NVIDIA A16 16GB vs NVIDIA A40 48GB

The A40 delivers 8.4x the FP16 throughput of the A16 (150 vs 18 TFLOPS dense). The A40 also carries 32 GB more memory (48 GB vs 16 GB).

A16: Ampere, 2021 A40: Ampere, 2021
FP16 dense lead
8.4x
A40: 150 vs 18 TFLOPS
Memory
16 vs 48 GB
32 GB more for the A40
Bandwidth
200 GB/s vs 696 GB/s
A40 moves data faster

Specifications

A16
A40
Architecture
AmpereAmpere
Launch Year
20212021
Form Factor
PCIePCIe
Memory
16 GB48 GB
Memory Bandwidth
200 GB/s696 GB/s
TDP
300 W
Process Node

Performance (TFLOPS)

A16
A40
FP64
No verified data available No verified data available
FP32
4.5 TFlops 37.4 TFlops
TF32
9.0 TFlops
18.0 TFLOPS sparse
74.8 TFlops
149.6 TFLOPS sparse
BF16
No verified data available149.7 TFlops
299.4 TFLOPS sparse
FP16
17.9 TFlops
35.9 TFLOPS sparse
149.7 TFlops
299.4 TFLOPS sparse
FP8
No verified data available No verified data available
FP6
No verified data available No verified data available
FP4
No verified data available No verified data available
INT8
35.9 TFlops
71.8 TOPS sparse
299.3 TFlops
598.6 TOPS sparse

FLOPS by Precision

What actually differs

Both GPUs launched in 2021: the A16 on Ampere and the A40 on Ampere.

Dense throughput for the A16 against the A40: FP32 4.5 vs 37 TFLOPS, FP16 18 vs 150 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.

Memory is 16 GB against 48 GB, fed at 200 GB/s versus 696 GB/s. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.

For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.

Frequently asked questions

Is the A16 faster than the A40?

At FP16 precision the A40 reaches 150 TFLOPS dense against 18 TFLOPS for the A16. The performance table on this page lists every published precision for both GPUs.

Which has more memory, the A16 or the A40?

The A40 carries 48 GB of memory versus 16 GB for the A16. Memory bandwidth is 200 GB/s for the A16 and 696 GB/s for the A40.

Can I rent the A16 or the A40 in the cloud?

Yes. Live cloud listings tracked by Flopper start at $0.35 per GPU hour. The rental pricing section on this page lists current providers and rates for both GPUs.

Where to Rent

NVIDIA A40 48GB

ProviderConfigurationPrice/GPU-hrChecked
RunPod
1× A40 48GB Community $0.35 5h ago View →
RunPod
1× A40 48GB $0.49 5h ago View →
All A40 listings and price history →

Get Comparison Updates

New GPUs added weekly. Be the first to see how they compare.