Ampere
NVIDIA A40
wins
1
The NVIDIA A40 offers 48GB GDDR6 at a lower price point than the A100.
Blackwell Ultra
NVIDIA GB300 NVL72
wins
4
The GB300 NVL72 is NVIDIA's Blackwell Ultra rack-scale system — 72 GPUs in a single rack delivering 7,500 TFLOPS FP16 per GPU.
Performance comparison
Visual bars — winner highlighted
NVIDIA A40
Metric
NVIDIA GB300 NVL72
FP16 Tensor TFLOPS (dense)
Bandwidth
VRAM
TDP (lower=better)
TFLOPS/watt
GB/s per watt
Full specification table
| Spec | NVIDIA A40 | NVIDIA GB300 NVL72 |
|---|---|---|
| Architecture | Ampere | Blackwell Ultra |
| VRAM | 48 GB GDDR6 | 288 GB HBM3e (per GPU) |
| VRAM type | GDDR6 | HBM3e (per GPU) |
| Memory bandwidth | 696 GB/s | 16,000 GB/s |
| FP16 Tensor TFLOPS (dense) | 149.7 TFLOPS | 7,500 TFLOPS |
| TDP | 300 W | 1,200 W |
| TFLOPS/watt | 0.50 T/W | 6.25 T/W |
| GB/s per watt | 2.32 GB/s/W | 13.33 GB/s/W |
| NVLink | No | Yes |
| Release year | 2020 | 2026 |
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 A40
NVIDIA GB300 NVL72
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 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
Choose NVIDIA GB300 NVL72 for:
- Next-gen frontier training
- Trillion-param inference
- Rack-scale AI
Weaknesses / Limitations
- Rack-scale system — not a single GPU, requires full NVL72 rack deployment
- Extremely high cost — rack-scale pricing only
- Very limited availability, primarily hyperscaler deployments
Popular GPU comparisons