Side-by-side GPU comparison: specs, memory, compute performance, and live cloud pricing.
Verdict
K80 has 4 GB more VRAM, making it better suited for large models and long context windows. For compute-bound workloads like training, K80 delivers 1.6× higher FP16 throughput. At $0.18/hr vs $0.60/hr, K80 is the more cost-efficient choice for inference. K80 supports a broader range of models (486 vs 443 from this catalog), giving more flexibility.
Specifications
| K80 | P4 | |
|---|---|---|
| VRAM | 12 GB | 8 GB |
| VRAM Type | GDDR5 | GDDR5 |
| Memory Bandwidth | 0.2 TB/s | 0.2 TB/s |
| FP16 Performance | 9 TFLOPS | 6 TFLOPS |
| Manufacturer | NVIDIA | NVIDIA |
| FP8 Support | No | No |
| FP4 Support | No | No |
Price / Performance
Based on cheapest single-GPU on-demand pricing. Lower $/TFLOP = better compute value; lower $/GB = better memory value.
| K80 | P4 | |
|---|---|---|
| $/hr (cheapest) | $0.18✓ best | $0.60 |
| $/TFLOP (compute value) | $0.0206✓ best | $0.1091 |
| $/GB VRAM (memory value) | $0.0150✓ best | $0.0750 |
Cloud Pricing
Cheapest on-demand price per provider (single GPU).
K80
| Provider | On-demand | Spot | Rent |
|---|---|---|---|
| Microsoft Azure | $0.18/hr | $0.10/hr |
P4
| Provider | On-demand | Spot | Rent |
|---|---|---|---|
| Google Cloud | $0.60/hr | $0.06/hr |
Model Compatibility
Models from the catalog that fit on each GPU, grouped by required precision.
K80 (486 models)
P4 (443 models)
You might also compare…
Pricing data refreshed hourly · Last updated April 11, 2026 · Browse all comparisons