Head to head

H100 PCIe vs H100 SXM

H100 PCIe and H100 SXM side by side: published specs, and what each one actually rents for per GPU-hour on the clouds this index tracks.

Prices last collected 2026-10-26 06:00 UTC.

Quick verdict

H100 PCIe is currently about $0.7000 per GPU-hour cheaper than H100 SXM — roughly 26% less.

Specifications side by side

Published specifications for H100 PCIe and H100 SXM (vendor data)
Attribute H100 PCIe H100 SXM
Memory 80 GB 80 GB
Architecture Hopper Hopper
Memory bandwidth 2.0 TB/s 3.35 TB/s
Dense FP16 throughput 756 TFLOPS* 989 TFLOPS*

Vendor-published figures, with sparsity where the vendor states it. Always confirm on the official datasheet before sizing a cluster.

What actually differs

  • Both cards carry 80 GB, so memory size is not what separates them.
  • Higher memory bandwidth: H100 SXM at 3.35 TB/s.
  • H100 PCIe is Hopper-generation silicon, H100 SXM is Hopper.
  • Cheaper on the price table today: H100 PCIe from $1.9900 per GPU-hour.

More head-to-head comparisons

H100 SXM vs A100 SXM H100 SXM vs H200 SXM H100 SXM vs B200 H100 SXM vs L40S

Frequently asked questions

Everything about renting cloud GPUs, affiliate tracking and how we collect pricing data.

How is the cheaper provider decided?

For every GPU model both providers list, we compare the normalised single-GPU hourly price. The average of those unit prices is shown for each side, and the lower one is labelled cheaper. We deliberately exclude models only one of the two offers, because that is not a like-for-like comparison.

Does the cheaper option always make sense?

Not always. Marketplace capacity can be reclaimed at short notice and its hosts vary in network and disk performance, while a dedicated cloud gives more predictable throughput. For multi-node training, interconnect and NVLink support usually matter more than a few cents per hour.

Are spot prices directly comparable between providers?

Only roughly. Each platform defines interruption rules, notice periods and reclaim behaviour differently. Treat spot prices as an indication of the cheapest achievable rate, not as a guaranteed price.

Why do you show per-hour rather than per-month prices?

Most providers bill by the second or the hour, and per-hour figures are the only ones that can be compared without guessing how many hours you will actually run. Use the GPU cost estimator to convert an hourly rate into a monthly budget.

Volume buyers

B2B bulk GPU enquiry

Reserving large amounts of GPU capacity? Tell us what you need and we will forward your request to the right provider team.

  • Forwarded by email only — nothing is stored on this site
  • Typical first reply from a provider team: 1–2 business days
  • Useful for reserved clusters, multi-month commitments and data-residency requirements

Submitting this form sends an email to our inbox. It is not written to any database.