AICloudCompareHardware comparison / GPU versus GPU
H100 SXM vs H100 PCIe
Same H100 name, different products. The SXM module uses HBM3 with higher bandwidth; the PCIe card uses HBM2e. Offers for the two are not interchangeable, so compare like with like.
Understand the hardware trade-offs before comparing provider configurations. This page compares GPU models, not cloud SKUs or availability.
What the hardware records show
- Both use the NVIDIA Hopper architecture.
- H100 SXM uses the SXM form factor and H100 PCIe uses PCIe. SXM modules sit on server boards built for multi-GPU systems; PCIe cards fit a standard slot. They are separate products with separate offers.
- Both have 80 GB of memory per GPU.
- H100 SXM has 1.7× the theoretical memory bandwidth (3,350 against 2,000 GB/s).
Reviewed 11/10/2026 · verified
Reviewed 11/10/2026 · verified
It does not by itself prove a lower monthly cost, higher tokens per second or a better result. Include model memory, context, batch size and measured performance in your decision.
Where each GPU appears in our catalogue
Providers with at least one dated, published configuration of the exact GPU. Availability is not checked; the provider page shows each price with its review date.
H100 SXM
- DigitalOcean · 2 configurations
- Hyperstack · 1 configuration
- Lambda · 2 configurations
- Modal · 1 configuration
- Nebius · 1 configuration
- Together AI · 1 configuration