GPUs & Accelerators
NVIDIA H100 NVL
A PCIe-based H100 NVL data-center accelerator for large language model inference, enterprise AI servers, and multi-GPU private AI infrastructure.
Vendor
NVIDIA
Price
Request quote
Quote required
Quick Summary
Why it matters
Specs
Simplified decision specs
GPU Memory
94GB HBM3 per GPU / 188GB with NVLink pair
AI / HPC Performance
3,341 FP8 TFLOPS / 835 TF32 Tensor TFLOPS
PCIe Server Profile
3.9 TB/s bandwidth / 350-400W TDP / dual-slot air-cooled
Best For
Where this product fits best
Used In
Example deployment contexts
Black Scarab POV
Where we see the real value
This is the H100 to list first because the NVL/PCIe form factor is easier to explain as a component than H100 SXM. It still belongs in certified server baskets with serious airflow, power, software, and support planning, not in a normal desktop workstation.
Category
GPUs & Accelerators
Power Profile
High (200W+)
Deployment Types
Indoor, Industrial
Vendor
NVIDIA
Price
Request quote
Quote required
Source
Official product page
Link Strategy
This catalog starts with official product and distributor links. Affiliate links can be added later without changing the page design; once a product gets an affiliate URL, the main outbound button can automatically use that tracked commercial link.