A100 SXM vs Tesla V100
Manufacturer specifications of the A100 SXM and the Tesla V100, side by side, per GPU. A value the manufacturer does not publish for the exact model is marked as such, never estimated.
NVIDIA A100 SXM 40GB
Ampere
- GPU memory
- 40 GB
- Memory bandwidth
- 1,555 GB/s
Live price on EverXCloud · from $0.34 per GPU-hour
Suited workloads
NVIDIA Tesla V100 32GB
Volta
- GPU memory
- 32 GB
- Memory bandwidth
- 900 GB/s
Live price on EverXCloud · from $0.10 per GPU-hour
Suited workloads
Key differences
- The A100 SXM has 1.3× the GPU memory of the Tesla V100 (40 GB vs 32 GB).
- The A100 SXM has 1.7× the memory bandwidth of the Tesla V100 (1,555 vs 900 GB/s), which usually sets the pace of LLM token generation.
Specifications
| Specification | A100 SXM | Tesla V100 |
|---|---|---|
| GPU memory | 40 GB (higher) | 32 GB |
| Memory type | HBM2 | HBM2 |
| Memory bandwidth | 1,555 GB/s (higher) | 900 GB/s |
| FP16 tensor | 312 TFLOPS | Not published |
| BF16 tensor | 312 TFLOPS | Not published |
| Max power | 400 W | Not published |
| Architecture | Ampere | Volta |
| Interconnect | NVLink 3 (600 GB/s) · PCIe 4.0 | Not published |
The higher value of each measured row is shown in bold. Manufacturer specifications, per GPU. Tensor throughput is shown without sparsity. Values the manufacturer does not publish for this exact model are left out.
What fits in memory, on one GPU
| Workload | A100 SXM | Tesla V100 |
|---|---|---|
| LLM inference, 16-bit | ≈ 15B parameters | ≈ 10B parameters |
| LLM inference, 4-bit | ≈ 50B parameters | ≈ 40B parameters |
| Full training, mixed precision | ≈ 2B parameters | ≈ 1B parameters |
Cautious estimates counting 80% of memory for model weights; context length, batch size and framework change the real limit. Several GPUs together hold larger models.
A100 SXM vs Tesla V100: frequently asked questions
Which has more memory, the A100 SXM or the Tesla V100?
The A100 SXM has 40 GB of GPU memory, against 32 GB for the Tesla V100.
Which is faster for LLM inference, the A100 SXM or the Tesla V100?
Generating tokens is usually bound by memory bandwidth, and the A100 SXM has 1.7× the bandwidth of the Tesla V100. Memory size then decides which models fit. Measure your own model before a large run.
Can I rent the A100 SXM and the Tesla V100 by the hour?
Yes, when sellers list them on the EverXCloud spot exchange: you buy GPU hours at the live price from a prepaid balance in euros, US dollars or pounds. Prices are shown excluding tax and EverXCloud is the merchant of record for every rental.
Be ready on launch day
Create your account now: you can top up your balance and rent the GPU that fits your workload as soon as the market opens.
Have GPU servers?
Businesses can join the EverXCloud Network and sell hours of their GPU servers on the spot exchange.