H200 NVL: 141 GB HBM3e for 70B+ parameter models on a single card
PNY NVIDIA H200 NVL is a compute accelerator on the Hopper architecture with 141 GB of HBM3e memory and 4.8 TB/s of bandwidth. The capacity is enough to keep a large language model entirely in the memory of one card without sharding - inference and fine-tuning of LLMs with 70B+ parameters run without offloading weights to system RAM.
14,592 CUDA cores and 456 4th generation Tensor cores with Transformer Engine and FP8 support deliver up to 3,341 TFLOPS in FP8 (with sparsity) and 1,671 TFLOPS in BFLOAT16/FP16. For scaling there is an NVLink bridge for 2 or 4 cards with 900 GB/s of bandwidth per GPU - this removes the inter-card exchange bottleneck in generative AI and HPC clusters.
- Memory: 141 GB HBM3e, 6016-bit bus, 4.8 TB/s, ECC
- Interface: PCIe 5.0 x16, NVLink 900 GB/s (2/4-way)
- Form factor: dual-slot FHFL ~267 mm, passive cooling for a server chassis
- Power: configurable TGP up to 600 W
- MIG: up to 7 isolated instances of ~16.5 GB
- Included: 5-year NVIDIA AI Enterprise subscription
The card is designed for NVIDIA-Certified servers with up to 8 GPUs, supports Confidential Computing and has 7 NVDEC decoders and 7 JPEG engines. This is a retail supply in PNY branded packaging with a full warranty from the official distributor. Selection of a multi-GPU configuration for your model - at +380 93,594 00 77.
Frequently Asked Questions
What power supply does this card need?
Auxiliary power connector: 1x16-pin. With two of these cards in one system, size the supply with headroom and check the circuit rating.
What model size fits into 141 GB of memory?
Roughly a 70B model at 8-bit or a 120B model at 4-bit, with a long context window. The exact figure depends on context length: longer conversations consume more cache. Memory: 141 GB HBM3e.
Which GPU does this card use?
NVIDIA H200. The GPU generation decides which compute formats are accelerated in hardware, and that affects inference speed more than raw memory capacity does.
What warranty is provided and what is in the box?
Warranty: 36 months. Ships in the manufacturer's original packaging. Ask us to confirm the bundle contents before ordering.
Not sure this card fits your system?
We will check compatibility with your platform, calculate the power draw for your configuration and tell you whether the memory is enough for your models.
The model you need is out of stock? ETE is listed among official PNY partners, so we bring professional NVIDIA cards to order with the manufacturer warranty.
Message us on Telegram or leave a request.
Read the review and measurements: NVIDIA H200 NVL: 141 GB of HBM3e — review and measurements →
