H100 PCIe 80GB

The H100 PCIe 80GB is a Hopper-generation NVIDIA GPU built on the GH100 architecture, manufactured on a TSMC 4N process node with 80 billion transistors. Released in 2022, it delivers 1513 TFLOPS of FP16 throughput and 1513 TFLOPS BF16 — the two precision formats most commonly used for transformer model training and inference. FP8 precision is supported at 3026 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 350W TDP, the H100 PCIe 80GB sits in the mid-range data center tier (350W), fitting standard GPU server form factors.

Memory capacity is 80GB of HBM3 with 2000 GB/s bandwidth. This determines which models can run without quantization: approximately 40B parameters at FP16 (2 bytes/param), 80B at INT8 (1 byte/param), or up to 160B 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 (1513 TFLOPS) to memory bandwidth (2000 GB/s) — is approximately 757 FLOP/byte. Transformer inference is typically memory-bound below this threshold, meaning the 2000 GB/s bandwidth figure is the primary determinant of tokens-per-second for most LLM serving workloads.

The H100 PCIe 80GB uses PCIe 5.0 for host connectivity. Without NVLink, VRAM cannot be pooled across multiple cards — the single-card 80GB capacity is the hard ceiling for model size without model sharding over slower PCIe. For workloads that exceed 80GB, the alternative is pipeline parallelism (splitting model layers across cards) rather than tensor parallelism, which introduces inter-card communication overhead at each layer boundary. This makes the H100 PCIe 80GB best suited for workloads that fit within a single card's VRAM budget.

The primary workloads for the H100 PCIe 80GB are Inference serving, Cost-efficient training, PCIe server deployments. PCIe form factor H100. Lower TDP and cost than SXM5. No NVLink. Ideal for inference-heavy workloads where SXM bandwidth is not needed. Key limitations to factor into your evaluation: Only 40GB VRAM — limits to ~30B models at FP16; Lower bandwidth than SXM5 variant (2 TB/s vs 3.35 TB/s); Higher cost than A100 for inference-only workloads. 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 H100 PCIe 80GB is a 4-year-old design that is increasingly being replaced by newer architectures in cloud deployments, though it remains available at competitive rental rates. CUDA compatibility is a significant advantage: the H100 PCIe 80GB 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.