H100 SXM
training · high-throughput inference
$1,352/ 30 days
View GPU H100 SXM vs L4
Compare memory, node configurations, and fixed weekly and monthly cost before choosing between NVIDIA H100 SXM and NVIDIA L4.
H100 SXM provides 56 GB more VRAM per card. L4 costs $1,152 less per 30-day term . Choose based on whether memory headroom or lower fixed cost is the binding constraint.
training · high-throughput inference
efficient inference · video encoding
| Specification | H100 SXM | L4 |
|---|---|---|
| Architecture | Hopper | Ada |
| VRAM | 80 GB HBM3 | 24 GB GDDR6 ECC |
| Host resources / card | 24 vCPU · 192 GB RAM | 8 vCPU · 64 GB RAM |
| Local NVMe / card | 3.84 TB | 1 TB |
| 7-day price / card | $380 | $56 |
| 30-day price / card | $1,352 | $200 |
| 4-GPU node / 30 days | $4,975 | $736 |
| Catalog quantity | 690 cards | 315 cards |