…
H100 SXM vs A100 SXM 80GB
Cloud rental price, compute per dollar and specs, side by side.
Open in the interactive comparer →
H100 SXM$1.66lowest on-demand per GPU·h
$1.68 per PFLOP·h · 28 providersA100 SXM 80GB$0.81lowest on-demand per GPU·h
$2.60 per PFLOP·h · 17 providersWhat actually differs
- The H100 SXM has 3.2× more dense FP16/BF16 throughput than the A100 SXM 80GB (0.989 PF vs 0.312 PF).
- The H100 SXM has 1.6× more memory bandwidth than the A100 SXM 80GB (3,350 vs 2,039 GB/s), which usually limits LLM token generation. Memory: 80 GB vs 80 GB.
- Only the H100 SXM has native FP8 tensor cores.
- The A100 SXM 80GB is 51% cheaper per hour than the H100 SXM at the lowest on-demand prices ($1.66 vs $0.81).
- Per unit of compute, the H100 SXM is the better deal: $1.68 vs $2.60 per PFLOP-hour of dense FP16/BF16.
| Spec / price | H100 SXM | A100 SXM 80GB |
|---|---|---|
| Memory | 80 GB HBM3 | 80 GB HBM2e |
| Memory bandwidth | 3,350 GB/sbetter | 2,039 GB/s |
| FP16 / BF16 (dense) | 0.989 PFbetter | 0.312 PF |
| FP8 (dense) | 1.98 PF | — |
| FP4 (dense) | — | — |
| INT8 (dense) | 1.98 PFbetter | 0.624 PF |
| TF32 (dense) | 0.495 PFbetter | 0.156 PF |
| FP32 (dense) | 0.067 PFbetter | 0.0195 PF |
| FP64 (dense) | 0.067 PFbetter | 0.0195 PF |
| Power | 700 W | 400 W |
| Architecture | Hopper (2022) | Ampere (2020) |
| Lowest on-demand | $1.66 | $0.81better |
| Median on-demand | $3.19 | $2.35better |
| Lowest spot | $0.79better | $0.90 |
| FP16/BF16 $ per PFLOP·h | $1.68better | $2.60 |
| Providers | 28 | 17 |
H100 SXM vs A100 SXM 80GB: questions
Is the H100 SXM faster than the A100 SXM 80GB?
The H100 SXM has 3.2× more dense FP16/BF16 compute than the A100 SXM 80GB. The H100 SXM has 1.6× more memory bandwidth than the A100 SXM 80GB. Real-world speedups depend on the workload: training and prefill scale with compute, while LLM decoding scales mostly with memory bandwidth.
Which is cheaper per FLOP, the H100 SXM or the A100 SXM 80GB?
Right now the H100 SXM is cheaper per unit of compute: $1.68 vs $2.60 per PFLOP-hour at the lowest on-demand prices.
How much does it cost to rent an H100 SXM vs an A100 SXM 80GB for a month?
Running one GPU around the clock at the lowest on-demand price costs about $1,212 per month for the H100 SXM and $591 for the A100 SXM 80GB.
Which has more memory, the H100 SXM or the A100 SXM 80GB?
Both have 80 GB of VRAM; the H100 SXM has more bandwidth.