…
V100 SXM2 32GB vs A100 SXM 40GB
Cloud rental price, compute per dollar and specs, side by side.
Open in the interactive comparer →
V100 SXM2 32GB$0.15lowest on-demand per GPU·h
$1.19 per PFLOP·h · 2 providersA100 SXM 40GB$0.54lowest on-demand per GPU·h
$1.71 per PFLOP·h · 8 providersWhat actually differs
- The A100 SXM 40GB has 2.5× more dense FP16/BF16 throughput than the V100 SXM2 32GB (0.125 PF vs 0.312 PF).
- The A100 SXM 40GB has 1.7× more memory bandwidth than the V100 SXM2 32GB (900 vs 1,555 GB/s), which usually limits LLM token generation. Memory: 32 GB vs 40 GB.
- The V100 SXM2 32GB is 72% cheaper per hour than the A100 SXM 40GB at the lowest on-demand prices ($0.15 vs $0.54).
- Per unit of compute, the V100 SXM2 32GB is the better deal: $1.19 vs $1.71 per PFLOP-hour of dense FP16/BF16.
| Spec / price | V100 SXM2 32GB | A100 SXM 40GB |
|---|---|---|
| Memory | 32 GB HBM2 | 40 GB HBM2better |
| Memory bandwidth | 900 GB/s | 1,555 GB/sbetter |
| FP16 / BF16 (dense) | 0.125 PF | 0.312 PFbetter |
| FP8 (dense) | — | — |
| FP4 (dense) | — | — |
| INT8 (dense) | — | 0.624 PF |
| TF32 (dense) | — | 0.156 PF |
| FP32 (dense) | 0.0157 PF | 0.0195 PFbetter |
| FP64 (dense) | 0.0078 PF | 0.0195 PFbetter |
| Power | 300 W | 400 W |
| Architecture | Volta (2018) | Ampere (2020) |
| Lowest on-demand | $0.15better | $0.54 |
| Median on-demand | $2.03 | $1.99better |
| Lowest spot | $0.42better | $0.55 |
| FP16/BF16 $ per PFLOP·h | $1.19better | $1.71 |
| Providers | 2 | 8 |
V100 SXM2 32GB vs A100 SXM 40GB: questions
Is the V100 SXM2 32GB faster than the A100 SXM 40GB?
The A100 SXM 40GB has 2.5× more dense FP16/BF16 compute than the V100 SXM2 32GB. The A100 SXM 40GB has 1.7× more memory bandwidth than the V100 SXM2 32GB. 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 V100 SXM2 32GB or the A100 SXM 40GB?
Right now the V100 SXM2 32GB is cheaper per unit of compute: $1.19 vs $1.71 per PFLOP-hour at the lowest on-demand prices.
How much does it cost to rent an V100 SXM2 32GB vs an A100 SXM 40GB for a month?
Running one GPU around the clock at the lowest on-demand price costs about $109 per month for the V100 SXM2 32GB and $391 for the A100 SXM 40GB.
Which has more memory, the V100 SXM2 32GB or the A100 SXM 40GB?
The A100 SXM 40GB has more memory: 40 GB vs 32 GB.