Blackwell
NVIDIA GB200 NVL72
wins
2
The GB200 NVL72 is NVIDIA's most powerful AI system.
Blackwell Ultra
NVIDIA GB300 NVL72
wins
2
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 GB200 NVL72
Metric
NVIDIA GB300 NVL72
FP16 Tensor TFLOPS (dense)
Bandwidth
VRAM
TDP (lower=better)
TFLOPS/watt
GB/s per watt
Full specification table
| Spec | NVIDIA GB200 NVL72 | NVIDIA GB300 NVL72 |
|---|---|---|
| Architecture | Blackwell | Blackwell Ultra |
| VRAM | 384 GB HBM3e (per GPU) | 288 GB HBM3e (per GPU) |
| VRAM type | HBM3e (per GPU) | HBM3e (per GPU) |
| Memory bandwidth | 16,000 GB/s | 16,000 GB/s |
| FP16 Tensor TFLOPS (dense) | 9,000 TFLOPS | 7,500 TFLOPS |
| TDP | 2,700 W | 1,200 W |
| TFLOPS/watt | 3.33 T/W | 6.25 T/W |
| GB/s per watt | 5.93 GB/s/W | 13.33 GB/s/W |
| NVLink | Yes | Yes |
| Release year | 2025 | 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 GB200 NVL72
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 GB200 NVL72 for:
- Trillion-parameter training
- AI factory scale
- Next-gen LLM inference
Weaknesses / Limitations
- Rack-scale system — not a single GPU, requires full NVL72 rack deployment
- 2,700W system TDP demands dedicated power and liquid cooling infrastructure
- Extremely limited availability — primarily hyperscaler and large enterprise
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