Compare · prices checked 2026-09-03
H100 SXM vs H100 PCIE: specs, price per hour, which to rent
NVIDIA H100 SXM (80 GB, $1.587/hr) against NVIDIA H100 PCIE (80 GB, $2.147/hr): public specs side by side, live fixed prices from our sheet, what fits in each card's VRAM, and a verdict written for real workloads — both are rentable right now.
H100 SXM vs H100 PCIE specifications
Public NVIDIA figures (dense, non-sparsity). The last column is H100 SXM relative to H100 PCIE.
| Spec | H100 SXM | H100 PCIE | Difference |
|---|---|---|---|
| Architecture | Hopper (2022) | Hopper (2022) | — |
| VRAM | 80 GB HBM3 | 80 GB HBM2e | same |
| Memory bandwidth | 3,350 GB/s | 2,000 GB/s | +68% |
| FP16 tensor (dense) | 990 TFLOPS | 756 TFLOPS | +31% |
| FP32 | 67.0 TFLOPS | 51.0 TFLOPS | +31% |
| CUDA cores | 16,896 | 14,592 | +16% |
| TDP | 700 W | 350 W | +100% |
| PCIe · NVLink | Gen 5.0 · NVLink | Gen 5.0 · no NVLink | — |
| PowerScore (RTX 3090 = 100) | 697 | 532 | +31% |
| Max GPUs per machine | 8× | 8× | — |
H100 SXM vs H100 PCIE price per hour
Fixed rates from our sheet — every on-demand price is the marketplace median × 0.70, rounded down. Monthly = 730 hours.
| Rate | H100 SXM | H100 PCIE | Cheaper |
|---|---|---|---|
| On-demand, per GPU-hour | $1.587 | $2.147 | H100 SXM (−26%) |
| Interruptible, per GPU-hour | $0.793 | $1.073 | H100 SXM |
| Reserved (3 mo), per GPU-hour | $1.031 | $1.395 | H100 SXM |
| On-demand, per month | $1,159 | $1,567 | H100 SXM |
| Market median (reference) | $2.27 | $3.07 | — |
| $ per 1,000 FP16 TFLOP-hours | $1.60 | $2.84 | H100 SXM (better value) |
| $ per GB of VRAM per hour | $0.0198 | $0.0268 | H100 SXM (better value) |
Try a full month with storage and bandwidth in the GPU cost calculator.
What fits in VRAM: 80 GB vs 80 GB
| Workload | H100 SXM | H100 PCIE |
|---|---|---|
| Largest LLM in FP16, one card | ~32B | ~32B |
| Largest LLM at 4-bit, one card | ~123B | ~123B |
| Flux dev (FP8, ~17 GB) | fits | fits |
| Wan 2.x 14B video (offloaded) | yes | yes |
| 70B 4-bit LLM on one card | yes | yes |
Rules of thumb: ~2.4 GB per billion parameters in FP16 all-in, ~0.62 GB in 4-bit. Full tables in the VRAM guide.
Verdict: which should you rent?
The same Hopper silicon in two packages: SXM brings 3.35 TB/s HBM3, 700 W and full NVLink; PCIe has 2 TB/s HBM2e and 350 W. Multi-GPU training wants SXM; single-GPU fine-tuning and inference get most of the value from the PCIe card at a lower rate.
- Cheaper per hour: H100 SXM ($1.587 vs $2.147, −26%).
- More VRAM: H100 SXM (80 GB vs 80 GB).
- More FP16 throughput: H100 SXM (about 1.3×).
- Best value per TFLOP-hour: H100 SXM.
- Best value per GB of VRAM: H100 SXM.
- Multi-GPU: H100 SXM with NVLink · H100 PCIE over PCIe.
Related comparisons
- H100 SXM vs A100 SXM4 $1.587 vs $0.583 per hour Compare
- H100 SXM vs H200 $1.587 vs $3.058 per hour Compare
- H100 SXM vs B200 $1.587 vs $4.204 per hour Compare
- H100 PCIE vs A100 PCIE $2.147 vs $0.662 per hour Compare
- L40S vs H100 PCIE $0.466 vs $2.147 per hour Compare
- RTX PRO 6000 WS vs H100 PCIE $1.097 vs $2.147 per hour Compare
H100 SXM vs H100 PCIE: FAQ
Deeper reading: H100 vs H200 vs B200, RTX 4090 vs RTX 5090, how cloud GPU pricing works.
Is the H100 SXM faster than the H100 PCIE?
On dense FP16 tensor throughput the H100 SXM leads by about 1.3× (990 vs 756 TFLOPS). Memory bandwidth matters as much for inference: H100 SXM 3,350 GB/s vs H100 PCIE 2,000 GB/s.
Which is cheaper to rent, the H100 SXM or the H100 PCIE?
The H100 SXM: $1.587/hr on-demand versus $2.147/hr — 26% less. Interruptible rates are $0.793 (H100 SXM) and $1.073 (H100 PCIE). Per TFLOP-hour the better value is the H100 SXM.
Which has more VRAM and what does that change?
The H100 SXM has 80 GB versus 80 GB. In LLM terms that is roughly a 32B FP16 model (or ~123B in 4-bit) on one card against 32B FP16 (~123B 4-bit). If the model does not fit, speed is irrelevant.
Can I rent both on PowerGPU right now?
Yes — 69 × H100 SXM and 60 × H100 PCIE are online as this page renders, deployable in about 30 seconds, billed per second, paid in crypto with no KYC.