Ampere
NVIDIA A30
wins
4
The NVIDIA A30 is a data center Ampere GPU with 24GB HBM2 and NVLink support.
Ampere
NVIDIA A40
wins
1
The NVIDIA A40 offers 48GB GDDR6 at a lower price point than the A100.
The central trade-off between the A40 and A30 is memory capacity: the A40 carries 48 GB GDDR6 while the A30 carries 24 GB HBM2. For training workloads, memory capacity determines whether a model fits at a given batch size and precision without offloading — the A40's 48 GB doubles the headroom for larger batches or longer sequences compared to the A30's 24 GB. Memory bandwidth and interconnect are separate considerations: the A30's HBM2 delivers higher bandwidth per GB than the A40's GDDR6, which benefits memory-bound inference workloads where bandwidth is the bottleneck rather than total capacity. Peak FP16 TFLOPS alone does not determine which GPU is faster for your workload — actual throughput depends on whether the workload is compute-bound or memory-bound, and on framework and driver support. Check the NVIDIA A30 specs page and NVIDIA A40 specs page for current cloud rental rates across providers.
Performance comparison
Visual bars — winner highlighted
NVIDIA A30
Metric
NVIDIA A40
FP16 TFLOPS
Bandwidth
VRAM
TDP (lower=better)
TFLOPS/watt
GB/s per watt
Full specification table
| Spec | NVIDIA A30 | NVIDIA A40 |
|---|---|---|
| Architecture | Ampere | Ampere |
| VRAM | 24 GB HBM2 | 48 GB GDDR6 |
| VRAM type | HBM2 | GDDR6 |
| Memory bandwidth | 933 GB/s | 696 GB/s |
| FP16 TFLOPS | 165 TFLOPS | 150 TFLOPS |
| TDP | 165 W | 300 W |
| TFLOPS/watt | 1.00 T/W | 0.50 T/W |
| GB/s per watt | 5.65 GB/s/W | 2.32 GB/s/W |
| NVLink | Yes | No |
| Release year | 2021 | 2020 |
Live cloud pricing
On-demand pricing across providers — updated in real time
Power efficiency analysis
TFLOPS/watt and GB/s/watt — critical for data center TCO
NVIDIA A30
NVIDIA A40
TFLOPS/watt measures compute efficiency — how much AI throughput you get per watt of power consumed. For data centers with PUE of 1.2–1.5, a 10% improvement in TFLOPS/watt translates directly to lower electricity costs and cooling requirements. GB/s/watt measures memory bandwidth efficiency, which is the binding constraint for memory-bound LLM inference workloads.
When to choose each GPU
Choose NVIDIA A30 for:
- Inference
- HPC
- Multi-instance GPU
Weaknesses / Limitations
- 24 GB VRAM limits model size
- Older Ampere architecture — lower TFLOPS than newer GPUs
- Being superseded by L4 and A10 on most inference platforms
Choose NVIDIA A40 for:
- Large model inference
- Professional rendering
- Multi-tenant AI
Weaknesses / Limitations
- Older Ampere architecture — lower FP16 TFLOPS than L40S at same VRAM
- GDDR6 bandwidth (696 GB/s) lower than HBM-based GPUs
- No NVLink on A40
Popular GPU comparisons