Side-by-side GPU comparison: specs, memory, compute performance, and live cloud pricing.
Verdict
A10 has 12 GB more VRAM, making it better suited for large models and long context windows. For compute-bound workloads like training, A10 delivers 7.2× higher FP16 throughput. At $0.18/hr vs $1.29/hr, K80 is the more cost-efficient choice for inference. A10 supports a broader range of models (566 vs 486 from this catalog), giving more flexibility.
Specifications
| A10 | K80 | |
|---|---|---|
| VRAM | 24 GB | 12 GB |
| VRAM Type | GDDR6 | GDDR5 |
| Memory Bandwidth | 0.6 TB/s | 0.2 TB/s |
| FP16 Performance | 63 TFLOPS | 9 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.
| A10 | K80 | |
|---|---|---|
| $/hr (cheapest) | $1.29 | $0.18✓ best |
| $/TFLOP (compute value) | $0.0206 | $0.0206✓ best |
| $/GB VRAM (memory value) | $0.0537 | $0.0150✓ best |
Cloud Pricing
Cheapest on-demand price per provider (single GPU).
A10
| Provider | On-demand | Spot | Rent |
|---|---|---|---|
| Lambda | $1.29/hr | — | |
| Microsoft Azure | $3.20/hr | $0.80/hr |
K80
| Provider | On-demand | Spot | Rent |
|---|---|---|---|
| Microsoft Azure | $0.18/hr | $0.10/hr |
Model Compatibility
Models from the catalog that fit on each GPU, grouped by required precision.
A10 (566 models)
K80 (486 models)
You might also compare…
Pricing data refreshed hourly · Last updated April 11, 2026 · Browse all comparisons