using the rental comparison table on this page, and factor in region availability if latency is a concern for your inference workload.