NVIDIA A40
The NVIDIA A40 is a Data Center graphics card based on the Ampere architecture. Released on 2020-10-05 with a launch price of $6,000.00, it features 10,752 CUDA cores and 48 GB of GDDR6 memory.
General info about A40
| Architecture | Ampere (GA102) |
| Market Segment | Data Center |
| Release Date | 2020-10-05 |
| Launch Price | $6,000.00 |
| Manufacturing Process | 8nm |
Technical specs of A40
| CUDA Cores | 10,752 |
| Base Clock Speed | 1305 MHz |
| Boost Clock Speed | 1740 MHz |
| Transistor Count | 28.3B |
| VRAM Capacity | 48 GB GDDR6 |
| Memory Bus Width | 384 bits |
| Memory Bandwidth | 696.0 GB/s |
| TDP | 300 W |
Key Technical Parameters
Key technical parameters of the NVIDIA A40 include its 10,752 CUDA cores, 1305 MHz base clock, and 1740 MHz boost clock, delivering high performance for data center applications.
Benchmark performance of A40
FP32 Performance
37.4
TFLOPS
FP16 Performance
74.8
TFLOPS
The NVIDIA A40 achieves 37.4 TFLOPS in FP32 performance, making it ideal for data center workloads.
Gaming performance of A40
Gaming Performance
The NVIDIA A40 is not designed for gaming. It is optimized for AI training, machine learning, and high-performance computing workloads.
Features and connectivity of A40
Tensor Cores
Gen 3
Transformer Engine
Disabled
Flash Attention
Not Supported
Connectivity
NVLink 3, PCIe Gen4
Physical dimensions of A40
| Length | 10.5in |
| Width | 4.4in |
| Height | 2-slot |
Rankings and efficiency of A40
| Ranking Position | #8 |
| Popularity Ranking | #12 |
| Cost Effectiveness | 1.3/5 |
| Power Efficiency | 1.4/5 |
Cloud Availability
Available Instances
1
Starting Price
$0.44/hr
Top GPUs for Training and Inference
| Category | Rank 1 | Rank 2 | Rank 3 |
|---|---|---|---|
| Best for Training | NVIDIA H200 | NVIDIA H100 | NVIDIA B200 |
| Best for Inference | NVIDIA A40 | NVIDIA A100 | NVIDIA A10 |
Compare GPU specifications and cloud instances to find the best GPU for your workload.